US20140074466A1

US20140074466A1 - Answering questions using environmental context

Info

Publication number: US20140074466A1
Application number: US13/626,439
Authority: US
Inventors: Matthew Sharifi; Gheorghe Postelnicu
Original assignee: Google LLC
Current assignee: Google LLC
Priority date: 2012-09-10
Filing date: 2012-09-25
Publication date: 2014-03-13
Also published as: US20160343371A1; US9786279B2; US20170133014A1; US9576576B2

Abstract

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving audio data encoding an utterance and environmental data, obtaining a transcription of the utterance, identifying an entity using the environmental data, submitting a query to a natural language query processing engine, wherein the query includes at least a portion of the transcription and data that identifies the entity, and obtaining one or more results of the query.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

This application claims the benefit of U.S. Provisional Patent Application No. 61/698,934, filed Sep. 10, 2012, the entire contents of the previous application is hereby incorporated by reference.

FIELD

The present specification relates to identifying results of a query based on a natural language query and environmental information, for example to answer questions using environmental information as context.

BACKGROUND

In general a search query includes one or more terms that a user submits to a search engine when the user requests the search engine to execute a search. Among other approaches, a user may enter query terms of a search query by typing on a keyboard or, in the context of a voice query, by speaking the query terms into a microphone of a mobile device. Voice queries may be processed using speech recognition technology.

SUMMARY

According to some innovative aspects of the subject matter described in this specification, environmental information, such as ambient noise, may aid a query processing system in answering a natural language query. For example, a user may ask a question about a television program that they are viewing, such as “What actor is in this movie?” The user's mobile device detects the user's utterance and environmental data, which may include the soundtrack audio of the television program. The mobile computing device encodes the utterance and the environmental data as waveform data, and provides the waveform data to a server-based computing environment.
The computing environment separates the utterance from the environmental data of the waveform data, and then obtains a transcription of the utterance. The computing environment further identifies entity data relating to the environmental data and the utterance, such as by identifying the name of the movie. From the transcription and the entity data, the computing environment can then identify one or more results, for example, results in response to the user's question. Specifically, the one or more results can include an answer to the user's question of “What actor is in this movie?” (e.g., the name of the actor). The computing environment can provide such results to the user of the mobile computing device.
Innovative aspects of the subject matter described in this specification may be embodied in methods that include the actions of receiving audio data encoding an utterance and environmental data, obtaining a transcription of the utterance, identifying an entity using the environmental data, submitting a query to a natural language query processing engine, wherein the query includes at least a portion of the transcription and data that identifies the entity, and obtaining one or more results of the query.
Other embodiments of these aspects include corresponding systems, apparatus, and computer programs, configured to perform the actions of the methods, encoded on computer storage devices.
These and other embodiments may each optionally include one or more of the following features. For instance, outputting a representation of at least one of the results. The entity is identified further using the utterance. Generating the query. Generating the query includes associating the transcription with the data that identifies the entity. Associating further includes tagging the transcription with the data that identifies the entity. Associating further includes substituting a portion of the transcription with the data that identifies the entity. Substituting further includes substituting one or more words of the transcription with the data that identifies the entity. Receiving the environmental data further includes receiving environmental audio data, environmental image data, or both. Receiving the environmental audio data further includes receiving additional audio data that includes background noise.
The details of one or more embodiments of the subject matter described in this specification are set forth in the accompanying drawings and the description below. Other potential features, aspects, and advantages of the subject matter will become apparent from the description, the drawings, and the claims.
DESCRIPTION OF DRAWINGS
FIG. 1 depicts an example system for identifying content item data based on environmental audio data and a spoken natural language query.
FIG. 2 depicts a flowchart for an example process for identifying content item data based on environmental audio data and a spoken natural language query.
FIGS. 3A-3B depicts portions of an example system for identifying content item.
FIG. 4 depicts an example system for identifying media content items based on environmental image data and a spoken natural language query.
FIG. 5 depicts a system for identifying one or more results based on environmental audio data and an utterance.
FIG. 6 depicts a flowchart for an example process for identifying one or more results based on environmental data and an utterance.
FIG. 7 depicts a computer device and a mobile computer device that may be used to implement the techniques described here.
Like reference symbols in the various drawings indicate like elements.

DETAILED DESCRIPTION

A computing environment that answers spoken natural language queries using environmental information as context may process queries using multiple processes. In an example of some processes, illustrated in FIGS. 1 to 4, the computing environment can identify media content based on environmental information, such as ambient noises. In an example of other processes, illustrated in FIGS. 5 and 6, a computing environment can augment the spoken natural language query with context that is derived from the environmental information, such as data that identifies media content, in order to provide a more satisfying answer to the spoken natural language query.
In more detail, FIG. 1 depicts a system 100 for identifying content item data based on environmental audio data and a spoken natural language query. Briefly, the system 100 can identify content item data that is based on the environmental audio data and that matches a particular content type associated with the spoken natural language query. The system 100 includes a mobile computing device 102, a disambiguation engine 104, a speech recognition engine 106, a keyword mapping engine 108, and a content recognition engine 110. The mobile computing device 102 is in communication with the disambiguation engine 104 over one or more networks. The mobile device 110 can include a microphone, a camera, or other detection mechanisms for detecting utterances from a user 112 and/or environmental data associated with the user 112.
In some examples, the user 112 is watching a television program. In the illustrated example, the user 112 would like to know who directed the television program that is currently playing. In some examples, the user 112 may not know the name of the television program that is currently playing, and may therefore ask the question “Who directed this show?” The mobile computing device 102 detects this utterance, as well as environmental audio data associated with the environment of the user 112.
In some examples, the environmental audio data associated with the environment of the user 112 can include background noise of the environment of the user 112. For example, the environmental audio data includes the sounds of the television program. In some examples, the environmental audio data that is associated with the currently displayed television program can include audio of the currently displayed television program (e.g., dialogue of the currently displayed television program, soundtrack audio associated with the currently displayed television program, etc.).
In some examples, the mobile computing device 102 detects the environmental audio data after detecting the utterance; detects the environmental audio data concurrently with detecting the utterance; or both. The mobile computing device 102 processes the detected utterance and the environmental audio data to generate waveform data 114 that represents the detected utterance and the environmental audio data and transmits the waveform data 114 to the disambiguation engine 104 (e.g., over a network), during operation (A). In some examples, the environmental audio data is streamed from the mobile computing device 110.
The disambiguation engine 104 receives the waveform data 114 from the mobile computing device 102. The disambiguation engine 104 processes the waveform data 114, including separating (or extracting) the utterance from other portions of the waveform data 114 and transmits the utterance to the speech recognition engine 106 (e.g., over a network), during operation (B). For example, the disambiguation engine 104 separates the utterance (“Who directed this show?”) from the background noise of the environment of the user 112 (e.g., audio of the currently displayed television program).
In some examples, the disambiguation engine 104 utilizes a voice detector to facilitate separation of the utterance from the background noise by identifying a portion of the waveform data 114 that includes voice activity, or voice activity associated with the user of the computing device 102. In some examples, the utterance relates to a query (e.g., a query relating to the currently displayed television program). In some examples, the waveform data 114 includes the detected utterance. In response, the disambiguation engine 104 can request the environmental audio data from the mobile computing device 102 relating to the utterance.
The speech recognition engine 106 receives the portion of the waveform data 114 that corresponds to the utterance from the disambiguation engine 104. The speech recognition engine 106 obtains a transcription of the utterance and provides the transcription to the keyword mapping engine 108, during operation (C). Specifically, the speech recognition engine 106 processes the utterance received from the speech recognition engine 106. In some examples, processing of the utterance by the speech recognition system 106 includes generating a transcription of the utterance. Generating the transcription of the utterance can include transcribing the utterance into text or text-related data. In other words, the speech recognition system 106 can provide a representation of language in written form of the utterance.
For example, the speech recognition system 106 transcribes the utterance to generate the transcription of “Who directed this show?” In some embodiments, the speech recognition system 106 provides two or more transcriptions of the utterance. For example, the speech recognition system 106 transcribes the utterance to generate the transcriptions of “Who directed this show?” and “Who directed this shoe?”
The keyword mapping engine 108 receives the transcription from the speech recognition engine 106. The keyword mapping engine 108 identifies one or more keywords in the transcription that are associated with a particular content type and provides the particular content type to the disambiguation engine 104, during operation (D). In some embodiments, the one or more content types can include ‘movie’, ‘music’, ‘television show’, ‘audio podcast’, ‘mage,’ ‘artwork,’ ‘book,’ ‘magazine,’ ‘trailer,’ ‘video podcast’, ‘Internet video’, or ‘video game’.
For example, the keyword mapping engine 108 identifies the keyword “directed” from the transcription of “Who directed this show?” The keyword “directed” is associated with the ‘television show’ content type. In some embodiments, a keyword of the transcription that is identified by the keyword mapping engine 108 is associated with two or more content types. For example, the keyword “directed” is associated with the ‘television show’ and ‘movie’ content types.
In some embodiments, the keyword mapping engine 108 identifies two or more keywords in the transcription that are associated with a particular content type. For example, the keyword mapping engines 108 identifies the keywords “directed” and “show” that are associated with a particular content type. In some embodiments, the identified two or more keywords are associated with the same content type. For example, the identified keywords “directed” and “show” are both associated with the ‘television show’ content type. In some embodiments, the identified two or more keywords are associated with differing content types. For example, the identified keyword “directed” is associated with the ‘movie’ content type and the identified keyword “show” is associated with the ‘television show’ content type. The keyword mapping engine 108 transmits (e.g., over a network) the particular content type to the disambiguation engine 108.
In some embodiments, the keyword mapping engine 108 identifies the one or more keywords in the transcription that are associated with a particular content type using one or more databases that, for each of multiple content types, maps at least one of the keywords to at least one of the multiple content types. Specifically, the keyword mapping engine 108 includes (or is in communication with) a database (or multiple databases). The database includes, or is associated with, a mapping between keywords and content types. Specifically, the database provides a connection (e.g., mapping) between the keywords and the content types such that the keyword mapping engine 108 is able to identify one or more keywords in the transcription that are associated with particular content types.
In some embodiments, one or more of the mappings between the keywords and the content types can include a unidirectional (e.g., one-way) mapping (i.e., a mapping from the keywords to the content types). In some embodiments, one or more of the mappings between the keywords and the content types can include a bidirectional (e.g., two-way) mapping (i.e., a mapping from the keywords to the content types and from the content types to the keywords). In some embodiments, the one or more databases maps one or more of the keywords to two or more content types.
For example, the keyword mapping engine 108 uses the one or more databases that maps the keyword “directed” to the ‘movie’ and ‘television show’ content types. In some embodiments, the mapping between the keywords and the content types can include mappings between multiple, varying versions of a root keyword (e.g., the word family) and the content types. The differing versions of the keyword can include differing grammatical categories such as tense (e.g., past, present, future) and word class (e.g., noun, verb). For example, the database can include mappings of the word family of the root word “direct” such as “directors,” “direction,” and “directed” to the one or more content types.
The disambiguation engine 104 receives data identifying the particular content type associated with the transcription of the utterance from the keyword mapping engine 108. Furthermore, as mentioned above, the disambiguation engine 104 receives the waveform data 114 from the mobile computing device 102 that includes the environmental audio data associated with the utterance. The disambiguation engine 104 then provides the environmental audio data and the particular content type to the content recognition engine 110, during operation (E).
For example, the disambiguation engine 104 transmits the environmental audio data relating to the currently displayed television program that includes audio of the currently displayed television program (e.g., dialogue of the currently displayed television program, soundtrack audio associated with the currently displayed television program, etc.) and the particular content type of the transcription of the utterance (e.g., ‘television show’ content type) to the content recognition engine 110.
In some embodiments, the disambiguation engine 104 provides a portion of the environmental audio data to the content recognition engine 110. In some examples, the portion of the environmental audio data can include background noise detected by the mobile computing device 102 after detecting the utterance. In some examples, the portion of the environmental audio data can include background noise detected by the mobile computing device 102 concurrently with detecting the utterance.
In some embodiments, the background noise (of the waveform data 114) is associated with a particular content type that is associated with a keyword of the transcription. For example, the keyword “directed” of the transcription “Who directed this show?” is associated with the ‘television show’ content type, and the background noise (e.g., the environmental audio data relating to the currently displayed television program) is also associated with the ‘television show’ content type.
The content recognition engine 110 receives the environmental audio data and the particular content type from the disambiguation engine 104. The content recognition engine 110 identifies content item data that is based on the environmental audio data and that matches the particular content type and provides the content item data to the disambiguation engine 104, during operation (F). Specifically, the content recognition engine 110 appropriately processes the environmental audio data to identify content item data that is associated with the environmental audio data (e.g., a name of a television show, a name of a song, etc.). Additionally, the content recognition engine 110 matches the identified content item data with the particular content type (e.g., content type of the transcription of the utterance). The content recognition engine 110 transmits (e.g., over a network) the identified content item data to the disambiguation engine 104.
For example, the content recognition engine 110 identifies content item data that is based on the environmental audio data relating to the currently displayed television program, and further that matches the ‘television show’ content type. To that end, the content recognition engine 110 can identify content item data based on dialogue of the currently displayed television program, or soundtrack audio associated with the currently displayed television program, depending on the portion of the environmental audio data received by the content recognition engine 110.
In some embodiments, the content recognition engine 110 is an audio fingerprinting engine that utilizes content fingerprinting using wavelets to identify the content item data. Specifically, the content recognition engine 110 converts the waveform data 114 into a spectrogram. From the spectrogram, the content recognition engine 110 extracts spectral images. The spectral images can be represented as wavelets. For each of the spectral images that are extracted from the spectrogram, the content recognition engine 110 extracts the “top” wavelets based on the respective magnitudes of the wavelets. For each spectral image, the content recognition engine 110 computes a wavelet signature of the image. In some examples, the wavelet signatures is a truncated, quantized version of the wavelet decomposition of the image.
For example, to describe an m×n image with wavelets, m×n wavelets are returned without compression. Additionally, the content recognition engine 110 utilizes a subset of the wavelets that most characterize the song. Specifically, the t “top” wavelets (by magnitude) are selected, where t<<m×n. Furthermore, the content recognition engine 110 creates a compact representation of the sparse wavelet-vector described above, for example, using MinHash to compute sub-fingerprints for these sparse bit vectors.
In some examples, when the environmental audio data includes at least the soundtrack audio associated with the currently displayed television program, the content recognition engine 110 identifies content item data that is based on the soundtrack audio associated with the currently displayed television program and that also matches the ‘television show’ content type. Thus, in some examples, the content recognition engine 110 identifies content item data relating to a name of the currently displayed television program. For example, the content recognition engine 110 can determine that a particular content item (e.g., a specific television show) is associated with a theme song (e.g., the soundtrack audio), and that the particular content item (e.g., the specific television show) matches the particular content type (e.g., ‘television show’ content type). Thus, the content recognition engine 110 can identify data (e.g., the name of the specific television show) that relates to the particular content item (e.g., the currently displayed television program) that is based on the environmental audio data (e.g., the soundtrack audio), and further that matches the particular content type (e.g., ‘television show’ content type).
The disambiguation engine 104 receives the identified content item data from the content recognition engine 110. The disambiguation engine 104 then provides the identified content item data to the mobile computing device 102, at operation (G). For example, the disambiguation engine 104 transmits the identified content item data relating to the currently displayed television program (e.g., a name of the currently displayed television program) to the mobile computing device 102.
In some examples, one or more of the mobile computing device 102, the disambiguation engine 104, the speech recognition engine 106, the keyword mapping engine 108, and the content recognition engine 110 can be in communication with a subset (or each) of the mobile computing device 102, the disambiguation engine 104, the speech recognition engine 106, the keyword mapping engine 108, and the content recognition engine 110. In some embodiments, one or more of the disambiguation engine 104, the speech recognition engine 106, the keyword mapping engine 108, and the content recognition engine 110 can be implemented using one or more computing devices, such as one or more computing servers, a distributed computing system, or a server farm or cluster.
In some embodiments, as mentioned above, the environmental audio data is streamed from the mobile computing device 110 to the disambiguation engine 104. When the environmental audio data is streamed, the above-mentioned process (e.g., operations (A)-(H)) is performed as the environmental audio data is received by the disambiguation engine 104 (i.e., performed incrementally). In other words, as each portion of the environmental audio data is received by (e.g., streamed to) the disambiguation engine 104, operations (A)-(H) are performed iteratively until content item data is identified.
FIG. 2 depicts a flowchart of an example process 200 for identifying content item data based on environmental audio data and a spoken natural language query. The example process 200 can be executed using one or more computing devices. For example, the mobile computing device 102, the disambiguation engine 104, the speech recognition engine 106, the keyword mapping engine 108, and/or the content recognition engine 110 can be used to execute the example process 200.
Audio data that encodes a spoken natural language query and environmental audio data is received (202). For example, the disambiguation engine 104 receives the waveform data 114 from the mobile computing device 102. The waveform data 114 includes the spoken natural query of the user (e.g., “Who directed this show?”) and the environmental audio data (e.g., audio of the currently displayed television program). The disambiguation engine 104 separates the spoken natural language query (“Who directed this show?”) from the background noise of the environment of the user 112 (e.g., audio of the currently displayed television program).
A transcription of the natural language query is obtained (204). For example, the speech recognition system 106 transcribes the natural language query to generate a transcription of the natural language query (e.g., “Who directed this show?”).
A particular content type that is associated with one or more keywords in the transcription is determined (206). For example, the keyword mapping engine 108 identifies one or more keywords (e.g., “directed”) in the transcription (e.g., “Who directed this show?”) that are associated with a particular content type (e.g., ‘television show’ content type). In some embodiments, the keyword mapping engine 108 determines the particular content type that is associated with one or more keywords in the transcription using one or more databases that, for each of multiple content types, maps at least one of the keywords to at least one of the multiple content types. The database provides a connection (e.g., mapping) between the keywords (e.g., “directed”) and the content types (e.g., ‘television show’ content type).
At least a portion of the environmental audio data is provided to a content recognition engine (208). For example, the disambiguation engine 104 provides at least the portion the environmental audio data encoded by the waveform data 114 (e.g., audio of the currently displayed television program) to the content recognition engine 110. In some examples, the disambiguation engine 104 also provides the particular content type (e.g. ‘television show’ content type) that is associated with the one or more keywords (e.g., “directed”) in the transcription to the content recognition engine 110.
A content item is identified that is output by the content recognition engine, and that matches the particular content type (210). For example, the content recognition engine 110 identifies a content item or content item data that is based on the environmental audio data (e.g., audio of the currently displayed television program) and that matches the particular content type (e.g. ‘television show’ content type).
FIGS. 3A and 3B depict portions 300 a and 300 b, respectively, of a system for identifying content item data. Specifically, FIGS. 3A and 3B include disambiguation engines 304 a and 304 b, respectively; and include content recognition engines 310 a and 310 b, respectively. The disambiguation engines 304 a and 304 b are similar to the disambiguation engine 104 of system 100 depicted in FIG. 1; and the content recognition engines 310 a and 310 b are similar to the content recognition engine 110 of system 100 depicted in FIG. 1.
FIG. 3A depicts the portion 300 a including the content recognition engine 310 a. The content recognition engine 310 a is able to identify content item data based on environmental data and that matches a particular content type. In other words, the content recognition engine 310 a is able to appropriately process the environmental data to identify content item data based on the environmental data, and further select one or more of the identified content item data such that the selected content item data matches the particular content type.
Specifically, the disambiguation engine 304 a provides the environmental data and the particular content type to the content recognition engine 310 a, during operation (A). In some embodiments, the disambiguation engine 304 a provides a portion of the environmental data to the content recognition engine 310 a.
The content recognition engine 310 a receives the environmental data and the particular content type from the disambiguation engine 304 a. The content recognition engine 310 a then identifies content item data that is based on the environmental data and that matches the particular content type and provides the identified content item data to the disambiguation engine 304 a, during operation (B). Specifically, the content recognition engine 310 a identifies content item data (e.g., a name of a television show, a name of a song, etc.) that is based on the environmental data. The content recognition engine 310 a then selects one or more of the identified content item data that matches the particular content type. In other words, the content recognition engine 310 a filters the identified content item data based on the particular content type. The content recognition engine 310 a transmits (e.g., over a network) the identified content item data to the disambiguation engine 304 a.
In some examples, when the environmental data includes at least soundtrack audio associated with a currently displayed television program, as mentioned above with respect to FIG. 1, the content recognition engine 310 a identifies content item data that is based on the soundtrack audio associated with the currently displayed television program. The content recognition engine 310 a then filters the identified content item data based on the ‘television show’ content type. For example, the content recognition engine 310 a identifies a ‘theme song name’ and a ‘TV show name’ associated with the soundtrack audio. The content recognition engine 310 a then filters the identified content item data such that the identified content item data also matches the ‘television show’ content type. For example, the content recognition engine 310 a selects the ‘TV show name’ identifying data, and transmits the ‘TV show name’ identifying data to the disambiguation engine 304 a.
In some examples, the content recognition engine 310 a selects a corpus (or index) based on the content type (e.g., ‘television show’ content type). Specifically, the content recognition engine 310 a can have access to a first index relating to the ‘television show’ content type and a second index relating to a ‘movie’ content type. The content recognition engine 310 a appropriately selects the first index based on the ‘television show’ content type. Thus, by selecting the first index (and not selecting the second index), the content recognition engine 310 a can more efficiently identify the content item data (e.g., a name of the television show).
The disambiguation engine 304 a receives the content item data from the content recognition engine 310 a. For example, the disambiguation engine 304 a receives the ‘TV show name’ identifying data from the content recognition engine 310 a. The disambiguation engine 304 a then provides the identifying data to a third party (e.g., the mobile computing device 102 of FIG. 1), during operation (C). For example, the disambiguation engine 304 a provides the ‘TV show name’ identifying data to the third party.
FIG. 3 b depicts the portion 300 b including the content recognition engine 310 b. The content recognition engine 310 b is able to identify content item data based on environmental data. In other words, the content recognition engine 310 b is able to appropriately process the environmental data to identify content item data based on the environmental data, and provide the content item data to the disambiguation engine 304 b. The disambiguation engine 310 b selects one or more of the identified content item data such that the selected content item data matches the particular content type.
Specifically, the disambiguation engine 304 b provides the environmental data to the content recognition engine 310 b, during operation (A). In some embodiments, the disambiguation engine 304 b provides a portion of the environmental data to the content recognition engine 310 b.
The content recognition engine 310 b receives the environmental data from the disambiguation engine 304 b. The content recognition engine 310 b then identifies content item data that is based on the environmental data and provides the identified content item data to the disambiguation engine 304 b, during operation (B). Specifically, the content recognition engine 310 b identifies content item data associated with two or more content items (e.g., a name of a television show, a name of a song, etc.) that is based on the environmental data. The content recognition engine 310 b transmits (e.g., over a network) two or more candidates representing the identified content item data to the disambiguation engine 304 b.
In some examples, when the environmental data includes at least soundtrack audio associated with a currently displayed television program, as mentioned above with respect to FIG. 1, the content recognition engine 310 b identifies content item data relating to two or more content items that is based on the soundtrack audio associated with the currently displayed television program. For example, the content recognition engine 310 b identifies a ‘theme song name’ and a ‘TV show name’ associated with the soundtrack audio, and transmits the ‘theme song name’ and ‘TV show name’ identifying data to the disambiguation engine 304 b.
The disambiguation engine 304 b receives the two or more candidates from the content recognition engine 310 b. For example, the disambiguation engine 304 b receives the ‘theme song name’ and ‘TV show name’ candidates from the content recognition engine 310 b. The disambiguation engine 304 b then selects one of the two or more candidates based on a particular content type and provides the selected candidate to a third party (e.g., the mobile computing device 102 of FIG. 1), during operation (C). Specifically, the disambiguation engine 304 b previously receives the particular content type (e.g., that is associated with an utterance), as described above with respect to FIG. 1. The disambiguation engine 304 b selects a particular candidate of the two or more candidates based on the particular content type. Specifically, the disambiguation engine 304 b selects the particular candidate of the two or more candidates that matches the particular content type. For example, the disambiguation engine 304 b selects the ‘TV show name’ candidate as the ‘TV show name’ candidate matches the ‘television show’ content type.
In some embodiments, the two or more candidates from the content recognition engine 310 b are associated with a ranking score. The ranking score can be associated with any scoring metric as determined by the disambiguation engine 304 b. The disambiguation engine 304 b can further adjust the ranking score of two or more candidates based on the particular content type. Specifically, the disambiguation engine 304 b can increase the ranking score of one or more of the candidates when the respective candidates are matched to the particular content type. For example, the ranking score of the candidate ‘TV show name’ can be increased as it matches the ‘television show’ content type. Furthermore, the disambiguation engine 304 b can decrease the ranking score of one or more of the candidates when the respective candidates are not matched to the particular content type. For example, the ranking score of the candidate ‘theme song name’ can be decreased as it does not match the ‘television show’ content type.
In some embodiments, the two or more candidates can be ranked based on the respective adjusted ranking scores by the disambiguation engine 304 b. For example, the disambiguation engine 304 b can rank the ‘TV show name’ candidate above the ‘theme song name’ candidate as the ‘TV show name’ candidate has a higher adjusted ranking score as compared to the adjusted ranking score of the ‘theme song name’ candidate. In some examples, the disambiguation engine 304 b selects the candidate ranked highest (i.e., has the highest adjusted ranking score).
FIG. 4 depicts a system 400 for identifying content item data based on environmental image data and a spoken natural language query. In short, the system 400 can identify content item data that is based on the environmental image data and that matches a particular content type associated with the spoken natural language query. The system 400 includes a mobile computing device 402, a disambiguation engine 404, a speech recognition engine 406, a keyword mapping engine 408, and a content recognition engine 410, analogous to that of the mobile computing device 102, the disambiguation engine 104, the speech recognition engine 106, the keyword mapping engine 108, and the content recognition engine 110, respectively, of system 100 illustrated in FIG. 1.
In some examples, the user 112 is looking at a CD album cover of a soundtrack of a movie. In the illustrated example, the user 112 would like to know what songs are on the soundtrack. In some examples, the user 112 may not know the name of the movie soundtrack, and may therefore ask the question “What songs are on this?” or “What songs play in this movie?” The mobile computing device 402 detects this utterance, as well as environmental image data associated with the environment of the user 112.
In some examples, the environmental image data associated with the environment of the user 112 can include image data of the environment of the user 112. For example, the environmental image data includes an image of the CD album cover that depicts images related to the movie (e.g., an image of a movie poster of the associated movie). In some examples, the mobile computing device 402 detects the environmental image data utilizing a camera of the mobile computing device 402 that captures an image (or video) of the CD album cover.
The mobile computing device 402 processes the detected utterance to generate waveform data 414 that represents the detected utterance and transmits the waveform data 414 and the environmental image data to the disambiguation engine 404 (e.g., over a network), during operation (A).
The disambiguation engine 404 receives the waveform data 414 and the environmental image data from the mobile computing device 402. The disambiguation engine 404 processes the waveform data 414 and transmits the utterance to the speech recognition engine 406 (e.g., over a network), during operation (B). In some examples, the utterance relates to a query (e.g., a query relating to the movie soundtrack).
The speech recognition system 406 receives the utterance from the disambiguation engine 404. The speech recognition system 406 obtains a transcription of the utterance and provides the transcription to the keyword mapping engine 408, during operation (C). Specifically, the speech recognition system 406 processes the utterance received from the speech recognition engine 406 by generating a transcription of the utterance.
For example, the speech recognition system 406 transcribes the utterance to generate the transcription of “What songs are on this?” In some embodiments, the speech recognition system 406 provides two or more transcriptions of the utterance. For example, the speech recognition system 406 transcribes the utterance to generate the transcriptions of “What songs are on this?” and “What sinks are on this?”
The keyword mapping engine 408 receives the transcription from the speech recognition engine 406. The keyword mapping engine 408 identifies one or more keywords in the transcription that are associated with a particular content type and provides the particular content type to the disambiguation engine 404, during operation (D).
For example, the keyword mapping engine 408 identifies the keyword “songs” from the transcription of “What songs are on this?” The keyword “songs” is associated with the ‘music’ content type. In some embodiments, a keyword of the transcription that is identified by the keyword mapping engine 408 is associated with two or more content types. For example, the keyword “songs” is associated with the ‘music’ and ‘singer’ content types. The keyword mapping engine 408 transmits (e.g., over a network) the particular content type to the disambiguation engine 408.
In some embodiments, analogous to that mentioned above, the keyword mapping engine 408 identifies the one or more keywords in the transcription that are associated with a particular content type using one or more databases that, for each of multiple content types, maps at least one of the keywords to at least one of the multiple content types. For example, the keyword mapping engine 408 uses the one or more databases that maps the keyword “songs” to the ‘music’ and ‘singer’ content types.
The disambiguation engine 404 receives the particular content type associated with the transcription of the utterance from the keyword mapping engine 408. Furthermore, as mentioned above, the disambiguation engine 404 receives the environmental image data associated with the utterance. The disambiguation engine 404 then provides the environmental image data and the particular content type to the content recognition engine 410, during operation (E).
For example, the disambiguation engine 404 transmits the environmental image data relating to the movie soundtrack (e.g., an image of the movie poster CD album cover) and the particular content type of the transcription of the utterance (e.g., ‘music’ content type) to the content recognition engine 410.
The content recognition engine 410 receives the environmental image data and the particular content type from the disambiguation engine 404. The content recognition engine 410 then identifies content item data that is based on the environmental image data and that matches the particular content type and provides the identified content item data to the disambiguation engine 404, during operation (F). Specifically, the content recognition engine 410 appropriately processes the environmental image data to identify content item data (e.g., a name of a content item). Additionally, the content recognition engine 410 matches the identified content item with the particular content type (e.g., content type of the transcription of the utterance). The content recognition engine 408 transmits (e.g., over a network) the identified content item data to the disambiguation engine 408.
For example, the content recognition engine 410 identifies data that is based on the environmental image data relating to the image of the movie poster CD album cover, and further that matches the ‘music’ content type.
In some examples, when the environmental image data includes at least the movie poster image associated with the CD album cover, the content recognition engine 410 identifies content item data that is based on the movie poster associated with the CD album cover and that also matches the ‘music’ content type. Thus, in some examples, the content recognition engine 410 identifies content item data relating to a name of the movie soundtrack. For example, the content recognition engine 410 can determine that a particular content item (e.g., a specific movie soundtrack) is associated with a movie poster, and that the particular content item (e.g., the specific movie soundtrack) matches the particular content type (e.g., ‘music’ content type). Thus, the content recognition 410 can identify data (e.g., the name of the specific movie soundtrack) that relates to the particular content item (e.g., the specific movie soundtrack) that is based on the environmental image data (e.g., the image of the CD album cover), and further that matches the particular content type (e.g., ‘music’ content type).
The disambiguation engine 404 receives the identified content item data from the content recognition engine 410. The disambiguation engine 404 then provides the identified content item data to the mobile computing device 402, at operation (G). For example, the disambiguation engine 404 transmits the identified content item data relating to the movie soundtrack (e.g., a name of the movie soundtrack) to the mobile computing device 402.
As noted above, FIGS. 1 to 4 illustrate several example processes in which the computing environment can identify media content (or other content) based on environmental information, such as ambient noises. Other processes for identifying content can also be used. Generally, FIGS. 5 and 6 illustrate other example processes in which a computing environment can augment the spoken natural language query with context that is derived from the environmental information, such as data that identifies media content, in order to provide a more satisfying answer to the spoken natural language query.
In more detail, FIG. 5 depicts a system 500 for identifying one or more results based on environmental audio data and an utterance. In some examples, the one or more results can represent one or more answers to a natural language query. The system 500 includes a mobile computing device 502, a coordination engine 504, a speech recognition engine 506, a content identification engine 508, and a natural language query processing engine 510. The mobile computing device 502 is in communication with the coordination engine 504 over one or more networks. The mobile device 510 can include a microphone, a camera, or other detection mechanisms for detecting utterances from a user 512 and/or environmental data associated with the user 512.
Similar to the system 100 of FIG. 1, the user 512 is watching a television program. In the illustrated example, the user 512 would like to know who directed the television program (e.g., an entity) that is currently playing. In some examples, the user 512 may not know the name of the television program that is currently playing, and may therefore ask the question “Who directed this show?” The mobile computing device 502 detects this utterance, as well as environmental data associated with the environment of the user 512.
In some examples, the environmental data associated with the environment of the user 512 can include background noise of the environment of the user 512. For example, the environmental data includes the sounds of the television program (e.g., an entity). In some examples, the environmental data that is associated with the currently displayed television program can include audio of the currently displayed television program (e.g., dialogue of the currently displayed television program, soundtrack audio associated with the currently displayed television program, etc.). In some examples, the environmental data can include environmental audio data, environmental image data, or both. In some examples, the mobile computing device 502 detects the environmental audio data after detecting the utterance; detects the environmental audio data concurrently with detecting the utterance; or both. The mobile computing device 502 processes the detected utterance and the environmental data to generate waveform data 514 that represents the detected utterance and detected environmental audio data (e.g., the sounds of the television program) and transmits the waveform data 514 to the coordination engine 504 (e.g., over a network), during operation (A).
The coordination engine 504 receives the waveform data 514 from the mobile computing device 502. The coordination engine 504 processes the waveform data 514, including separating (or extracting) the utterance from other portions of the waveform data 514 and transmits the portion of the waveform data 514 corresponding to the utterance to the speech recognition engine 506 (e.g., over a network), during operation (B). For example, the coordination engine 504 separates the utterance (“Who directed this show?”) from the background noise of the environment of the user 512 (e.g., audio of the currently displayed television program). In some examples, the coordination engine 504 utilizes a voice detector to facilitate separation of the utterance from the background noise by identifying a portion of the waveform data 514 that includes voice activity. In some examples, the utterance relates to a query (e.g., a query relating to the currently displayed television program).
The speech recognition engine 506 receives the portion of the waveform data 514 corresponding to the utterance from the coordination engine 504. The coordination engine 506 obtains a transcription of the utterance and provides the transcription to the coordination engine 504, during operation (C). Specifically, the speech recognition system 506 appropriately processes the portion of the waveform data 514 corresponding to the utterance received from the coordination engine 504. In some examples, processing of the portion of the waveform data 514 corresponding to the utterance by the speech recognition engine 506 includes generating a transcription of the utterance. Generating the transcription of the utterance can include transcribing the utterance into text or text-related data. In other words, the speech recognition engine 506 can provide a representation of language in written form of the utterance.
For example, the speech recognition engine 506 transcribes the utterance to generate the transcription of “Who directed this show?” In some embodiments, the speech recognition engine 506 provides two or more transcriptions of the utterance. For example, the speech recognition engine 506 transcribes the utterance to generate the transcriptions of “Who directed this show?” and “Who directed this shoe?”
The coordination engine 504 receives the transcription of the utterance from the speech recognition engine 506. Furthermore, as mentioned above, the coordination engine 504 receives the waveform data 514 from the mobile computing device 502 that includes the environmental audio data associated with the utterance. The coordination engine 504 then identifies an entity using the environmental data. Specifically, the coordination engine 504 obtains data that identifies an entity from the content identification engine 508. To that end, the coordination engine 504 provides the environmental audio data and the portion of the waveform 514 corresponding to the utterance to the content identification engine 508 (e.g., over a network), during operation (D).
For example, the coordination engine 504 transmits the environmental data relating to the currently displayed television program (e.g., the entity) that includes audio of the currently displayed television program (e.g., dialogue of the currently displayed television program, soundtrack audio associated with the currently displayed television program, etc.) and the portion of the waveform 514 corresponding to the utterance (“Who directed this show?”) to the content identification engine 508.
In some embodiments, the coordination engine 504 provides a portion of the environmental data to the content identification engine 508. In some examples, the portion of the environmental data can include background noise detected by the mobile computing device 502 after detecting the utterance. In some examples, the portion of the environmental data can include background noise detected by the mobile computing device 502 concurrently with detecting the utterance.
The content identification engine 508 receives the environmental data and the portion of the waveform 514 corresponding to the utterance from the coordination engine 504. The content identification engine 508 identifies data that identifies the entity (e.g., content item data) that is based on the environmental data and the utterance and provides the data that identifies the entity to the coordination engine 504 (e.g., over a network), during operation (E). Specifically, the content identification engine 508 appropriately processes the environmental data and the portion of the waveform 514 corresponding to the utterance to identify data that identifies the entity (e.g., content item data) that is associated with the environmental data (e.g., a name of a television show, a name of a song, etc.).
For example, the content identification engine 508 processes the environmental audio data to identify content item data that is associated with the currently displayed television program. In some embodiments, the content identification engine 508 is the system 100 of FIG. 1.
The coordination engine 504 receives the data that identifies the entity (e.g., the content item data) from the content identification engine 508. Furthermore, as mentioned above, the coordination engine 504 receives the transcription from the speech recognition engine 506. The coordination engine 504 then provides a query including the transcription and the data that identifies the entity to the natural language query processing engine 510 (e.g., over a network), during operation (F). For example, the coordination engine 504 submits a query to the natural language query processing engine 510 that includes the transcription of the utterance (“Who directed this show?”) and the content item data (‘television show name’) to the natural language query processing engine 510.
In some examples, the coordination engine 504 generates the query. In some examples, the coordination engine 504 obtains the query (e.g., from a third-party server). For example, the coordination engine 504 can submit the transcription of the utterance, and the data that identifies the entity to the third-party server, and receive back the query based on the transcription and the data that identifies the entity.
In some embodiments, generating the query by the coordination engine 504 can include associating the transcription of the utterance with the data that identifies the entity (e.g., the content item data). In some examples, associating the transcription of the utterance with the content item data can include tagging the transcription with the data that identifies the entity. For example, the coordination engine 504 can tag the transcription “Who directed this show?” with the ‘television show name’ or other identifying information associated with the content item data (e.g., an identification (ID) number). In some examples, associating the transcription of the utterance with the data that identifies the entity can include substituting a portion of the transcription with the data that identifies the entity. For example, the coordination engine 504 can substitute a portion of the transcription “Who directed this show?” with the ‘television show name’ or data identifying the ‘television show name.’ In some examples, substituting a portion of the transcription with the data that identifies the entity can include substituting one or more words of the transcription of the utterance with the data that identifies the entity. For example, the coordination engine 504 can substitute the ‘television show name’ or data identifying the ‘television show name’ in the transcription of “Who directed this show?” For example, the substitution can result in the transcription including “Who directed ‘television show name’?” or “Who directed ‘ID number’?”
The natural language query processing engine 510 receives the query that includes the transcription and the data that identifies the entity (e.g., the content item data) from the coordination engine 504. The natural language query processing engine 510 appropriately processes the query and based on the processing, provides one or more results to the coordination engine 504 (e.g., over a network), during operation (G). In other words, the coordination engine 510 obtains one or more results of the query (e.g., from the natural language query processing engine 510).
Specifically, the natural language query processing engine 510 obtains information resources (from a collection of information resources) relevant to the query (the transcription of the utterance and the content item data). In some examples, the natural language query processing engine 510 matches the query against database information (e.g., text documents, images, audio, video, etc.) and a score is calculated on how well each object in the database matches the query. The natural language query processing engine 510 identifies one or more results based on the matched objects (e.g., objects having a score above a threshold score).
For example, the natural language query processing engine 510 receives the query that includes the transcription of the utterance “Who directed this show” and the ‘television show name’ (or other identifying information). The natural language query processing engine 510 matches the query against database information, and provides one or more results that match the query. The natural language query processing engine 510 calculates a score of each of the matching objects.
The coordination engine 504 receives the one or more results from the natural language processing engine 510. The coordination engine 504 then provides the one or more results to the mobile computing device 502 (e.g., over a network), at operation (H). For example, the coordination engine 504 transmits the one or more results (e.g., the name of the director of the television show) to the mobile computing device 502.
In some examples, one or more of the mobile computing device 502, the coordination engine 504, the speech recognition engine 506, the content identification engine 508, and the natural language query processing engine 510 can be in communication with a subset (or each) of the mobile computing device 502, the coordination engine 504, the speech recognition engine 506, the content identification engine 508, and the natural language query processing engine 510. In some embodiments, one or more of the coordination engine 504, the speech recognition engine 506, the content identification engine 508, and the natural language query processing engine 510 can be implemented using one or more computing devices, such as one or more computing servers, a distributed computing system, or a server farm or cluster.
FIG. 6 depicts a flowchart of an example process 600 for identifying one or more results based on environmental data and an utterance. The example process 600 can be executed using one or more computing devices. For example, the mobile computing device 502, the coordination engine 504, the speech recognition engine 506, the content identification engine 508, and/or the natural language query processing engine 510 can be used to execute the example process 600.
Audio data that encodes an utterance and environmental data is received (602). For example, the coordination engine 504 receives the waveform data 514 from the mobile computing device 502. The waveform data 514 includes the utterance of the user (e.g., “Who directed this show?”) and the environmental data (e.g., audio of the currently displayed television program). In some examples, receiving the environmental data can include receiving environmental audio data, environmental image data, or both. In some examples, receiving the environmental data includes receiving additional audio data that includes background noise.
A transcription of the utterance is obtained (604). For example, the coordination engine 504 obtains a transcription of the utterance using the speech recognition engine 506. The speech recognition engine 506 transcribes the utterance to generate a transcription of the utterance (e.g., “Who directed this show?”).
An entity is identified using the environmental data (606). For example, the coordination engine 504 obtains data identifying the entity using the content identification engine 508. The content identification engine 508 can appropriately process the environmental data (e.g., the environmental audio data associated with the displayed television program) to identify data identifying the entity (e.g., content item data) that is associated with the environmental data (e.g., a name of a television show, a name of a song, etc.). In some examples, the content identification engine 508 can further process the waveform 514 corresponding to the utterance (concurrently or subsequently to processing the environmental data) to identify the entity.
In some examples, the coordination engine 504 generates a query. In some examples, generating of the query by the coordination engine 504 can include associating the transcription of the utterance with the data that identifies the entity. In some examples, associating the transcription of the utterance with the content item data can include substituting a portion of the transcription with the data that identifies the entity. In some example, substituting a portion of the transcription with the data that identifies the entity can include substituting one or more words of the transcription of the utterance with the data that identifies the entity.
The query is submitted to a natural language processing engine (608). For example, the coordination engine 504 submits the query to the natural language query processing engine 510. The query can include at least a portion of the transcription and the data that identifies the entity (e.g., the content item data). For example, the coordination engine 504 submits a query to the natural language query processing engine that includes the transcription of the utterance (“Who directed this show?”) and the content item data (television show name') to the natural language query processing engine 510.
One or more results of the query are obtained (610). For example, the coordination engine 510 obtains one or more results (e.g., the name of the director of the television show) of the query from the natural language query processing engine 510. In some examples, the coordination engine 504 then provides the one or more results to the mobile computing device 502.
FIG. 7 shows an example of a generic computer device 700 and a generic mobile computer device 750, which may be used with the techniques described here. Computing device 700 is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. Computing device 750 is intended to represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the inventions described and/or claimed in this document.
Computing device 700 includes a processor 702, memory 704, a storage device 706, a high-speed interface 708 connecting to memory 704 and high-speed expansion ports 710, and a low speed interface 712 connecting to low speed bus 714 and storage device 706. Each of the components 702, 704, 706, 708, 710, and 712, are interconnected using various busses, and may be mounted on a common motherboard or in other manners as appropriate. The processor 702 may process instructions for execution within the computing device 700, including instructions stored in the memory 704 or on the storage device 706 to display graphical information for a GUI on an external input/output device, such as display 716 coupled to high speed interface 708. In other implementations, multiple processors and/or multiple buses may be used, as appropriate, along with multiple memories and types of memory. Also, multiple computing devices 700 may be connected, with each device providing portions of the necessary operations (e.g., as a server bank, a group of blade servers, or a multi-processor system).
The memory 704 stores information within the computing device 700. In one implementation, the memory 704 is a volatile memory unit or units. In another implementation, the memory 704 is a non-volatile memory unit or units. The memory 704 may also be another form of computer-readable medium, such as a magnetic or optical disk.
The storage device 706 is capable of providing mass storage for the computing device 700. In one implementation, the storage device 706 may be or contain a computer-readable medium, such as a floppy disk device, a hard disk device, an optical disk device, or a tape device, a flash memory or other similar solid state memory device, or an array of devices, including devices in a storage area network or other configurations. A computer program product may be tangibly embodied in an information carrier. The computer program product may also contain instructions that, when executed, perform one or more methods, such as those described above. The information carrier is a computer- or machine-readable medium, such as the memory 704, the storage device 706, or a memory on processor 702.
The high speed controller 708 manages bandwidth-intensive operations for the computing device 700, while the low speed controller 712 manages lower bandwidth-intensive operations. Such allocation of functions is exemplary only. In one implementation, the high-speed controller 708 is coupled to memory 704, display 716 (e.g., through a graphics processor or accelerator), and to high-speed expansion ports 710, which may accept various expansion cards (not shown). In the implementation, low-speed controller 712 is coupled to storage device 706 and low-speed expansion port 714. The low-speed expansion port, which may include various communication ports (e.g., USB, Bluetooth, Ethernet, wireless Ethernet) may be coupled to one or more input/output devices, such as a keyboard, a pointing device, a scanner, or a networking device such as a switch or router, e.g., through a network adapter.
The computing device 700 may be implemented in a number of different forms, as shown in the figure. For example, it may be implemented as a standard server 720, or multiple times in a group of such servers. It may also be implemented as part of a rack server system 724. In addition, it may be implemented in a personal computer such as a laptop computer 722. Alternatively, components from computing device 700 may be combined with other components in a mobile device (not shown), such as device 750. Each of such devices may contain one or more of computing device 700, 750, and an entire system may be made up of multiple computing devices 700, 750 communicating with each other.
Computing device 750 includes a processor 752, memory 764, an input/output device such as a display 754, a communication interface 766, and a transceiver 768, among other components. The device 750 may also be provided with a storage device, such as a microdrive or other device, to provide additional storage. Each of the components 750, 752, 764, 754, 766, and 768, are interconnected using various buses, and several of the components may be mounted on a common motherboard or in other manners as appropriate.
The processor 752 may execute instructions within the computing device 650, including instructions stored in the memory 764. The processor may be implemented as a chipset of chips that include separate and multiple analog and digital processors. The processor may provide, for example, for coordination of the other components of the device 750, such as control of user interfaces, applications run by device 750, and wireless communication by device 750.
Processor 752 may communicate with a user through control interface 658 and display interface 756 coupled to a display 754. The display 754 may be, for example, a TFT LCD (Thin-Film-Transistor Liquid Crystal Display) or an OLED (Organic Light Emitting Diode) display, or other appropriate display technology. The display interface 756 may comprise appropriate circuitry for driving the display 754 to present graphical and other information to a user. The control interface 758 may receive commands from a user and convert them for submission to the processor 752. In addition, an external interface 762 may be provide in communication with processor 752, so as to enable near area communication of device 750 with other devices. External interface 762 may provide, for example, for wired communication in some implementations, or for wireless communication in other implementations, and multiple interfaces may also be used.
The memory 764 stores information within the computing device 750. The memory 764 may be implemented as one or more of a computer-readable medium or media, a volatile memory unit or units, or a non-volatile memory unit or units. Expansion memory 754 may also be provided and connected to device 750 through expansion interface 752, which may include, for example, a SIMM (Single In Line Memory Module) card interface. Such expansion memory 754 may provide extra storage space for device 750, or may also store applications or other information for device 750. Specifically, expansion memory 754 may include instructions to carry out or supplement the processes described above, and may include secure information also. Thus, for example, expansion memory 754 may be provide as a security module for device 750, and may be programmed with instructions that permit secure use of device 750. In addition, secure applications may be provided via the SIMM cards, along with additional information, such as placing identifying information on the SIMM card in a non-hackable manner.
The memory may include, for example, flash memory and/or NVRAM memory, as discussed below. In one implementation, a computer program product is tangibly embodied in an information carrier. The computer program product contains instructions that, when executed, perform one or more methods, such as those described above. The information carrier is a computer- or machine-readable medium, such as the memory 764, expansion memory 754, memory on processor 752, or a propagated signal that may be received, for example, over transceiver 768 or external interface 762.
Device 750 may communicate wirelessly through communication interface 766, which may include digital signal processing circuitry where necessary. Communication interface 766 may provide for communications under various modes or protocols, such as GSM voice calls, SMS, EMS, or MMS messaging, CDMA, TDMA, PDC, WCDMA, CDMA2000, or GPRS, among others. Such communication may occur, for example, through radio-frequency transceiver 768. In addition, short-range communication may occur, such as using a Bluetooth, WiFi, or other such transceiver (not shown). In addition, GPS (Global Positioning System) receiver module 750 may provide additional navigation- and location-related wireless data to device 750, which may be used as appropriate by applications running on device 750.
Device 750 may also communicate audibly using audio codec 760, which may receive spoken information from a user and convert it to usable digital information. Audio codec 760 may likewise generate audible sound for a user, such as through a speaker, e.g., in a handset of device 750. Such sound may include sound from voice telephone calls, may include recorded sound (e.g., voice messages, music files, etc.) and may also include sound generated by applications operating on device 750.
The computing device 750 may be implemented in a number of different forms, as shown in the figure. For example, it may be implemented as a cellular telephone 780. It may also be implemented as part of a smartphone 782, personal digital assistant, or other similar mobile device.
Various implementations of the systems and techniques described here may be realized in digital electronic circuitry, integrated circuitry, specially designed ASICs (application specific integrated circuits), computer hardware, firmware, software, and/or combinations thereof. These various implementations may include implementation in one or more computer programs that are executable and/or interpretable on a programmable system including at least one programmable processor, which may be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.
These computer programs (also known as programs, software, software applications or code) include machine instructions for a programmable processor, and may be implemented in a high-level procedural and/or object-oriented programming language, and/or in assembly/machine language. As used herein, the terms “machine-readable medium” “computer-readable medium” refers to any computer program product, apparatus and/or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and/or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term “machine-readable signal” refers to any signal used to provide machine instructions and/or data to a programmable processor.
To provide for interaction with a user, the systems and techniques described here may be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user may provide input to the computer. Other kinds of devices may be used to provide for interaction with a user as well; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic, speech, or tactile input.
The systems and techniques described here may be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a client computer having a graphical user interface or a Web browser through which a user may interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (“LAN”), a wide area network (“WAN”), and the Internet.
The computing system may include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
While this disclosure includes some specifics, these should not be construed as limitations on the scope of the disclosure or of what may be claimed, but rather as descriptions of features of example implementations of the disclosure. Certain features that are described in this disclosure in the context of separate implementations can also be provided in combination in a single implementation. Conversely, various features that are described in the context of a single implementation can also be provided in multiple implementations separately or in any suitable subcombination. Moreover, although features may be described above as acting in certain combinations and even initially claimed as such, one or more features from a claimed combination can in some cases be excised from the combination, and the claimed combination may be directed to a subcombination or variation of a subcombination.
Similarly, while operations are depicted in the drawings in a particular order, this should not be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. In certain circumstances, multitasking and parallel processing may be advantageous. Moreover, the separation of various system components in the implementations described above should not be understood as requiring such separation in all implementations, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.
Thus, particular implementations of the present disclosure have been described. Other implementations are within the scope of the following claims. For example, the actions recited in the claims can be performed in a different order and still achieve desirable results. A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the disclosure. For example, various forms of the flows shown above may be used, with steps re-ordered, added, or removed. Accordingly, other implementations are within the scope of the following claims.

Claims

1. A computer-implemented method comprising:

receiving audio data encoding (i) an utterance and (ii) background audio data;

obtaining a transcription of the utterance;

identifying an entity using the background audio data;

submitting a query to a natural language query processing engine, wherein the query includes at least a portion of the transcription and data that identifies the entity that is identified using the background audio data; and

obtaining one or more results of the query.

2. The computer-implemented method of claim 1, further comprising outputting a representation of at least one of the results.

3. The computer-implemented method of claim 1, wherein the entity is identified further using the utterance.

4. The computer-implemented method of claim 1, further comprising generating the query.

5. The computer-implemented method of claim 4, wherein generating the query comprises associating the transcription with the data that identifies the entity.

6. The computer-implemented method of claim 5, wherein associating further includes tagging the transcription with the data that identifies the entity.

7. The computer-implemented method of claim 5, wherein associating further includes substituting a portion of the transcription with the data that identifies the entity.

8. The computer-implemented method of claim 7, wherein substituting further includes substituting one or more words of the transcription with the data that identifies the entity.

9-10. (canceled)

11. A system comprising:

one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:

receiving, from a computing device, audio data encoding an utterance recorded by the computing device, and image data encoding an image captured by the computing device;

obtaining a transcription of the utterance;

identifying an entity using the image data;

submitting a query to a natural language query processing engine, wherein the query includes at least a portion of the transcription and data that identifies the entity that is identified using the image data; and

obtaining one or more results of the query.

12. The system of claim 11, the operations further include generating the query, wherein generating the query comprises associating the transcription with the data that identifies the entity.

13. The system of claim 12, wherein associating further includes tagging the transcription with the data that identifies the entity.

14. The system of claim 12, wherein associating further includes substituting a portion of the transcription with the data that identifies the entity.

15. The system of claim 14, wherein substituting further includes substituting one or more words of the transcription with the data that identifies the entity.

16-17. (canceled)

18. A computer-readable storage device storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising:

receiving (i) audio data encoding an utterance, and (ii) environmental data;

obtaining a transcription of the utterance;

identifying a title of an item of media content using the environmental data;

submitting a query to a natural language query processing engine, wherein the query includes at least a portion of the transcription and data that identifies the title of the item of media content that is identified using the environmental data; and

obtaining one or more results of the query.

19. The computer-readable storage device of claim 18, the operations further comprise generating the query, wherein generating the query comprises associating the transcription with the data that identifies the entity.

20. The computer-readable storage device of claim 19, wherein associating further includes tagging the transcription with the data that identifies the entity.

21. The computer-readable storage device of claim 19, wherein associating further includes substituting a portion of the transcription with the data that identifies the entity.

22. The computer-readable storage device of claim 21, wherein substituting further includes substituting one or more words of the transcription with the data that identifies the entity.

23. The computer-implemented method of claim 1, further comprising:

providing, to a query generation engine, the transcription of the utterance and the data that identifies the entity; and

receiving, from the query generation engine, the query based on providing the transcription of the utterance and the data that identifies the entity to the query generation engine.

24. The computer-implemented method of claim 1, further comprising:

determining, for each result of the one or more results, a score based on a matching between the query and the result;

comparing, for each result of the one or more results, the score to a threshold score; and

based on comparing the score to the threshold score, identifying one or more particular results of the one or more results associated with a score that satisfied the threshold score.