Abstract: A technique for ordering textual messages in a graphical user interface (GUI) of a communication application based on text content can include receiving a textual message, and determining an insertion point in the GUI based on the text content of the received textual message. In some implementations, determining the insertion point can include utilizing a language model to determine a probability that the text content of the textual message is associated with each preceding textual message. Additionally or alternatively, determining an insertion point can include utilizing a timestamp corresponding to a time that the received textual message was initiated by a sender user. The technique can further include displaying the textual message at the determined insertion point of the GUI. In some implementations, the displaying of the textual message can include providing an active indication of the received textual message being inserted at the determined insertion point.

Abstract: Methods, systems, and media for recommending content based on network conditions are provided. In some embodiments, the method comprises: receiving, from a first user device, a request to present media content recommendations on the first user device; in response to receiving the request, determining information indicating a user context associated with the first user device and network connectivity information associated with a connection status of the first user device over a communications network; identifying a group of media content items to recommend based on the user context and the network connectivity information; and causing recommendations for the group of media content items to be presented on the first user device.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for a dynamic threshold for speaker verification are disclosed. In one aspect, a method includes the actions of receiving, for each of multiple utterances of a hotword, a data set including at least a speaker verification confidence score, and environmental context data. The actions further include selecting from among the data sets, a subset of the data sets that are associated with a particular environmental context. The actions further include selecting a particular data set from among the subset of data sets based on one or more selection criteria. The actions further include selecting, as a speaker verification threshold for the particular environmental context, the speaker verification confidence score. The actions further include providing the speaker verification threshold for use in performing speaker verification of utterances that are associated with the particular environmental context.

Abstract: In some examples, a method includes receiving, first notification data from a first application and second notification data from a second application. The method may include generating, based at least in part on comparing the first notification data to the second notification data, an affinity score that indicates whether content from the first notification data has a relationship to at least a portion of content from the second notification data. The method may include, generating a bundle notification graphical element that includes at least a portion of the content from the first notification data and a graphical indicator that indicates the content from the first and second notification data are related. The method may include outputting a set of notification graphical elements comprising the bundle notification graphical element accompanied by the graphical indicator.

Abstract: A technique for ordering textual messages in a graphical user interface (GUI) of a communication application based on text content can include receiving a textual message, and determining an insertion point in the GUI based on the text content of the received textual message. In some implementations, determining the insertion point can include utilizing a language model to determine a probability that the text content of the textual message is associated with each preceding textual message. Additionally or alternatively, determining an insertion point can include utilizing a timestamp corresponding to a time that the received textual message was initiated by a sender user. The technique can further include displaying the textual message at the determined insertion point of the GUI. In some implementations, the displaying of the textual message can include providing an active indication of the received textual message being inserted at the determined insertion point.

Abstract: In some implementations, a language proficiency of a user of a client device is determined by one or more computers. The one or more computers then determines a text segment for output by a text-to-speech module based on the determined language proficiency of the user. After determining the text segment for output, the one or more computers generates audio data including a synthesized utterance of the text segment. The audio data including the synthesized utterance of the text segment is then provided to the client device for output.

Abstract: In some implementations, a language proficiency of a user of a client device is determined by one or more computers. The one or more computers then determines a text segment for output by a text-to-speech module based on the determined language proficiency of the user. After determining the text segment for output, the one or more computers generates audio data including a synthesized utterance of the text segment. The audio data including the synthesized utterance of the text segment is then provided to the client device for output.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for correcting a corrupted data sample using a trained deep neural network, the method including obtaining a feature representation of a corrupted data sample; and modifying the feature representation of the corrupted data sample to generate a feature representation of a corrected data sample by iteratively processing a current version of the feature representation of the corrupted data sample using the trained deep neural network to generate a current corruption score for the current version of the feature representation of the corrupted data sample and generating a less-corrupted version of the feature representation by performing an iteration of gradient descent against the current version of the feature representation of the corrupted data sample to reduce the current corruption score.

Abstract: Computer-implemented techniques can include obtaining, by a client computing device, a digital media item and a request for a processing task on the digital item and determining a set of operating parameters based on (i) available computing resources at the client computing device and (ii) a condition of a network. Based on the set of operating parameters, the client computing device or a server computing device can select one of a plurality of artificial neural networks (ANNs), each ANN defining which portions of the processing task are to be performed by the client and server computing devices. The client and server computing devices can coordinate processing of the processing task according to the selected ANN. The client computing device can also obtain final processing results corresponding to a final evaluation of the processing task and generate an output based on the final processing results.

Abstract: In some examples, a method includes receiving, first notification data from a first application and second notification data from a second application. The method may include generating, based at least in part on comparing the first notification data to the second notification data, an affinity score that indicates whether content from the first notification data has a relationship to at least a portion of content from the second notification data. The method may include, generating a bundle notification graphical element that includes at least a portion of the content from the first notification data and a graphical indicator that indicates the content from the first and second notification data are related. The method may include outputting a set of notification graphical elements comprising the bundle notification graphical element accompanied by the graphical indicator.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving audio data corresponding to an utterance, determining that the audio data corresponds to a hotword, generating a hotword audio fingerprint of the audio data that is determined to correspond to the hotword, comparing the hotword audio fingerprint to one or more stored audio fingerprints of audio data that was previously determined to correspond to the hotword, detecting whether the hotword audio fingerprint matches a stored audio fingerprint of audio data that was previously determined to correspond to the hotword based on whether the comparison indicates a similarity between the hotword audio fingerprint and one of the one or more stored audio fingerprints that satisfies a predetermined threshold, and in response to detecting that the hotword audio fingerprint matches a stored audio fingerprint, disabling access to a computing device into which the utterance was spoken.

Abstract: The timing for displaying a mid-roll advertisement during a video is determined based on analysis of viewer retention data for the video. In an aspect, a system is provided that includes a reception component configured to receive viewer retention information for a video, the viewer retention information indentifying amounts of viewers that have watched respective frames of the video, an identification component configured to identify high interest segments of the video based on analysis of the viewer retention information, wherein the high interest segments are associated with relatively higher viewer interest over other segments of the video. The system further includes an advertisement component configured to select a point to insert a mid-roll advertisement during playback of the video at or near at least one of the high interest segments.

Abstract: In some implementations, a language proficiency of a user of a client device is determined by one or more computers. The one or more computers then determines a text segment for output by a text-to-speech module based on the determined language proficiency of the user. After determining the text segment for output, the one or more computers generates audio data including a synthesized utterance of the text segment. The audio data including the synthesized utterance of the text segment is then provided to the client device for output.

Abstract: In some implementations, a language proficiency of a user of a client device is determined by one or more computers. The one or more computers then determines a text segment for output by a text-to-speech module based on the determined language proficiency of the user. After determining the text segment for output, the one or more computers generates audio data including a synthesized utterance of the text segment. The audio data including the synthesized utterance of the text segment is then provided to the client device for output.

Abstract: Methods, including computer programs encoded on a computer storage medium, for collaborative language model biasing. In one aspect, a method includes receiving (i) data including a set of terms associated with a target user, and, (ii) from each of multiple other users, data including a set of terms associated with the other user, selecting a particular other user based at least on comparing the set of terms associated with the target user to the sets of terms associated with the other users, selecting one or more terms from the set of terms that is associated with the particular other user, obtaining, based on the selected terms that are associated with the particular other user, a biased language model, and providing the biased language model to an automated speech recognizer.

Abstract: A digital content server provides bias scores used for biasing display of sections of a digital content item, such as an e-book, audio track, or video, during scrubbing on a client device. For each user, the server compiles a user profile which includes information such as the user's search and browsing history, stated interests, and location. The server determines a collection of similar user profiles and analyzes them to determine a relevance score for each section of the digital content item. For each section, the server also identifies individual entities, and compares the identified entities against the user profile to determine a second relevance score. The server combines the relevance scores to determine an aggregate bias score for each section of the digital content item. The bias scores are provided to a client device containing a scrubber module, which uses the scores to bias display of sections during scrubbing.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for a dynamic threshold for speaker verification are disclosed. In one aspect, a method includes the actions of receiving, for each of multiple utterances of a hotword, a data set including at least a speaker verification confidence score, and environmental context data. The actions further include selecting from among the data sets, a subset of the data sets that are associated with a particular environmental context. The actions further include selecting a particular data set from among the subset of data sets based on one or more selection criteria. The actions further include selecting, as a speaker verification threshold for the particular environmental context, the speaker verification confidence score. The actions further include providing the speaker verification threshold for use in performing speaker verification of utterances that are associated with the particular environmental context.

Abstract: A system is described that, after executing a parameterless search query to identify results that are determined to be relevant to a user of a device, determines a respective target property, for each of the results, that indicates whether a result is associated with audio content, visual content, audio and visual content, haptic content, or textual content. The system also determines one or more dynamic properties, of the device, that indicate one or more of a network connection strength, connection status to an audio output device, battery level, or connection status to a multimedia streaming device.

Abstract: In some examples, a method includes receiving, first notification data from a first application and second notification data from a second application. The method may include generating, based at least in part on comparing the first notification data to the second notification data, an affinity score that indicates whether content from the first notification data has a relationship to at least a portion of content from the second notification data. The method may include, generating a bundle notification graphical element that includes at least a portion of the content from the first notification data and a graphical indicator that indicates the content from the first and second notification data are related. The method may include outputting a set of notification graphical elements comprising the bundle notification graphical element accompanied by the graphical indicator.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for receiving audio data corresponding to an utterance, determining that the audio data corresponds to a hotword, generating a hotword audio fingerprint of the audio data that is determined to correspond to the hotword, comparing the hotword audio fingerprint to one or more stored audio fingerprints of audio data that was previously determined to correspond to the hotword, detecting whether the hotword audio fingerprint matches a stored audio fingerprint of audio data that was previously determined to correspond to the hotword based on whether the comparison indicates a similarity between the hotword audio fingerprint and one of the one or more stored audio fingerprints that satisfies a predetermined threshold, and in response to detecting that the hotword audio fingerprint matches a stored audio fingerprint, disabling access to a computing device into which the utterance was spoken.