Abstract: Approaches are described for detecting when an electronic device (such as a mobile phone) has been stolen or is otherwise being used by someone other than an authorized user of the device. At least one sensor of the device can obtain data during a current use of the device, and the device can determine from the data a set of available features. The features can be compared to a corresponding model associated with an owner (or other authorized user) of the device to generate a confidence value indicative of whether the current user operating the device is likely the owner of the device. The confidence value can be compared to at least one confidence threshold, for example, and based on the comparison, the current user can be provided access to at least a portion of functionality of the device and/or a security action can be performed when the confidence value does not at least meet at least one confidence threshold.

Abstract: Power consumption for a computing device may be managed by one or more keywords. For example, if an audio input obtained by the computing device includes a keyword, a network interface module and/or an application processing module of the computing device may be activated. The audio input may then be transmitted via the network interface module to a remote computing device, such as a speech recognition server. Alternately, the computing device may be provided with a speech recognition engine configured to process the audio input for on-device speech recognition.

Abstract: Features are disclosed for reducing errors in speech recognition processing. Methods for reducing errors can include receiving multiple speech recognition hypotheses based on an utterance indicative of a command or query of a user and determining a command or query within a grammar having a least amount of difference from one of the speech recognition hypotheses. The determination of the least amount of difference may be based at least in part on a comparison of individual subword units along at least some of the sequence paths of the speech recognition hypotheses and the grammar. For example, the comparison may be performed on the phoneme level instead of the word level.

Abstract: Features are disclosed for active learning to identify the words which are likely to improve the guessing and automatic speech recognition (ASR) after manual annotation. When a speech recognition system needs pronunciations for words, a lexicon is typically used. For unknown words, pronunciation-guessing (G2P) may be included to provide pronunciations in an unattended (e.g., automatic) fashion. However, having manually (e.g., by a human) annotated pronunciations provides better ASR than having automatic pronunciations that may, in some instances, be wrong. The included active learning features help to direct these limited annotation resources.

Abstract: Features are disclosed for generating acoustic models from an existing corpus of data. Methods for generating the acoustic models can include receiving at least one characteristic of a desired acoustic model, selecting training utterances corresponding to the characteristic from a corpus comprising audio data and corresponding transcription data, and generating an acoustic model based on the selected training utterances.

Abstract: A system and method for using speech recognition, natural language understanding, image processing, and facial recognition to automatically analyze the audio and video data of video content and generate enhanced data relating to the video content and characterize the aspects or events of the video content. The results of the analysis and characterization of the aspects of the video content may be used to annotate and enhance the video content to enhance a user's viewing experience by allowing the user to interact with the video content and presenting the user with information related to the video content.

Abstract: Natural language controlled devices may be configured to activate command recognition in response to one or more wake words. Techniques are provided to receive a candidate word for evaluation as a wake word that activates a natural language control functionality of a computing device. The candidate word may include one or more words or sounds. Values for multiple wake word metrics are then determined. The candidate word is evaluated based on the various wake word metrics.

Abstract: Features are disclosed for generating and using personalized named entity recognition models. A personalized model can be trained for a particular user, and then interpolated with a general model for use in named entity recognition. In some embodiments, a model may be trained for a group of users, where the users share some similarity relevant to language processing. In some embodiments, various base models may be trained so as to provide better accuracy for certain types of language input than a general model. Users may be associated with any number of base models, and the associated based models may then be interpolated for use in named entity recognition on input from the corresponding user.

Abstract: A method and apparatus for identifying a document in a set of stored documents based on a pattern of characteristics in the document is presented. A digital image including at least a portion of the a document is acquired. A pattern of characteristics is then identified in the digital image. The pattern is matched to the set of stored documents to identify the document in the digital image from the set of stored documents.

Abstract: An automatic speech recognition (ASR) device may be configured to predict pronunciations of textual identifiers (for example, song names, etc.) based on predicting one or more languages of origin of the textual identifier. The one or more languages of origin may be determined based on the textual identifier. The pronunciations may include a hybrid pronunciation including a pronunciation in one language, a pronunciation in a second language and a hybrid pronunciation that combines multiple languages. The pronunciations may be added to a lexicon and matched to the content item (e.g., song) and/or textual identifier. The ASR device may receive a spoken utterance from a user requesting the ASR device to access the content item. The ASR device determines whether the spoken utterance matches one of the pronunciations of the content item in the lexicon. The ASR device then accesses the content when the spoken utterance matches one of the potential textual identifier pronunciations.