Abstract: An entailment pattern pair extension apparatus 50 extends entailment pairs by generating an n-term entailment pair from an m-term entailment pair, where m and n are integers not smaller than 0 and satisfying m<n. Each entailment pair includes a first language pattern and a second language pattern entailed by the first language pattern. Entailment pattern pair extension apparatus 50 includes: a generation rule storage unit 110 storing generation rules for generating n-term entailment pairs from m-term entailment pairs, and a binary pair adding unit 112 receiving m-term entailment pair, determining, for each generation rule stored in the generation rule storage unit 110, whether its condition is satisfied by the m-term pair, and if the condition is satisfied, applying a modification rule of the generation rule to each language pattern constituting the m-term entailment pair.

Abstract: Data associated with a selectively offline capable voice action is locally persisted in a voice-enabled electronic device whenever such an action cannot be competed locally due to the device being offline to enable the action to later be completed after online connectivity has been restored. Synchronization with an online service and/or another electronic device, and/or retrieval of context sensitive data from an online service may be performed after online connectivity has been restored to enable the voice action to thereafter be completed.

Abstract: The present disclosure provides a voiceprint authentication method and a voiceprint authentication apparatus. The method includes: displaying a tip text to a user, the tip text being a combination of a preregistered phrase; obtaining a speech of the tip text read by the user; obtaining a pre-established registration model and determining a result of a voiceprint authentication according to the speech of the tip text and the pre-established registration model, if the speech of the tip text corresponds to the tip text.

Abstract: Exemplary embodiments provide techniques for evaluating when words or phrases of a translation were generated with a low degree of confidence, and conveying this information when the translation is presented. For example, if a source language word is encountered in source material for translation, but the source language word was only encountered a few times (or not at all) in the training data used to train the translation system, then the resulting translation may be flagged as being of low confidence. Other situations, such as the generation of two equally-likely translations, or translation system model disagreement, may also indicate a questionable translation. When the translation is displayed, questionable words and phrases may be flagged, and possible alternative translations may be presented. If one of the alternatives is selected, this information may be used to update the translation system's models in order to improve translation quality in the future.

Abstract: An audio capture device that incorporates a beamformer and beam-specific trigger word detection. Audio data from each beam is processed by a low power trigger word detector, such as a neural network or other trained model to detect if audio data (such as an audio frame or feature vector corresponding thereto) likely includes part of a trigger word. The beam that either most strongly represents a trigger word portion or represents a trigger word portion most early in time may be selected for further processing such as speech processing or confirmation by a more robust power intensive trigger word detector.

Abstract: A language validation system includes a language validation module (LVM) and a linguistic intelligence engine (LIE). The LVM may perform language validation on text in a source document. The LVM may identify a language of origin for each word or sentence in the document. The LVM may determine whether a word/sentence not of the expected language constitutes a true defect. The LVM may provide input to an LIE database as the LVM identifies new words/sentences. The LIE may remain dormant for an initial interval until it obtains a sufficient vocabulary. Eventually, the LIE may be sufficiently knowledgeable to perform automated translations requests to translate original documents in the original language into one or more source documents in one or more source language. The LIE may adapt or otherwise implement a particular style in accordance with a style guide or analogous information.

Abstract: Exemplary embodiments relate to detecting, removing, and/or replacing objectionable words and phrases in a machine-generated translation. A classifier identifies translations containing target words or phrases. The classifier may be applied to the output translation to remove target words and phrases from the translation, or to prevent target words and phrases from being automatically presented. Further, the classifier may be applied to a translation model to prevent the target words and phrases from appearing in the output translation. Still further, the classifier may be applied to training data so that the translation model is not trained using the target words of phrases. The classifier may remove target words or phrases only when the target words or phrases appear in the output translation but not the source language input data. The classifier may be provided as a standalone service, or may be employed in the context of a machine translation system.

Abstract: Applications and services are disclosed for transforming spoken thoughts to a visual representation of related topics found in the spoken thoughts. Transforming the spoken thoughts can include receiving audio input of a spoken thought of a user and converting the audio input to text of the spoken thought. The system can identify a plurality of topics from the text using word frequency, and can create a topic graph for the plurality of topics by determining relationships between topics of the plurality of topics, and assigning a weight to each of the plurality of topics based at least on the determined relationships. The system can generate a visual display of the topic graph for the plurality of topics by representing relevant topics of the plurality of topics using topic shapes, which can be connected by the determined relationships and sized by the assigned weight.

Abstract: Techniques for joining a device of a third user to a communication between a device of a first user and a device of a second user are described herein. For instance, two or more users may utilize respective computing devices to engage in a telephone call, a video call, an instant-messaging session, or any other type of communication in which the users communicate with each other audibly and/or visually. In some instances, a first user of the two users may issue a voice command requesting to join a device of a third user to the communication. One or more computing devices may recognize this voice command and may attempt to join a device of a third user to the communication.

Abstract: An input method editor (IME) is associated with a local user. Memory stores local data and a processor, coupled to the memory, is configured to receive input from a local, first user, obtain shared data associated with at least a remote, second user from a remote server and generate prediction candidates and conversion candidates based on the input provided by the local, first user and correlation of the input and the obtained shared data.

Abstract: Exemplary embodiments relate to detecting, removing, and/or replacing objectionable words and phrases in a machine-generated translation. A classifier identifies translations containing target words or phrases. The classifier may be applied to the output translation to remove target words and phrases from the translation, or to prevent target words and phrases from being automatically presented. Further, the classifier may be applied to a translation model to prevent the target words and phrases from appearing in the output translation. Still further, the classifier may be applied to training data so that the translation model is not trained using the target words of phrases. The classifier may remove target words or phrases only when the target words or phrases appear in the output translation but not the source language input data. The classifier may be provided as a standalone service, or may be employed in the context of a machine translation system.

Abstract: One embodiment provides a method, comprising: receiving, at an information handling device, user input comprising a word that activates a digital assistant associated with the information handling device; determining, using a processor, whether the word comprises a false positive wake word, wherein the false positive wake word comprises a stored wake word; and responsive to determining that the word is associated with a false positive wake word, providing output associated with an alternative wake word. Other aspects are described and claimed.

Abstract: One or more content items can be received at a data recognition module. The data recognition module can utilize, individually or in any combination, image recognition (e.g., OCR, object recognition, etc.), audio recognition (e.g., speech recognition, music identification, etc.), and/or text recognition (e.g., text crawling) in order to identify or recognize at least a portion of the one or more content items. Based on the identified content portion(s), the one or more content items and/or their respective source(s) can be classified. In one example, an image containing a not yet machine-readable curse word can be included in a source webpage. The image can be received at the data recognition module. The curse word contained in the image can be recognized/identified using an OCR process. Based, at least in part, on the recognized/identified curse word, the image and/or the webpage can be classified as likely being associated with inappropriate material.

Abstract: Techniques for integrating a virtual assistant into a spoken conversation session, the techniques including receiving an utterance information that expresses an utterance spoken by a first participant included in a plurality of participants of a spoken conversation session; processing the utterance information using at least one machine-trained model to determine an intent or content for a command or query included in the utterance; selectively identifying a recipient subset of one or more of the plurality of participants based on at least the determined intent or content for the utterance; generating a response for the command or query; and providing, during the spoken conversation session, the response to the identified recipient subset.

Abstract: In one embodiment, a method includes receiving, from a client device that corresponds to a user of an online social network, an input that comprises free-form text; determining, through application of natural-language processing of the free-form text, an affinity declaration for an object associated with the online social network; determining an affinity coefficient between respective user and the object; adjusting the determined affinity coefficient based on social-networking information of the user, wherein the social-networking information reinforces or reduces the determined affinity coefficient; and upon determining that the determined affinity coefficient is above a threshold coefficient, creating or modifying an edge connection in a social graph between a user node corresponding to the user and a concept node corresponding to the object.

Abstract: Devices, systems and methods are disclosed for estimating a prior probability for speech recognition by taking into account a number of observations of a particular word and a prior probability for a group of words having a similar number of observations. For example, a prior probability may be determined by combining a number of correct results and a number of observations for a group of words and calculating a prior probability of the entire group. Further, a prior probability may be determined for a word that was not previously observed by determining a prior probability for a group of words that have been observed once. The prior probability for a particular word may be determined differently as the number of observations increases and may transition from the group prior probability to an individual prior probability when the number of observations exceeds a threshold.

Abstract: A terminal comprises: a speech receiving unit that receives speech in a locked state; a voiceprint authentication unit that performs voiceprint authentication based on the speech received in the locked state and determining whether or not a user is legitimate; a speech recognition unit that performs speech recognition of the speech received in the locked state; and an execution unit that executes an application using a result of the speech recognition.

Abstract: Speech generating devices, communication systems, and methods for communicating using the devices and systems are disclosed herein. In certain examples, the speech generating device includes a display device and an input device configured to generate a communication to be displayed on the display device, wherein the speech generating device is configured to allow a user to select between playing the generated communication through a speaker and transmitting the generated communication via a communication network to a computing device separate from the speech generating device. In certain examples, the computing device may be designated with other computing devices of other conversation partners within a conversation group.

Abstract: Disclosed is a scalable method for automatically deriving the topics discussed most prevalently in unstructured, multilingual text, and simultaneously revealing which topics are more biased towards one or another ‘information space’. The concept of the ‘information space’ is derived from Russian strategic doctrine on information warfare; an example of an ‘information space’ would be the portion of social media in which the Russian language is used. The disclosed method leverages this concept, in conjunction with unsupervised multilingual machine learning, to determine, automatically and without any built-in bias or preconceived notions of what is important, which topics are more discussed, for example, in one language than another. An analyst's attention can then be focused on the most important differences between national discourses, and insight more quickly gained into the areas (both topics and geographic regions) in which propaganda of the sort envisaged in Russian strategic doctrine may be taking hold.

Abstract: A terminal comprises: a speech receiving unit that receives speech in a locked state; a voiceprint authentication unit that performs voiceprint authentication based on the speech received in the locked state and determining whether or not a user is legitimate; a speech recognition unit that performs speech recognition of the speech received in the locked state; and an execution unit that executes an application using a result of the speech recognition.