Abstract: A method includes receiving text input data for conversion into synthesized speech and determining, using a hotword-aware model trained to detect a presence of a hotword assigned to a user device, whether a pronunciation of the text input data includes the hotword. The hotword is configured to initiate a wake-up process on the user device for processing the hotword and/or one or more other terms following the hotword in the audio input data. When the pronunciation of the text input data includes the hotword, the method also includes generating an audio output signal from the text input data and providing the audio output signal to an audio output device to output the audio output signal. The audio output signal when captured by an audio capture device of the user device, configured to prevent initiation of the wake-up process on the user device.

Abstract: An information processing apparatus includes an acquisition unit configured to acquire acceleration of a vehicle, a controller configured to execute a speech dialogue with a driver of the vehicle, an input unit configured to receive a speech input by the driver, and an output unit configured to execute a speech output to the driver. The controller dynamically controls a response time in the speech dialogue with the driver based on the acceleration of the vehicle.

Abstract: This disclosure describes computer-based techniques for administering a spoken patterned passphrase. A passphrase processing unit running on an administrator computer generates passphrase data for a secure system using acoustic data and video data representing a spoken phrase by a speaker. This passphrase includes a pattern of words or speech segments that are audible and words or speech segments that are inaudible. During authentication, a passphrase administration unit on the administrator computer receives acoustic and visual data of a spoken phrase by a person attempting to access the secure system and evaluates whether the spoken phrase includes the pattern of audible and inaudible words or speech segments associated with the account. In this way, the techniques discussed herein may enable the administrator computer to administer spoken passphrases with an additional degree of protection than a system that is limited to using linguistic or biometric content in passwords or passphrases.

Abstract: Systems and methods are described herein for disambiguating a voice search query by determining whether the user made a gesture while speaking a quotation from a content item and whether the user mimicked or approximated a gesture made by a character in the content item when the character spoke the words quoted by the user. If so, a search result comprising an identifier of the content item is generated. A search result representing the content item from which the quotation comes may be ranked highest among other search results returned and therefore presented first in a list of search results. If the user did not mimic or approximate a gesture made by a character in the content item when the quotation is spoken in the content item, then a search result may not be generated for the content item or may be ranked lowest among other search results.

Abstract: Techniques are described for detecting a conversation between at least two people, and for reducing noise during the conversation. In certain embodiments, at least one speech metric is generated based on spectral analysis of an audio signal and is used to determine that the audio signal represents speech from a first person. Responsive to determining that the speech is part of a conversation between the first person and a second person an operating state of a device in a physical environment is adjusted such that a volume level of sound contributed by or associated with the device is reduced. The sound contributed by or associated with the device corresponds to noise, at least for the duration of the conversation. Therefore, reducing the volume level of sound contributed by or associated with the device reduces the overall noise level in the environment, resulting in a reduction in conversational effort.

Abstract: Systems, computer-implemented methods, and computer program products that can facilitate word entity disambiguation are provided. According to an embodiment, a system can comprise a memory that stores computer executable components and a processor that executes the computer executable components stored in the memory. The computer executable components can comprise a language model component that employs an artificial intelligence model to generate a profile vector of an entity based on one or more binary values representing profile data of the entity and a word vector of a word entity in a dialogue based on one or more second word entities adjacent to the word entity in the dialogue. The computer executable components can further comprise a dialogue management component that disambiguates the word entity based on the profile vector and the word vector.

Abstract: Methods and apparatuses are described in which unstructured computer text is analyzed for identification and classification of complaint-specific user interactions. A data store receives unstructured computer text corresponding to current user interactions. A server filters the unstructured computer text to identify messages that comprise a potential complaint.

Abstract: A system for distinguishing between a human voice generated command and an electronic speaker generated command is provided. An exemplary system comprises a microphone array for receiving an audio signal collection, preprocessing circuitry configured for converting the audio signal collection into processed recorded audio signals, energy balance metric determination circuitry configured for calculating a final energy balance metric based on the processed recorded audio signals, and energy balance metric evaluation circuitry for outputting a command originator signal based at least in part on the final energy balance metric.

Abstract: A system that performs automatic gain control (AGC) using different decay rates. The system may select a slow decay rate to track a loudness level within speech (e.g., within an utterance), improving audio quality and maintaining dynamic range for an individual voice, while selecting a fast decay rate to track the loudness level after a gap of silence (e.g., no voice activity detected for a duration of time) or during large level changes (e.g., actual speech loudness is lower than estimated speech loudness for a duration of time). This improves an accuracy of the loudness estimate and therefore a responsiveness of the automatic gain control, resulting in an improved user experience.

Abstract: Methods, systems and computer program products for natural language context embedding are provided herein. A computer-implemented method includes extracting a document anatomy and document elements from a given structured document, identifying semantic references in the given structured document, and generating an ontology comprising (i) a hierarchy of concepts and (ii) relations connecting the concepts, each concept comprising attributes for a document element. The computer-implemented method also includes generating natural language text context for a given document element by utilizing the ontology to combine (i) attributes of a given concept corresponding to the given document element with (ii) attributes of another concept, the other concept corresponding to another document element, the other concept being connected to the given concept by at least one relation.

Abstract: A method for encoded-sound determination performed by a computer includes: executing a first process that includes obtaining information indicating intensities of sound signals, the frequencies being calculated from the sound signals and corresponding to frequencies; and executing a second process that includes determining whether or not the sound signals are signals of encoded sound, based on whether or not the intensities of the sound signals in predetermined frequency bands that are adjacent to each other in a frequency direction have a difference that is larger than or equal to a predetermined threshold.

Abstract: A method for automatically detecting block legal billing is described where the technique analyzes each billing entry in the legal bill for visual or textual aspects that indicate that a list of billing items is included in the block. The technique utilizes a combination of textual analysis for punctuation characters, count of the number of verbs, or a search for conjunctions. A visual analysis is match the image of the billing item with a predetermined image of a list. Essentially, a novel natural language processing technique is described that identifies lists in a block of text, where the block of text is in the context of a legal bill.