Robert Bosch

Kui Xu

Zhe Feng

Fuliang Weng

Lin Zhao

Zhongnan Shen

Yifan He

Liu Ren

Baoshi Yan

Zhengyu Zhou

Yen-Lin Chen

Haibo Ding

Lawrence Cavedon

Hong Luo

Fei Liu

Chin-Hui Lee

Danilo Mirkovic

Tobias Scheideck

Qi Zhang

Xiaohua Jiang

Katrina Li

Ben Wang

Jianjie Zhang

Zhenyu Zhou

Wenwei Guo

Tobias Menne

Kehuang Li

Matthew Purver

Florin Ratiu

An automatic terminology linking system includes a candidate generator configured to identify candidate nodes for each terminology that is to be linked to a node of the knowledge base. A pseudo-candidate generator is configured to identify pseudo-candidate nodes for candidate-less terminologies. A candidate scorer is configured to respectively score the candidate nodes and the pseudo-candidate nodes by collective inference using occurrence statistics and co-occurrence statistics for these nodes. The pseudo-candidate generator is configured to identify knowledge base nodes that are semantically-related to candidate-less terminology as the pseudo-candidate nodes for the candidate-less terminology.

System and method for domain-independent terminology linking

A method of automatically generating a terminology definition knowledge base (KB) includes mapping each word in a word sequence to a real value dense vector using dense vector representations. The word sequence is then processed using a Convolutional Neural Network (CNN) model to identify whether the word sequence includes a terminology definition and to label the word sequence with a label indicating whether a terminology definition exists within the word sequence. The word sequence is then processed using a Conditional Random Field (CRF) model to identify boundaries of the terminology definition in the word sequence. The terminology definition is then extracted and added to the terminology definition KB.

System and method for domain- and language-independent definition extraction using deep neural networks

A method for automated aspect-based sentiment analysis includes parsing reviews from a first domain to generate rhetorical structure trees and extracting rhetorical rules from the rhetorical structure trees, each rhetorical rule including a path extracted from at least one span in at least one of the rhetorical structure trees associated with a probability that the path corresponds to a positive or negative sentiment based on annotation data. The method further includes parsing reviews from a second domain to generate a second plurality of rhetorical structure trees, generating training data that associates at least one aspect in the review from the second domain with a sentiment associated with a rhetorical rule in the plurality of rhetorical rules, and training a classifier to identify sentiments in reviews from the second domain using the second plurality of reviews and the training data.

System and method for domain-independent aspect level sentiment detection

A glove for use in collecting data for sign language recognition comprises: multiple azimuth sensors arranged on the glove at positions corresponding to the phalanges and metacarpal bones of the hand and used for sensing postures of the hand. The azimuth sensors are only arranged on the glove at positions corresponding to the phalanges of the hand other than the distal phalange in proximity to the fingertip of at least one finger among the middle finger, the index finger, the ring finger, and the little finger. The glove reduces the number of the azimuth sensors arranged on the glove at positions corresponding to the phalanges of the hand, thus reducing costs while not affecting detection performance.

Glove for use in collecting data for sign language recognition

An in-vehicle infotainment system, smart home information access and device control unit, or mobile system presents summarized information to a user based on a user preference model that is associated with the user. The system modifies the presentation of information to the user based on environmental context data about the vehicle and user context data about the activity of the user. During presentation of the information, the system modifies the content and presentation of the summarized information in response to multi-modal input requests from the user.

System and method for dialog-enabled context-dependent and user-centric content presentation

A method for sign language recognition includes receiving, with an input device, an input based on a plurality of hand movements and postures of a user that correspond to a sequence of signs, extracting a plurality of features from the input corresponding to the plurality of hand movements and postures, identifying, a start of the sequence of signs in the input based on a first set of features in the plurality of features and a first Hidden Markov Model (HMM) stored in the memory and identifying a first sign in the input based on a second set of features in the plurality of features and a second HMM stored in the memory. The method also includes generating an output corresponding to the first sign from the input.

System and method for automated sign language recognition

A method of providing information assistance services includes generating a plurality of service requests for a plurality of request elements that are generated from a single client request received by a processor. The service requests are sent to both software application service providers that are executed by the processor and remote service providers that are connected to the local processor through a data network. The processor receives a plurality of service responses from the service providers, generating at least one output message element corresponding to the service responses, and sending the output message data to at least one output device that is operatively connected to the processor to produce a response to the client request.

System and method of conversational assistance in an interactive information system

A method for speech recognition of partial proper names is described which includes natural language processing (NLP), partial name candidate generation, speech recognition and post processing. Natural language processing techniques including shallow and deep parsing are applied to long proper names to identify syntactic units (for example, noun phrases). The syntactic units form a basis for generating a candidate list of partial names for each original full name. A partial name is part of the original name, with some words omitted, or word order changed, or even word substitution. After candidate partial names are generated, their phonetic transcriptions are incorporated into a model for a speech recognizer to recognize the partial names in a speech recognition system.

Speech recognition of partial proper names by natural language processing

A speech recognition method includes providing a processor communicatively coupled to each of a local speech recognition engine and a server-based speech recognition engine. A first speech input is inputted into the server-based speech recognition engine. A first recognition result from the server-based speech recognition engine is received at the processor. The first recognition result is based on the first speech input. The first recognition result is stored in a memory device in association with the first speech input. A second speech input is inputted into the local speech recognition engine. The first recognition result is retrieved from the memory device. A second recognition result is produced by the local speech recognition engine. The second recognition result is based on the second speech input and is dependent upon the retrieved first recognition result.

Methods and systems for adapting grammars in hybrid speech recognition engines for enhancing local SR performance

In accordance with one embodiment, a method of generating language models for speech recognition includes identifying a plurality of utterances in training data corresponding to speech, generating a frequency count of each utterance in the plurality of utterances, generating a high-frequency plurality of utterances from the plurality of utterances having a frequency that exceeds a predetermined frequency threshold, generating a low-frequency plurality of utterances from the plurality of utterances having a frequency that is below the predetermined frequency threshold, generating a grammar-based language model using the high-frequency plurality of utterances as training data, and generating a statistical language model using the low-frequency plurality of utterances as training data.

Speech recognition using multiple language models

Embodiments of an interface system that enables a call center agent to access and intervene in an interaction between an automated call center system and a caller whenever necessary for complex application tasks is described. The system includes a user interface that presents the agent with one or more categories of information, including the conversation flow, obtained semantic information, the recognized utterances, and access to the utterance waveforms. This information is cross-linked and attached with a confidence level for better access and navigation within the dialog system for the generation of appropriate responses to the caller.

System and method for interacting with live agents in an automated call center

Representation-neutral dialogue systems and methods ('RNDS') are described that include multi-application, multi-device spoken-language dialogue systems based on the information-state update approach. The RNDS includes representation-neutral core components of a dialogue system that provide scripted domain-specific extensions to routines such as dialogue move modeling and reference resolution, easy substitution of specific semantic representations and associated routines, and clean interfaces to external components for language-understanding (i.e., speech-recognition and parsing) and language-generation, and to domain-specific knowledge sources. The RNDS also resolves multi-device dialogue by evaluating and selecting among candidate dialogue moves based on features at multiple levels. Multiple sources of information are combined, multiple speech recognition and parsing hypotheses tested, and multiple device and moves considered to choose the highest scoring hypothesis overall. Confirmation and clarification behavior can be governed by the overall score.

Dialogue management using scripts and combined confidence scores

Embodiments of an ontological determination method for use in natural language processing applications are described. In one embodiment, shallow lexico-syntactic patterns are applied to identify relations by extracting term features to distinguish relation terms from non-relation terms, identifying coordinate relations for every adjacent terms; identifying short-distance ontological (e.g., hypernym or part-whole relations) for other adjacent terms based on term features and lexico-syntactic patterns; and then inferring long-distance hypernym and part-whole relations based on the identified coordinate relations and the short-distance relations.

Growing community of inventors

Sunnyvale, CA, United States of America

Kui Xu