International Business Machines Corporation

University of Michigan

Intenrational Business Machines Corporation

Lazaros Polymenakos

David Nahamoo

Raimo Bakis

Chulaka Gunasekara

Jatin Ganhotra

Song Feng

Li Zhu

Sunil D Shashidhara

Cheng Wu

John Zakos

Ponani S Gopalakrishnan

Rangachari Anand

Ladislav Kunc

Dimitrios B Dimitriadis

Kshitij P Fadnis

Emily Mower Provost

Ashima Arora

Mukund Panmanabhan

Zakaria Aldeneh

Stephane Herman Maes

Rui Zhang

Lalit R Bahl

Ramesh Ambat Gopinath

Honglak Lee

David Echeverria Ciaurri

Dragomir Radkov Radev

Nathaniel Mills

Sunil Davangere Shashidhara

A method, apparatus and computer program product for presenting a user interface for a conversational system is described. A unified contextual graph for use by the conversational system, the unified contextual graph comprising components based on database queries of the conversational system and a user dialog between a user and the conversational system. For each of a set of user utterances produced in a dialog with the conversational system, a semantic meaning representation is determined. The semantic meaning representations are converted to respective sentential concept graphs. The unified contextual graph is updated based on new sentential concept graphs while the dialog with the conversational system progresses.

Semantic representation and realization for conversational systems

Systems, computer-implemented methods, and computer program products that can facilitate predicting a source of a subsequent spoken dialogue 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 speech receiving component that can receive a spoken dialogue from a first entity. The computer executable components can further comprise a speech processing component that can employ a network that can concurrently process a transition type and a dialogue act of the spoken dialogue to predict a source of a subsequent spoken dialogue.

End-of-turn detection in spoken dialogues

Embodiments relate to a system, program product, and method directed at natural language (NL) and a virtual dialog platform. An NL statement is detected and analyzed to identify one or more entities expressed in the statement. The identified entities are leveraged to parse the statement into keywords. The intent of the received statement is represented as a relationship between two or more of the keywords. A knowledge representation is identified to represent the statement with respect to a formatted module having two or more components and a component relationship structure. Each statement keyword is assigned to a designated module component based on an alignment of the component relationship with the keyword relationship. The statement intent is expressed based on the relationship between the keywords, and a statement response is inferred. The inferred statement is communicated to the virtual dialog platform.

Conversational agents over domain structured knowledge

Embodiments provide a computer implemented method, in a data processing system including a processor and a memory including instructions which are executed by the processor to cause the processor to train an enhanced chatflow system, the method including: ingesting a corpus of information including at least one user input node corresponding to a user question and at least one variation for each user input node; for each user input node: designating the node as a class; storing the node in a dialog node repository; designating each of the at least one variations as training examples for the designated class; converting the classes and the training examples into feature vector representations; training one or more training classifiers using the one or more feature vector representations of the classes; and training classification objectives using the one or more feature vector representations of the training examples.

System and method for enhanced chatflow application

A computer-implemented conversational system framework to perform tasks associated with a client request. A conversation application executing on a hardware processor provides application workflow orchestration, the conversation application receiving a client request and sending one or more application requests based on the application workflow orchestration. A conversation system executing on a hardware processor provides conversation workflow orchestration, the conversation system receiving the one or more application requests. The conversation application and the conversation system develop dialog context and store the dialog context in a memory device. The conversation application and the conversation system develop the dialog context by invoking at least one micro-service to perform tasks associated with the one or more application requests. The conversation application generates a response to the client request based on the developed dialog context.

Conversational framework

A system and method performs automated domain concept discovery and clustering using word embeddings by receiving a set of documents for natural language processing for a domain, representing a plurality of entries in the set of documents as continuous vectors in a high dimensional continuous space, applying a clustering algorithm based on a mutual information optimization criterion to form a set of clusters, associating each entry of the plurality of entries with each cluster in the set of clusters through formalizing an evidence based model of each cluster given each entry, calculating a mutual information metric between each entry and each cluster using the evidence based model, and identifying a nominal center of each cluster by maximizing the mutual information.

Domain concept discovery and clustering using word embedding in dialogue design

A method, apparatus and computer program product for presenting a user interface for a conversational system is described. A user input is received in a dialog between a user and the conversational system, the user input in a natural language. A domain trained semantic matcher is used to determine a set of entities and a user intent from the user input. One or more queries is generated to selected ones of a plurality of knowledge sources, the knowledge sources created from domain specific knowledge. The results from the one or more queries are ranked based on domain specific knowledge. A system response is presented in the dialog based on at least a highest ranked result from the plurality of knowledge sources.

Implicit dialog approach operating a conversational access interface to web content

A method, apparatus and computer program product for presenting a user interface for a conversational system is described. A unified semantic representation of conversational content between a user and conversational system is created as a contextual graph of concepts and relations. A set of subgraph components of the semantic contextual graph dynamically identified based on a current dialog activity. The identified set of subgraph components in a user interface as a set of graphical elements representing respective concepts and relations.

A method, program product and computer system to predict utterances in a dialog system includes receiving a set of utterances associated with a dialog between a client device and a dialog system, mapping the utterances to vector representations of the utterances, and identifying at least one cluster to which the utterances belong from among a plurality of possible clusters. A next cluster is predicted based upon a conditional probability of the next cluster following a set of a predetermined number of previous clusters using a language model. A next utterance is predicted from among a plurality of possible utterances within the predicted next cluster.

Quantized dialog language model for dialog systems

A method, apparatus and computer program product for creating a dialog system for web content is described. Knowledge is extracted from a target web application for the dialog system. The knowledge includes an organizational structure of the target web application and domain knowledge pertinent to the target web application. A deep learning process associates the domain knowledge with the organization structure of the target application. A plurality of knowledge sources of different respective types are created from the domain knowledge and the organizational structure. Each of the knowledge sources is used for providing answers to user queries to the dialog system. As part of the invention, a semantic matcher is provided to select among the answers provided by the plurality of knowledge sources for a best answer to a user query.

Implicit dialog approach for creating conversational access to web content

Embodiments provide a computer implemented method of training an enhanced chatflow system, comprising: ingesting a corpus of information comprising at least one user input node corresponding to a user question and at least one expert-designed variation for each user input node; matching one or more user inputs to one or more corresponding dialog nodes using regular expressions and delimiters; ingesting one or more usage logs from a deployed dialog system, each usage log comprising at least one user input node; for each user input node: designating the node as a class; storing the node in a dialog node repository; designating each of the at least one variations as training examples for the designated class; converting the classes and the training examples into feature vector representations; training one or more classifiers and one or more classification objectives using the feature vector representations.

An information processing system, a computer program product, and methods for modeling multi-party dialog interactions. A method includes learning, directly from data obtained from a multi-party conversational channel, to identify particular multi-party dialog threads as well as participants in one or more conversations. Each participant utterance being converted to a continuous vector representation updated in a model of the multi-party dialog relative to each participant utterance and according to each participant's role selected from the set of: sender, addressee, or observer. The method trains the model to choose a correct addressee and a correct response for each participant utterance, using a joint selection criterion. The method learns directly from the data obtained from the multi-party conversational channel, which dialog turns belong to each particular multi-party dialog thread.

Modeling multiparty conversation dynamics: speaker, response, addressee selection using a novel deep learning approach

A technique to improve the recognition accuracy when transcribing speech data that contains data from a wide range of environments. Input data in many situations contains data from a variety of sources in different environments. Such classes include: clean speech, speech corrupted by noise (e.g., music), non-speech (e.g., pure music with no speech), telephone speech, and the identity of a speaker. A technique is described whereby the different classes of data are first automatically identified, and then each class is transcribed by a system that is made specifically for it. The invention also describes a segmentation algorithm that is based on making up an acoustic model that characterizes the data in each class, and then using a dynamic programming algorithm (the viterbi algorithm) to automatically identify segments that belong to each class. The acoustic models are made in a certain feature space, and the invention also describes different feature spaces for use with different classes.

Transcription of speech data with segments from acoustically dissimilar

A method and apparatus for removing the effect of background music or noise from speech input to a speech recognizer so as to improve recognition accuracy has been devised. Samples of pure music or noise related to the background music or noise that corrupts the speech input are utilized to reduce the effect of the background in speech recognition. The pure music and noise samples can be obtained in a variety of ways. The music or noise corrupted speech input is segmented in overlapping segments and is then processed in two phases: first, the best matching pure music or noise segment is aligned with each speech segment; then a linear filter is built for each segment to remove the effect of background music or noise from the speech input and the overlapping segments are averaged to improve the signal to noise ratio. The resulting acoustic output can then be fed to a speech recognizer.

Growing community of inventors

West Harrison, NY, United States of America

Lazaros Polymenakos