International Business Machines Corporation

Nuance Communications, Inc.

Ladislav Kunc

Saloni Potdar

Jan Kleindienst

Ming Tan

Yang Yu

Haoyu Wang

Martin Labsky

Tomas Macek

Haode Qi

Martin Labský

Jan Vystrcil

Tomás Macek

David Nahamoo

Raimo Bakis

Lazaros Polymenakos

John Zakos

Donna Karen Byron

Mary Swift

Benjamin L Johnson

Nils Lenke

Abhishek Shah

Navneet N Rao

Lin Pan

Bart D'hoore

A method, computer system, and computer program product for multi-lingual chatlog training are provided. The embodiment may include receiving, by a processor, a plurality of data related to conversational data in multiple languages. The embodiment may also include assigning an intent label to each conversational data. The embodiment may further include assigning a language label to each conversational data. The embodiment may also include paring the plurality of the data related to the conversational data according to the intent label and the language label. The embodiment may further include training a machine learning model using a multi-lingual and multi-intent conversational data pairing. The embodiment may also include training the machine learning model using a single language and multi-intent conversational data paring.

Conversational AI with multi-lingual human chatlogs

Mechanisms are provided for optimizing an automated machine learning (AutoML) operation to configure parameters of a machine learning model. AutoML logic is configured based on an initial default value and initial range for sampling of a parameter of the machine learning (ML) model and an initial AutoML process is executed on the ML model based on a plurality of datasets comprising a plurality of domains of data elements, utilizing the initially configured AutoML logic. For each domain, a cross-dataset default value and cross-dataset value range are derived from results of the execution of the initial AutoML process. For each domain, an entry is stored in a data structure, the entry storing the derived cross-dataset default value and cross-dataset value range for the domain. The AutoML logic performs a subsequent AutoML process on a new dataset based on one or more entries of the data structure.

Mechanisms for continuous improvement of automated machine learning

A mechanism is provided to implement an abnormal entity detection mechanism that facilitates detecting abnormal entities in real-time response systems through weak supervision. For each first intent from an entity labeled workspace that matches a second intent in labeled chat logs, when the entity score associated with each first entity or second entity is above a predefined significance level the first entity or the second entity is recorded. For each first intent from the entity labeled workspace that matches the second intent in the labeled chat logs: responsive to the first entity being recorded and the second entity failing to be recorded, that first entity is removed from the training data as being mistakenly included; or, responsive to the second entity being recorded and the first entity failing to be recorded, that second entity is added as a potential business case to the training data.

Weak supervised abnormal entity detection

A computing device receives a natural language input from a user. The computing device routes the natural language input from an active domain node of multiple domain nodes of a multi-domain context-based hierarchy to a leaf node of the domain nodes by selecting a parent domain node in the hierarchy until an off-topic classifier labels the natural language input as in-domain and then selecting a subdomain node in the hierarchy until an in-domain classifier labels the natural language input with a classification label, each of the plurality of domain nodes comprising a respective off-topic classifier and a respective in-domain classifier trained for a respective domain node. The computing device outputs the classification label determined by the leaf node.

Routing text classifications within a cross-domain conversational service

A computer-implemented method for building a semantic analysis model. In one embodiment, the computer-implemented method includes creating proxy tags comprising a set of surface form variants. The computer-implemented method creates training examples comprising a combination of terminal tokens and at least one of the proxy tags. The computer-implemented method builds the semantic analysis model using the training examples.

Abstraction and portability to intent recognition

An online version of a sentence representation generation module updated by training a first sentence representation generation module using first labeled data of a first corpus. After training the first sentence representation generation module using the first labeled data, a second corpus of second labeled data is obtained. The second corpus is distinct from the first corpus. A subset of the first labeled data is identified based on similarities between the first corpus and the second corpus. A second sentence representation generation module is trained using the second labeled data of the second corpus and the subset of the first labeled data.

Updating an online multi-domain sentence representation generation module of a text classification system

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 method includes obtaining a training data set including a plurality of training examples. The method includes generating, for each training example, multiple feature vectors corresponding, respectively, to multiple feature types. The method includes applying weighting factors to feature vectors corresponding to a subset of the feature types. The weighting factors are determined based on one or more of: a number of training examples, a number of classes associated with the training data set, an average number of training examples per class, a language of the training data set, a vocabulary size of the training data set, or a commonality of the vocabulary with a public corpus. The method includes concatenating the feature vectors of a particular training example to form an input vector and providing the input vector as training data to a machine-learning intent classification model to train the model to determine intent based on text input.

Weighting features for an intent classification system

A method includes providing input text to a plurality of multi-task learning (MTL) models corresponding to a plurality of domains. Each MTL model is trained to generate an embedding vector based on the input text. The method further includes providing the input text to a domain identifier that is trained to generate a weight vector based on the input text. The weight vector indicates a classification weight for each domain of the plurality of domains. The method further includes scaling each embedding vector based on a corresponding classification weight of the weight vector to generate a plurality of scaled embedding vectors, generating a feature vector based on the plurality of scaled embedding vectors, and providing the feature vector to an intent classifier that is trained to generate, based on the feature vector, an intent classification result associated with the input text.

Cross-domain multi-task learning for text classification

According to some aspects, a method of searching for content in response to a user voice query is provided. The method may comprise receiving the user voice query, performing speech recognition to generate N best speech recognition results comprising a first speech recognition result, performing a supervised search of at least one content repository to identify one or more supervised search results using one or more classifiers that classify the first speech recognition result into at least one class that identifies previously classified content in the at least one content repository, performing an unsupervised search of the at least one content repository to identify one or more unsupervised search results, wherein performing the unsupervised search comprises performing a word search of the at least one content repository, and generating combined results from among the one or more supervised search results and the one or more unsupervised search results.

Systems and methods for speech-based searching of content repositories

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.

Methods, systems, computer-readable media, and apparatuses for providing search disambiguation using contextual information and domain ontologies are presented. In some embodiments, a computing device may receive a natural language input from a user. The computing device may identify a plurality of hypotheses for the natural language input. The computing device may map the plurality of hypotheses to one or more concepts of a plurality of concepts of an ontology by annotating the one or more concepts. The ontology may include the plurality of concepts respectively connected by a plurality of relations. The computing device may determine that there is an imperfect match between the annotated one or more concepts and annotations of answers. In response, the computing device may disambiguate the annotated one or more concepts using the ontology. The computing device may present output to the user based on the disambiguation.

Contextual search disambiguation

Aspects described herein provide solutions to problems posed by a user. Input that includes a specified subject may be received from a user. A specified descriptor for the specified subject may be obtained from a hierarchical taxonomy associated with the specified subject. An information repository may be searched based on the specific descriptor, and entries stored at the information repository that are associated with the specific descriptor may be indicated in a list of results. The specific descriptor may be iteratively generalized to obtain generalized descriptors, and the information repository may be searched based on the generalized descriptors. The generalized descriptors may also be specialized to obtain specialized descriptors, and the information repository may be searched based on the specialized descriptors. The list of results may include entries stored at the information repository that are respectively associated with the generalized descriptors and the specialized descriptors.

Recommendation system with hierarchical mapping and imperfect matching

Methods and apparatus to receive a first data stream, such as a public broadcast, and receive a second data stream, such as a private data stream, containing emails, for example. A user profile can be used to generate a data output stream for the user from the first and second data streams. The user profile can contain preferences for segments within the first and second data streams.

Methods and apparatus for providing mixed audio streams

Methods, devices, and computer program products for recognizing and responding to natural language input are described herein. Natural language input is received at a natural language input interface of a computing device and transformed into computer-usable text. A natural language input recognizer obtains evidence from one or more evidence source and generates an evidence graph based on the evidence obtained. Evidence may be obtained asynchronously, and the natural language input recognizer may update the evidence graph upon receipt of additional evidence. The natural language input recognizer generates a set of recognition hypotheses based on the evidence graph and selects one of the recognition hypotheses as a recognition result for the natural language input. Semantic models, evidence models, and response models may be employed to generate the evidence graph and respond to the recognition result selected for the natural language input.

Growing community of inventors

Prague, Czechia

Ladislav Kunc