Fmr Corp.

Hua Hao

Byung Chun

Pu Li

Tieyi Guo

Yu Zhang

Dong Han

Chao Yu

Liang Chen

Yucai Yang

Pengrui Sun

Hao Fu

Yuanjun Liu

Yong Zou

Wenbin Ren

Junmei Liu

Chen Bi

Chuanlu Yu

Sijing Lv

Bing Cui

Ou Li

Maohua Sun

Yachao He

Methods and apparatuses for multimodal enhancement of interactions in conversation service applications include a server that establishes a chat-based communication session between a conversation service application and a client computing device. Query input is captured from a user during the session. The server converts an audio segment corresponding to a video into a video transcript using speech recognition, and determines a transcript portion for the video that is responsive to the query input using a dense passage retriever model. The server selects video frames from the video that correspond to the transcript portion and determines, for the selected frames, a predicted relevancy between the selected frame and the query input. The server generates a response to the query input comprising the selected frames based upon the relevancy. The server transmits the response to the client device for display.

Multimodal enhancement of interactions in conversation service applications

Methods and apparatuses are described for enhancing chatbot recognition of user intent through graph analysis. A computing device classifies user messages associated with a chat session as either comprehensible user messages that are mapped to a user intent or incomprehensible utterances that are not mapped to a user intent. The computing device arranges the comprehensible messages into first clusters and arranges the incomprehensible messages into second clusters. The computing device generates a first sub-graph for each first cluster and generates a second sub-graph for each second cluster. The computing device calculates a graph kernel matrix using the sub-graphs and trains a classification model using the graph kernel matrix. The computing device executes the trained classification model using the graph kernel matrix as input to generate a predicted intent identifier for the incomprehensible messages.

Enhancing chatbot recognition of user intent through graph analysis

Methods and apparatuses are described for automatic refinement of intent classification for virtual assistant applications. A computing device creates utterance groups comprising messages. The device determines, using a first machine learning (ML) classifier, a first predicted intent associated with each utterance group and determines, using a second ML classifier, a second predicted intent associated with each utterance group. The device combines the first and second predicted intents to generate a final predicted intent, determines, for each incomprehensible utterance group, whether the final predicted intent overlaps with the final predicted intent for other utterance groups, and selects messages having no overlapping intent to create new intents. The device determines, for each comprehensible utterance group, whether the final predicted intent overlaps with a preassigned intent for other utterance groups and combines the messages that have an overlapping intent to create and label a new utterance group with the overlapping intent.

Automatic refinement of intent classification for virtual assistant applications

Methods and apparatuses are described for unstructured computer text is analyzed for masking of confidential information using artificial intelligence. A client device generates a message comprising unstructured computer text including confidential information. A server trains a word embedding model using the unstructured text. The server generates a multidimensional vector for each word in the unstructured text, generates a mapping table comprising a predetermined set of words corresponding to confidential information from the unstructured text, and determines one or more neighboring words in the trained word embedding model using the predetermined set of words. The server updates the mapping table to incorporate the one or more neighboring words and executes rules on the unstructured text that filter out one or more words, and applies the updated mapping table to match words in the updated mapping table with words in the filtered text and mask the matching words in the unstructured text.

Automated masking of confidential information in unstructured computer text using artificial intelligence

Methods and apparatuses are described for automated computer text classification and routing using artificial intelligence transfer learning. A server trains a word embedding model using one-hot vectors of word pairs from a filtered first corpus of unstructured computer text and a filtered second corpus of unstructured computer text, using an artificial intelligence neural network. The server trains a long short-term memory model using vector matrices that correspond to sentences in the filtered second corpus of unstructured computer text, and labels. The server receives a message, generates a matrix for each sentence in the message by applying the trained word embedding model, generates one or more labels, and a probability for each label, for each sentence in the message by applying the trained long short-term memory model, and routes the message to a second client computing device based upon an assigned label.

Automated computer text classification and routing using artificial intelligence transfer learning

A computer-implemented method of training, using a computer log file, an application error prediction engine to identify one or more application errors includes parsing the computer log file into a plurality of data sets. Each data set is associated with a unique computing session having a session identifier and ending in an application or. The method also includes extracting, from each data set, values for a specified set of parameters in each data set. The method also includes encoding the extracted values for each data set into a corresponding data structure. The method also includes generating, for each data structure, a corresponding vector, the corresponding vectors collectively forming a matrix. The method also includes calculating, based on the matrix, a set of clusters, each cluster corresponding to a known error type, the set of clusters used to create a model used to identify new error types.

Automated log analysis and problem solving using intelligent operation and deep learning

Growing community of inventors

Dalian, China

Hua Hao