Search International and National Patent Collections

1. (WO2018009490) CONVERSATIONAL RELEVANCE MODELING USING CONVOLUTIONAL NEURAL NETWORK

Pub. No.:    WO/2018/009490    International Application No.:    PCT/US2017/040626
Publication Date: Fri Jan 12 00:59:59 CET 2018 International Filing Date: Wed Jul 05 01:59:59 CEST 2017
IPC: G06F 17/27
Applicants: MICROSOFT TECHNOLOGY LICENSING, LLC
Inventors: WU, Bowen
WANG, Baoxun
PENG, Shuang
ZENG, Min
ZHOU, Li
Title: CONVERSATIONAL RELEVANCE MODELING USING CONVOLUTIONAL NEURAL NETWORK
Abstract:
Non-limiting examples of the present disclosure describe a convolutional neural network (CNN) architecture configured to evaluate conversational relevance of query-response pairs. A CNN model is provided. The CNN model comprises: a first branch, a second branch, and multilayer perceptron (MLP) layers. The first branch comprises a plurality of convolutional layers with dynamic pooling to process a query. The second branch comprises a plurality of convolutional layers with dynamic pooling to process candidate responses for the query. The MLP layers are configured to rank query-response pairs based on conversational relevance. The query and the candidate responses are processed in parallel using the CNN model. Pairwise ranking of the query-response pairs are generated using the MLP layers based on a first input propagated from the first branch and second input propagated from the second branch. A ranking of one or more query-response pairs may be output. Other examples are also described.