Method for semantic matching, learning method for semantic matching model, and server

Abstract

Embodiments of the invention disclose a method for semantic matching. The method comprises: receiving a to-be-matched first statement; inputting the first statement as a word segmentation model, through the word segmentation model, outputting a first word sequence after word segmentation is performed on the first statement; converting the first word sequence to a first word vector sequence; inputting the first word vector sequence to a twinborn bidirectional long short-term memory LSTM network, and outputting a second statement matched with the first statement through the twinborn bidirectional long short-term memory LSTM network. An embodiment also provides a semantic matching server, used for improving accuracy of semantic matching.

Description

technical field [0001] The application relates to the field of computer applications, in particular to a semantic matching method, a model learning method and a server. Background technique [0002] With the rapid development of mobile terminals, the mobile Internet has gradually become an important part of the Internet. The way users obtain information is also constantly changing, such as the question answering system, through which users can obtain the information they need. Question answering systems are also an important part of the field of natural language processing. [0003] The current question answering system may mainly include a question understanding part and an answer generation part, wherein the question understanding part is mainly responsible for analyzing and understanding questions raised by users, and generating some candidate query items to assist subsequent modules. Understanding user intent is the key, and it is also the core of the question-answerin...

AI Technical Summary

Problems solved by technology

In the current technology, the learning model receives the sequence of input questions. When processing the input data at this moment, it is only affected by the current input word and the input words before this moment. However, there are contextual connections in the sentences spoken by people in daily life. , resulting in inaccurate matching with the input sequence, thus reducing the matching degree between the output answer and the question

Images