An implicit inter-sentence relationship analysis method based on multi-task bi-directional long-short time memory network

A technology of long and short-term memory and inter-sentence relationship, applied in the computer field, can solve the problem of low accuracy in identifying the relationship between texts and sentences, and achieve the effect of high practicability and high accuracy.

CN109460466AInactive Publication Date: 2019-03-12UNIV OF ELECTRONICS SCI & TECH OF CHINA
4 Cites 16 Cited by

Patent Information

Authority / Receiving Office
CN · China
Current Assignee / Owner
Publication Date
2019-03-12
Estimated Expiration
Not applicable · inactive patent

Smart Images

  • Figure 1
    Figure 1
  • Figure 2
    Figure 2
  • Figure 3
    Figure 3
Patent Text Reader

Abstract

The invention provides an implicit inter-sentence relationship analysis method based on a multi-task bi-directional long-short time memory network. The method comprises the following steps: obtaininga Chinese text-level semantic relationship corpus, comprising implicit inter-sentence relationship statements and explicit inter-sentence relationship statements; obtaining a Chinese text-level semantic relationship corpus. Using the method of multi-task learning, the input sequence of the model is obtained by using implicit inter-sentence relation recognition task as the main task and explicit inter-sentence relation recognition task as the auxiliary task. Enter both primary and secondary tasks into Bi-LSTM recurrent neural network, through learning the implicit relationship between sentencesrecognition model; The implicit inter-sentence relation recognition model adopts a fusion word embedding method and introduces a priori knowledge so as to make full use of text features and obtain abetter recognition result. The present invention fully utilizes the semantic connection between the implicit sentence-to-sentence relation sentence and the explicit sentence-to-sentence relation sentence, and solves the problem that the implicit sentence-to-sentence relation sentence does not have good characteristics, which leads to bad implicit sentence-to-sentence relation recognition effect.
Need to check novelty before this filing date? Find Prior Art

Description

technical field

[0001] The invention relates to the field of computer technology, in particular to an implicit inter-sentence relationship analysis method based on a multi-task bidirectional long-short-term memory network. Background technique

[0002] Sentence is an important research level besides words, words and phrases in natural language processing. Discourse Rela-tion Recognition is an indispensable part of sentence-level research. The main task of discourse inter-sentence relationship recognition is to study the logical relationship between two consecutive arguments in a text (for example: comparative relationship, extended relationship, parallel relationship and causal relationship, etc.). This task is a fundamental research problem in natural language understanding. Correctly judging the logical relationship between text sentences means being able to effectively understand the semantic relationship of the text.

[0003] The difficulty of identifying the implicit...

Examples

Embodiment Construction

[0011] The implementation of the present invention is divided into two parts: the training of the model and the use of the model. The specific implementation manners of the present invention will be described in further detail below according to the drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0012] figure 1 It is a schematic diagram of the model training framework of an embodiment of the present invention.

[0013] The implicit inter-sentence relationship recognition model based on multi-task learning recurrent neural network is as follows: figure 1 shown. Among them, task1 is the task of identifying the relationship between implicit sentences, and task2 is the task of identifying the relationship between explicit sentences. The model has a total of three Bi-LSTMs. The upper and lower networks are exclusive to task1 and task2, respectively, and are used to e...