End-to-end text event extraction method and system based on deep learning

Abstract

An end-to-end chapter event extraction method and system based on deep learning, the method comprising: candidate entity identification, document level coding, candidate entity relationship coding and event decode.The end-to-end chapter event extraction technology proposed by the invention can effectively solve the problem of event element dispersion in a chapter document, and improve the extraction efficiency. Meanwhile, multi-event prediction can be flexibly performed in parallel, candidate word embedding is performed by adopting an attention mechanism, word embedding of [CLS] sentence header identification is used as sentence embedding, semantic representation of word embedding and sentence embedding is well stored, and the final event extraction performance is improved.

Description

technical field [0001] The invention relates to the field of natural language processing in computer data processing technology, in particular to an end-to-end text event extraction method and system based on deep learning. Background technique [0002] With the continuous development of modern information technology, various industries have accumulated a large amount of information data, and the world has entered the era of big data. Accurately obtaining useful information from massive data in real time is of great significance in reality. Event extraction aims to extract structured events from unstructured text. It is widely used in intelligence work in commercial and military fields. Event extraction technology makes it possible to automatically extract structured event information from massive data. Extracting the events and their elements mentioned in the plain text faces two specific challenges. One is that the elements of an event are scattered, and the elements that...

AI Technical Summary

Problems solved by technology

Shun Zheng et al in the document Doc2EDAG: An End-to-End Document-level Framework for Chinese Financial Event Extraction, and Hang Yang et al in the document Document-level Event Extraction via Parallel Prediction Networks.Proceedings of the 59 th Annual Meeting of the Association for Computational Linguistics and the 11 th In the International Joint Conference on Natural Language Processing, all attempts are made to extract chapter events. The events of the former are decoded by way of path extension, and event extraction cannot be performed in parallel. The latter uses a parallel multi-event decoder, which needs to pre-specify the events to be extracted. The number of events is likely to cause problems such as event loss and waste of computing resources. At the same time, the performance of the current chapter event extraction method is low, and the F1 value is lower than 80%, which cannot reach the level of commercial use.

Images