A scenic spot abnormal event extraction method based on a composite neural network

Abstract

The invention discloses a scenic spot abnormal event extraction method based on a composite neural network. The method comprises the following steps of carrying out data preprocessing on an obtained event original text corpus, converting an event sentence into a word vector, transmitting a sequence of the word vector to a bidirectional long-short time memory network, and training by utilizing thebidirectional long-short time memory network to obtain semantic characteristics of each candidate trigger word; transmitting the event sentence sequence represented by the word vector into a convolutional neural network, and performing training by using the convolutional neural network to obtain the global features of the event sentence where the candidate trigger word is located. According to themethod, the semantic features of candidate trigger words and the global features of sentences where the candidate trigger words are located are synthesized, the softmax is used as a classifier to classify each candidate trigger word, so that the trigger words of scenic spot abnormal events are found out, and the event types are classified according to the manually marked trigger word types. According to the method, the scenic spot abnormal events can be quickly and accurately extracted, the abnormal events in complex redundant texts can be processed, the efficiency is high, and the universality is good.

Description

technical field [0001] The invention relates to the technical field of natural language processing, in particular to a method for extracting abnormal events in scenic spots based on a composite neural network. Background technique [0002] With the popularity of domestic computer Internet technology and the increasing number of tourists in scenic spots, the event monitoring inside the scenic spot is also facing increasing pressure. How to extract and classify useful abnormal events from the massive information texts obtained has become an urgent problem to be solved. As a part of information extraction, event extraction is a research hotspot in information extraction, and its research content is to automatically obtain specific types of events and elements from natural texts. [0003] Extracting corresponding events from text is usually achieved by identifying event trigger words, so artificially marked event trigger words are the key elements of time recognition. [0004]...

AI Technical Summary

Problems solved by technology

[0004] In the current extraction methods for abnormal emergencies, text frames are often used for formatted extraction and support vector machine SVM classifiers for text classification, but these methods cannot achieve good results for nonlinear data sets. Classification effect, and rely too much on trigger words, resulting in judgment errors when candidate trigger words are vague; moreover, frequent event information cannot standardize grammar, and the current method lacks research on event extraction from non-standard sentences

Images