New event theme extraction method

Abstract

The invention belongs to the technical field of network information and relates to a new event theme extraction method. The vectorization representation is carried out on the news event text data setbased on BERT; the context connection is closer; expression ways are more accurate, a bidirectional long-short memory network of an attention mechanism is utilized to learn news texts with large datavolume in the network; to discover new events, efficient and accurate utilization of the data is realized; compared with a single mode, a mode of combining a supervised method and an unsupervised method is more efficient, the method is simple, semantic information can be extracted deeply, news texts in a network can be analyzed and mined, new events can be discovered, real-time mastering of the new events by related supervision departments and personal users is facilitated, and subsequent work is facilitated.

Description

Technical field: [0001] The invention belongs to the field of network information technology, and relates to a method for extracting new event topics, especially a bidirectional long-short memory network training new event discovery model based on BERT and attention mechanism and multi-feature fusion topic modeling analysis to extract new event topics Methods. Background technique: [0002] With the development of the Internet in the era of big data, people are surrounded by a large number of news information from a wide range of sources, such as newspapers and the Internet. The most common carrier of news is text, which is the easiest way to obtain valuable information. Due to the variety of news information generated from different sources, the format of news texts and the information contained in them are often messy, and the amount of news information generated is also extremely large. It is almost impossible to completely rely on manual detection of Chinese news events....

AI Technical Summary

Problems solved by technology

However, this method still has the following disadvantages: first, this method can only be used for specific domain data sets, and is not suitable for general data sets in various fields; Professionalism; Finally, this method uses deep learning methods to label and classify text data. This operation can only be aimed at specific fields, and the performance of new data models in new fields is poor.

Images