Text classification method for research report text

Abstract

The invention relates to a text classification method for research report texts, which comprises the following steps: firstly, collecting a certain number of research reports, and marking collected research report paragraphs to form a sample; sending the labeled samples to a machine learning framework for training to obtain a comprehensive training model; and finally, performing content extractionand text noise reduction processing on an original research and report file to be identified, and finishing extraction and classification of research and report contents by the comprehensive trainingmodel. According to the method, the accuracy of extracting and classifying the research report paragraphs is effectively improved, and the text analysis capability of research reports is improved.

Description

technical field

[0001] This patent application belongs to the field of machine learning technology, and more specifically, relates to a text classification method for research paper texts. Background technique

[0002] At present, the existing mature natural language processing technology can identify the entities in the research reports, and can classify the research reports, such as individual stock research reports, industry research reports, futures research reports, etc., but if you need to classify each research report If each paragraph is classified, for example, a stock research report includes core viewpoints, objective expositions, profit forecasts, and risk warnings, then the existing text classification technology obviously cannot meet the needs.

[0003] At the same time, the current deep learning models mainly include TextCnn, LSTM, FastText and other models. These models are all deep learning models based on neural networks. They are good at single text classi...

Images