Method for extracting science and technology text problem method based on extremely simple abstract strategy

Abstract

The invention relates to a computer technology, in particular to a method for extracting a science and technology text problem method based on extremely simple abstract strategy, which comprises the following steps of: obtaining a science and technology document data set; preprocessing the unstructured text to obtain a training corpus label; utilizing a BERT pre-training model to carry out vectorization representation on the preprocessed text; constructing a deep neural network of a seq2seq architecture by adopting a Transformer model to serve as an encoder and a decoder, and generating an extremely simple abstract for limiting contents and styles; and extracting problem method words in the generated extremely simple abstract by applying part-of-speech analysis and syntactic analysis algorithms. The method comprises data crawling, natural language processing and deep learning, large-scale science and technology text automatic processing can be achieved, and problem words and method words with corresponding relations are extracted from the science and technology text automatic processing.

Description

technical field [0001] The invention belongs to the field of computer technology, and in particular relates to a method for extracting scientific and technical text questions based on a minimalist summary strategy. Background technique [0002] The increasing number of accessible digital book resources has made it increasingly difficult to accurately retrieve information and quickly acquire knowledge. In order to facilitate the indexing of documents and the acquisition of knowledge, the existing symbol system has formulated a wide range of classification and indexing frameworks to improve retrieval efficiency. However, the retrieval strategy with documents as granular units cannot meet the fine-grained and oriented knowledge acquisition needs of readers. Studies have shown that the information acquisition behavior of researchers is often driven by goals and tasks, and they pay more attention to specific content such as problems, methods or results in the literature. Theref...

AI Technical Summary

Problems solved by technology

However, the problem method extraction under the mode of "manually labeled corpus + machine learning algorithm" relies on large-scale, high-quality labeled corpus, and the high cost of training data acquisition makes the performance of the model quite constrained

Secondly, for scientific and technological texts involving multiple problems and methods, it is difficult for existing methods to distinguish the corresponding relationship between problems and methods.

Images