Research of Chinese Keyword Extraction Based on Weibo Information

Weibo content has the characteristic of complex and changeable. And the traditional technology of keyword extraction has the limitation of ignoring the meaning of words. This paper combines different weights of speech to improve the TF-IDF formula and uses deep learning method to train text language model. Transforming the process of Weibo content into vector computation in vector space, the semantic similarity between words can be obtained by calculating the similarity between vectors. Then, the similarity and term frequency features of words in the sliding window are added to Textrank's iterative formula for edge weight optimization. By changing the number of keyword extraction and the size of sliding window, the optimal extraction effect of the algorithm is obtained. The algorithm proposed in this paper can solve the problems of the traditional Textrank algorithm, such as equal probability jumping of keyword extraction and the lack of meaning. It has the advantages of low time complexity and less corpus training.