Research On Keywords Extraction From Weibos Based On Semantic Association Between Image And Text

The keywords extraction is an important technique to express the theme of an article.With the rapid development of Internet,the information increases very fast.The keywords can help people quickly understand the core content of an article.At the same time,keywords extraction also plays an important role in text classification,text clustering,and information retrieval.This paper is aimed to improve weibos' keywords extraction using semantic association between image and text.The text of weibo may not be related with image.In this paper,we first propose a method to reveal the semantic association between image and text,and then propose a method to extract the keywords of weibo using the semantic association of image and text.To reveal the semantic association of Chinese image-weibo,three methods have been used to calculate the semantic similarity and three kinds of machine learning methods have been used to recognize.The experimental results show that the BP neural network combined with the Word-Embedding method to calculate the semantic similarity,adding the textual features and social features can obtain best results.As keywords extraction,three methods are used to calculate text candidate word weights,and then improve the result using image information.The experimental results show that the TFIDF method to extract the keywords of weibo is better.After adding semantic association information,the results are significantly improved.