Tendency analysis method of public opinion based on word2vec

Abstract

The invention provides a tendency analysis method of a public opinion based on word2vec. The method comprises a vector training phase, a key sentence extraction phase and a tendency discrimination phase; by extracting news key sentences, the discriminant feature space is reduced, contents with relatively large relevance with the original theme are remained, useless information is eliminated and the accuracy of tendency analysis of the public opinion is improved; a depth learning model word2vec is introduced into the tendency analysis of the public opinion, and used for comparing the semantic similarity between words and comparing the semantic similarity through word vectors, so that words with the same emotional tendency but not in an emotion dictionary can be well identified, thus even if the emotion dictionary is not complete, a better analysis effect can be obtained, and meanwhile, weighted calculation is performed on the emotional tendency of the key sentences by fusing a grammatical rule; and combined with contextual information, the limitation of simply using the semantic similarity is compensated, and the tendency is analyzed integrally from the sentences, so that the emotional tendency and emotional intensity of news texts of the text level are accurately discriminated.

Description

technical field [0001] The invention relates to the field of computer applications, in particular to a word2vec-based public opinion tendency analysis method. Background technique [0002] Public opinion is the abbreviation of "public opinion situation". Social attitudes generated and held by people and their political, social, and moral orientations. In the process of news dissemination, more and more tendencies are disclosed to the public. These positive or negative emotional tendencies subtly affect people's views on current events and the trend of public opinion. It is in this situation that the research on the tendency analysis of public opinion was born. It aims to dig out the potential tendency of news through the analysis of news texts, hoping to help the public understand the latest social trends, current affairs hot spots and the latest development of public opinion and public opinion. and trends; hope to help regulatory authorities discover social problems in a ...

AI Technical Summary

Problems solved by technology

There are several main problems in the bag-of-words model: 1) The dimensionality of the sample space is too high, which increases the complexity of calculation and easily causes the phenomenon of "dimension disaster"; 2) the bag-of-words model does not consider the semantic relationship between words, For example, "computer" and "computer" are actually the same thing, but the similarity between them is 0 through the bag-of-words model, and the granularity of the evaluation object in the prior art is at the word level, which cannot be achieved at the chapter level. Therefore, there is an urgent need for a new technical means that can realize the judgment of the emotional tendency and emotional intensity of article-level news texts.

Images