Document sentiment analysis system and method with automatic updating function

Abstract

The invention discloses a document sentiment analysis system and method with an automatic updating function. The method comprises the following steps of: collecting initial sentiment words, calculating a vector value of each initial sentiment word and constructing an initial sentiment word bank according to the initial sentiment words and the corresponding vector values; when document sentiment analysis is carried out, extracting sentiment words in a target document, matching the extracted document sentiment words with the initial sentiment words according to the vector values and calculating vector value matching degrees between the document sentiment words and the initial sentiment words; carrying out statistical scoring on the document sentiment words according to the vector value matching degrees so as to obtain a sentiment value of the target document; and adding the document sentiment words, the vector value matching degrees of which are greater than a preset threshold value, into the sentiment word bank to obtain an updated sentiment word bank. According to the system and method disclosed by the invention, the correctness and objectivity of document sentiment scoring can be improved, and the sentiment word bank can be automatically updated, so that the management is more convenient.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to an automatically updated document sentiment analysis system and a corresponding method. Background technique [0002] The commonly used document emotion recognition technology usually adopts: first collect a large number of emotional words, such as happy, excited, angry, sad, etc., and quantify the score of each emotional word, and then form these words into an emotional lexicon. When using it, use the emotional thesaurus to match the target document. When the emotional word in the emotional thesaurus matches the word that appears in the target document, it will be scored according to the quantified score of the word quantified before, and finally the scores will be summarized. Get the sentiment index of the document. [0003] Examples are as follows: [0004] Happy (1 point), Angry (-1 point)... [0005] There are 50 times of happiness and 10 times of anger in...

AI Technical Summary

Problems solved by technology

[0006] If the full score is 100 points, when the score is above 80, above 90 or negative below 90, below 80, the result is relatively accurate; however, if the score is around 50, it is difficult to judge, because there will be In many cases, for example, the entire document contains neutral words, or half of the positive emotional words and half of the negative emotional words, the final score may be 50 points

[0007] Therefore, there are many disadvantages in adopting the above-mentioned document sentiment scoring method. First, the sentiment lexicon cannot be updated automatically, and it is necessary to artificially add sentiment words into it, and artificially assign scores to it, which affects the objectivity of the results; secondly, pure scores do not It may not be able to fully reflect the real situation of the sentiment of the document

Images