| With the rapid development of the Internet, social media is becoming more and more widely used as part of people’s lifestyle. Data accumulated through social media is experiencing explosive growth. Compared to traditional investigation method of the questionnaire in the sociological research, obtaining information through mining social media text can get more realistic, large amount of data, and has the advantages of low cost, so it is more and more widely used.But in the social media text mining, traditional lexicon method has the problem of low accuracy. To solve this problem, this paper presents a dependency parsingbased method and a method based on machine learning, which can use more natural language processing information such as part of speech tagging and dependency parsing information in addition to the word segmentation information. Experiment was carried out on the eating behavior recognition and it proves that the dependency parsing-based method has increased in accuracy compared to the lexicon method and the method based on machine learning has higher accuracy compared to the other two methods, although the recall rate is lower than the lexicion method, it is still the best overall performance of the models.Using the model based on machine learning, diet behavior can be identified from large scale microblog corpus and be corresponded to the original microblog attributes. And then, we conducted an eating habits analysis and achieved results on gender, region, time, including cross-analysis results, and used the form of the word cloud to visualize our results. And the heat distribution of a food in region and time is calculated, which is also description of the diet habits from a certain degree. In addition, we explores how to analyze diet expression habits of different user groups. Finally, several dimensions are used to describe diet expression habits, and the results of the analysis of diet expression habits are demonstrated using word clouds. |
PDF Full Text Request