Research On Personalized Tag Extraction Of The Users Of Micro-blog

Because the users' tags of micro-blog can reflect the characteristic and the preference of the users, and the tags of the users have potential values to the recommendation of the advertisement to the users, the clustering of the users and the search of the users. So it is meaningful to study the user's tag extraction. The personalize of the personalized tag in this paper have two meanings, the first meaning is that the tag can reflect the individual character of the micro-blog users, the second meaning is that the tag will have some individual character themselves. In this paper, we estimate the level of how better the tags extracted by the machine can reflect the user's individual characters by compare the tags extracted by the machine to the tags extracted by the micro-blog users themselves. The individual character of the tags refers to that we will do more works to the tags extracted by the machine, namely we will add some classification attributes to the tags extracted by the machine. The classification attributes can make the tags of the users have the same attributes, and this can make the search of the users of micro-blog more easily, and this can also be used to make clustering of the users of micro-blog. Through studying the tags extracted by the micro-blog users found that there are three kinds of tags of basic types in the tags extracted by the micro-blog users themselves, they are called basic tags, classified tags and followed tags respectively in this paper, then through studying the characteristics of the tags of each basic types designed the extract method for the tags of each basic types, and then studied how to comprehensive the tags ofthe basic types together to get the users' personalized tags better. So in this paper there are seven methods in the procedure of getting the user's tags of micro-blog, among them, the tags that extracted by three methods are basic types, and the other four kinds of tags are mixed tags mixed by the tags of the three basic types, except the method of extracting the basic tags of the basic types that used in this paper is the algorithm of TextRank, the other six methods of extracting the user's tags are first proposed in this paper. According to the experiment, we found that the mixed tags mixed by the tags of the three basic types is the best in these seven kinds of tags. So, there are some improvement in the user's individual tag extraction with the method proposed in this paper. Besides, after the tags that extracted by the machine appended some classified tags, it makes the user's tags have the same attributes, and this makes the clustering of the users, the classifying of the users, and the searching of the users more easily,so this makes the use of the personalized tags of the users of micro-blog more widely.