Social Intent Recognition And Classification In Weibo

With the increase of the size of netizens and the rapid development of Weibo platforms,the scale of Weibo users has shown a rapid growth pattern,and a large amount of online social data has been generated,providing a data foundation for analyzing Weibo users.The analysis of Weibo users' relationships,interests,and behaviors has gradually become a hot topic in academic circles.Among them,most blog posts posted by Weibo users have a certain social intent.Social intent analysis refers to mining an intent that users want to express through text.We have analyzed a large number of Weibo studies and found that the social intent of Weibo is basically valid.It is divided into marketing recommendation,news comment,knowledge dissemination,psychosocial awareness,and daily sharing.Based on social intent recognition on Weibo,accurate classification can provide basic support services for related applications.For example,the user's blog posts can be used to determine his identity background and social purpose.Users who always broadcast marketing advertisements can infer that they are merchants.The purpose is to sell products to followers and achieve profitability.Frequent promotion and academic research Users of relevant information can be classified as researchers,whose purpose is to maintain academic reputation and disseminate knowledge.Now that Weibo is the mainstream social media,in addition to judging the user 's identity background and social purpose,social intent recognition and classification can also be used as the basis for distinguishing styles,providing behavioral dimension references for the style portraits of Weibo users,such as: The level of style of the publishers who promote ads in the circle of friends will not be very high.It can also add new dimensions to the classification of Weibo users and provide accurate support for Weibo blog posts.In order to more fully and accurately mine the social intentions of Weibo users,the original information of Weibo users and the reposted Weibo content are comprehensively considered.In this paper,the recognition is transformed into a technical simplification combining semantic mining and text classification.It reduces complexity and has little effect on downstream applications.It is a very clever processing method.From the perspective of text processing,social intent recognition is based on the sentence level and uses the semantic features of the sentence for classification,and the keywords in the sentence assist the definition of the intent label.Therefore,a method based on word and sentence joint training encoder and decoder model to identify intents of short texts on Weibo is proposed.In addition to sentences that can reflect intent categories in microblog texts,nouns and verbs in sentences can also be used for intent categories.Provide a certain guidance,so we use the method of word and sentence joint modeling to more accurately identify the social intent categories of Weibo short text;finally,in order to more accurately identify the social intent categories of Weibo users,we analyze A large amount of Weibo user information builds a library of intent categories.A large number of real Weibo users' blog posts were collected as data sets by the python crawler program,and the experiments performed by us verified the effectiveness of the proposed encoder-decoder model based on word-sentence joint training.The experimental results show that the encoder and decoder model based on the joint training of words and sentences can accurately identify the social intent categories of Weibo users' blog posts;and compare the effectiveness of this method and the BERT pre-trained language model on this task.Experiments show that this method can obtain the best accuracy and the best F1 value on our data set.