Machine learning based user profile recognition for popular short videos on social platforms

Considering the emotion and opinion tendentiousness contained in user comment data, a user comment data mining model is based on bidirectional encoder representation from transformers - bidirectional long short term memory - conditional random field (BERT-Bi LSTM-CRF). BERT-Bi LSTM-CRF is designed to obtain text sequence features and annotation terms. In addition, to recognise the emotional polarity of aspect words, a classification model based on BiAtt-GCN is constructed. For the dataset SemEval Task 4, the proposed model achieved an accuracy improvement of 0.83% and 3.2%, respectively, compared to the BiSLTM CNN model and CMLA model, and an increase of 0.68% and 1.26% in the recall index. Therefore, the model proposed in the study is effective in the analysis of user comment data.