Research On Key Technologies Of The Construction Of Sentiment Knowledge Graph In Social Network

Sentiment analysis is an important research,which monitors public opinion based on social network.Text-based multi-dimensional sentiment analysis belongs to the research field of natural language processing technology.Meanwhile,knowledge graph plays an important role in research of natural language processing.However,there is a lack of research on Sentiment knowledge Graph in China and abroad.Name Entity Recognition and Relation Extraction are the most basic steps when constructing the Knowledge Graph.The entities,attributes,and relations in Sentiment knowledge Graph are different from the traditional Knowledge Graph,which lacks consideration of the semantic features of the sentiment.Therefore,our study focuses on Named Entity Recognition and Relation Extraction of Sentiment knowledge Graph.In summary,the traditional research work is difficult to meet the needs of Sentiment knowledge Graph construction.In this paper,we offer a model with multilayer neural network,it is suitable for Named Entity Recognition of Sentiment knowledge Graph.In addition,we convert Relationship Extraction into two subtasks,which share an encoder in order to extract attitudes at different levels.The main work of this paper includes:(1)We offer the BBC-LSTM model,which takes emotional semantics into account.In addition,this model utilizes their relations between entity and opinion terms.(2)We offer the SE-Learning model,which models Relation Extraction as two tasks:sequence labeling and text classification.This model can extract attitudes at different levels in the same time.(3)We conduct different experiments on datasets to verify that our model achieves better performance for Relation Extraction and Named Entity Recognition than tradition models.Meanwhile,it is analyses that the shortcomings of models.