Research On Short Text Emotion Assessment Method Based On Knowledge Graph

The knowledge graph is the most effective method of organizing one’s body of information since it uses graphs to explain the things and the interactions between them in the real world.At the moment,the technology behind Knowledge Atlas is being utilized in a wide variety of cuttingedge research domains,such as intelligent question answering and recommendation systems,amongst others.Data mining of the short text kind is another key topic of research within the field of internet information data mining.In this article,knowledge maps are utilized in the process of brief text emotional perception analysis.This methodology is founded on the reasoning performance enhancement made possible by knowledge map triplet query.The primary research contents and research results presented in this thesis are as follows:(1)We propose a method of sentiment assessment analysis based on tweet graphs to effectively assess emotional polarity with positive and negative attitudes in short texts.This method is based on the following hypothesis: In particular,we use the Twitter graph to represent the knowledge base in the form of a graph.Then,we extract the properties of the graph in order to generate the feature matrix for the machine learning model.In the knowledge graph representation learning process,the similarity of the graph is intended to be related to emotion prediction.In order to get a greater overall effect of assessment,the graph features that relate to the Horn clause and any other logical assertions have been eliminated.(2)In order to improve the effectiveness and precision of the training of short text word vectors,a word vector training model that is based on Skip-Gram has been constructed.The research creates the word vector model on top of the widely used Word2 Vec model in order to achieve the goal of achieving a more effective training impact.This study analyzes and evaluates the effectiveness of several corpus training methods,basing its findings on a variety of tasks that are analogous to one another.According to the findings,the Word2 Vec Skip-Gram model is the most effective way for word vector training.The Word2 Vec Skip-Gram model is selected as the modeling technique for this paper,and the concrete implementation of the model is provided.The results of the experiments demonstrate that the word vector model that was built in this work has improved validity and accuracy,in addition to improved scalability.For instance,it is able to locate appropriate synonyms in order to comprehend the possible new meanings of common phrases found on the internet.(3)The optimization model of sentiment analysis based on the Bi-LSTM model is meant to improve the effectiveness and efficiency of sentiment analysis performed on short texts.The objective of the bidirectional long short-term memory(Bi-LSTM)model is to exploit important information irrespective of the proximity of the information or the nature of the connection between the pieces of information.Quantitative application is the key to unlocking its full potential.The rationality of the information utilization can be improved by paying attention to the significance of the input features and the real-time convergence of the model encoder to certain features of the input sequence.These considerations are then reflected in the calculation of the output vector in the form of weights,which brings the process full circle.According to the findings of the experiment,the FI score,precision,and recall are,in that order,0.675,0.876,and 0.812,respectively.This demonstrates that the model that was built in this paper is accurate.