Twitter Sentiment Analysis Using Social-Spider Lex Feature-Based Syntactic-Senti Rule Recurrent Neural Network Classification

Social platforms have become one of the major sources of unstructured text. Investigating the unstructured text and interpreting the meaning is a complex job. Sentiment Analysis is an emerging approach as the social platforms have lot of opinionated data.(1) It uses language processing, classification of texts and linguistics to retrieve the opinions from the text. Twitter is a micro blogging site which is popular amongst the social users as it is a vast open data-platform and it witnesses lot of sentiments. Twitter Sentiment Analysis is a process of automatic mining of user tweets for opinions, emotions, attitude to derive useful insights into community opinions and classify the opinions as well. Due to the enormous increase in the number of collaborative tweets, it has become complex to identify the terms that carries sentiments. Also, the unstructured tweets may have non-relevant terms and reduce the classification accuracy. To address these issues, we propose a Social-Spider Lex Feature Ensemble Model-Based Syntactic-Senti Rule prediction Recurrent Neural Network Classifier ((SLFEM)-L-2-S2RRNN) to obtain better classification accuracy. Twitter is used as source of data and we have extracted the tweets using Twitter API. Initially, data pre-processing is done to remove unwanted data, symbols and content terms are extracted to improvise the dataset. Then, the significant lexical content terms are extracted employing the proposed Social Spider Lex Feature Ensemble Model ((SLFEM)-L-2) based on Syntactic-Senti Rule Prediction. The semantics(4) of the terms are analysed on the verbs, subjectivity of the tweet patterns to count the overall weightage of tweets. Based on tweet weightage Recurrent Neural Network is trained to classify the tweets int to positive, negative and neutral. The experiment results show that the proposed classifier outperforms the existing models for sentiment classification in terms of accuracy with a performance score 94.1%.