CLANG: A CHARACTER GRAM MULTILINGUAL GRAPH-BASED MODEL FOR SENTIMENT CLASSIFICATION

For extended texts, such as full-length publications or news stories, traditional methods of sentiment analysis are formulated generally. Microblogs like Facebook, the most common microblogging site, often find short texts that have some nuances such as using slang phrases, abbreviations, broad noises and multilingualism. Past methods for classifying emotions cannot isolate essential features and result in low classification accuracy when applied to short texts. A character gram multilingual graph-based model named Clang for emotion classification is proposed in this paper for dealing with the above-mentioned problems. The suggested solution mainly used N-gram character graph to represent the data on the microblog. The Clang offers a method for managing the multilingual and noisy data. When doing content analysis, the order as well as the location of characters and words in a text is important. Thus, the method suggested extracts the longest possible subsequence of characters and their degree of similarity to extract the significant features of textual details. The PageRank algorithm is used to evaluate the edge weights, and the long-short term memory (LSTM) deep neural network technique is used to analyse sentiments. The Twitter dataset is used to test and equate the character N-gram Clang output to standard word n-gram model. The experimental result shows that the LSTM Clang model performs better than the traditional method.