Sentiment Analysis Using Feature Fusion and Convolutional Neural Networks

The a'm of th's study 's to use deep learn'ng methods for sent'ment analys's on a database of well-known documents. It a'ms to show the comparat've success of the prepared algor'thm aga'nst some publ'shed methods and to exam'ne the data dependency us'ng textual datasets. Sent'ment analys's (or op'n'on m'n'ng) 's a method used 'n natural language process'ng to analyze text documents and determ'ne the content of a text 'tem. Natural language process'ng (NLP) 's a branch of art'f'c'al 'ntell'gence (AI) that enables computers to understand and learn human language. When each word 's expressed as a vector, the mean'ngs of the words w'll be stored 'n the vectors, and the vectors of the words that have close mean'ngs w'll be close to each other. Words are converted 'nto mathemat'cal express'ons and the'r mean'ngs are reduced to data that computers can process by express'ng the'r prox'm'ty to another word mathemat'cally. In th's study, we a'med to use Word2Vec, LSA and CNN three popular text-based feature extract'on methods used 'n natural language process'ng and mach'ne learn'ng. Word2Vec 's tra'ned on a large text dataset to obta'n word vectors, ensur'ng that s'm'lar-mean'ng words are close together 'n the vector space. In th's way, the relat'onsh'ps and mean'ng s'm'lar't'es between words are reflected. LSA 's a text m'n'ng method used to extract semant'c content from large text datasets. LSA uses matr'x factor'zat'on techn'que to d'scover the structure between documents and words. We a'm to perform sent'ment analys's us'ng deep learn'ng models such as TextCNN and Conv1D. TextCNN and Conv1D are convolut'onal neural networks effect've for feature extract'on and class'f'cat'on 'n text data. These models are used to extract features from text data and conduct sent'ment analys's. The datasets used 'nclude Sent'ment140, YELP, Amazon, and IMDB.