A Deep Learning based Approach for Analyzing the Sentiments of Financial Text

Data mining and computational semantics include subfields such as sentimental text analysis. Information collected from opinion-rich sources, such as the news, social networks, review websites, etc., must be analyzed and interpreted. The manual examination of data is no longer practical in today’s fast-paced environment, as data volume is constantly growing while remaining vital to all enterprises. Therefore, there is a need to employ AI and other data-mining methods. Financial Market fluctuations are due, in part, to the gains or losses a company experiences, among other factors. To better comprehend the market scenario, this paper seeks to create a novel LSTM-based algorithm for analyzing financial language and presenting its related people’s sentiments. The proposed method employs the LSTM model to analyze texts and categorize their emotional tone as positive, negative, or neutral. Kaggle’s public data is mined for insights. The data is cleaned and preprocessed using the Doc2Vec method to create embeddings. Finally, the LSTM model is fed these embeddings. When pitted against a three-layer CNN, the LSTM model is shown to perform more effectively. It achieved an accuracy of 99.61% during training and 73.17% during validation.