Deep learning-based sentiment analysis of gaming tweets on twitter using LSTM and BiLSTM models

The fast development of social media platforms has created a massive amount of user-generated text, especially in the form of short posts like tweets. These tweets are a valuable source of information because they reflect users’ opinions, experiences, and preferences. This is particularly useful in the gaming industry, where players regularly share feedback about games, updates, and their overall experience. However, tweets are often unstructured and include informal words, abbreviations, emojis, and unnecessary noise, which makes manual analysis difficult and time-consuming. Sentiment Analysis, which is a major area of Natural Language Processing (NLP), helps in automatically detecting and classifying the emotions present in textual data. This study focuses on analysing tweets related to gaming in order to understand user opinions and behaviour patterns. The dataset used for this work contains nearly 43,000 tweets, which are categorized into positive and negative sentiments.To make the data suitable for analysis, several preprocessing techniques are applied. These include cleaning the text, normalizing words, breaking sentences into tokens, removing stop words, and applying stemming. These steps help transform raw and noisy data into a more structured format. After preprocessing, the text data is converted into numerical form using methods like sequence padding and word embeddings.In this research, a deep learning approach based on the Bidirectional Long Short-Term Memory (BiLSTM) model is used for sentiment classification. Unlike traditional machine learning techniques that require manual feature selection, the BiLSTM model can understand the context of the text by processing it in both directions. This allows the model to capture complex patterns in language, such as negation and contextual meaning.The model is trained using the Adam optimizer along with the Binary Cross-Entropy loss function. Dropout is also used to reduce overfitting and improve generalization. The performance of the model is evaluated using metrics such as accuracy, precision, recall, and F1-score. The experimental results indicate that the proposed BiLSTM model achieves strong performance and provides better results compared to traditional methods for sentiment classification.