Sentiment analysis with echo state network and augmented water cycle algorithm.

Area of Science:

• Natural Language Processing
• Machine Learning
• Computational Linguistics

Background:

• Sentiment analysis is crucial for understanding public opinion but remains challenging due to data complexity.
• Existing machine learning models often struggle with noisy text data, including misspellings and abbreviations found in datasets like IMDb movie reviews.

Purpose of the Study:

• To develop and evaluate a novel sentiment analysis model using an Echo State Network (ESN) optimized with the Augmented Water Cycle Algorithm (AWCA).
• To assess the performance of the proposed ESN-AWCA model using both GloVe and Word2Vec word embedding techniques on the IMDb movie reviews dataset.

Main Methods:

• The study involved extensive data preprocessing of 50,000 IMDb movie reviews, including cleaning, tokenization, stemming, and part-of-speech tagging.
• Two word embedding models, GloVe and Word2Vec, were employed for text vectorization.
• An Echo State Network (ESN) was utilized and optimized using the Augmented Water Cycle Algorithm (AWCA) for hyperparameter tuning.

Main Results:

• The ESN-AWCA model achieved high performance metrics, with GloVe yielding 96.37% F1-score, 96.39% accuracy, 95.87% recall, and 96.87% precision.
• Using Word2Vec, the model achieved 96.23% F1-score, 96.12% accuracy, 95.76% recall, and 96.71% precision.
• Statistical validation (p=0.001, d>1.1) confirmed the model's significant superiority over other evaluated methods.

Conclusions:

• The proposed ESN-AWCA model demonstrates robust and superior performance for sentiment analysis tasks, particularly on challenging datasets.
• Both GloVe and Word2Vec word embeddings are effective when used with the ESN-AWCA architecture, highlighting the model's versatility.
• The optimized ESN-AWCA approach offers a promising solution for accurate and reliable sentiment classification in natural language processing.