Market Sentiment Analysis Using NLP: Understanding Trends and Buyer Preferences in Real Estate and Environmental Sectors

The explosive growth of digital platforms has resulted in an explosion of unstructured textual data, such as customer reviews, social media posts, and feedback forms, which can provide significant insights into consumer preferences and industry trends. To improve market sentiment research in the real estate and environmental sectors, this study investigates the use of Natural Language Processing (NLP) approaches, including word embedding models like Word2Vec, FastText, GloVe, and custom-developed embeddings. The study's goal is to use these models to transform raw textual data into structured insights that reveal consumer attitudes and behavior patterns. The research highlights the unique challenges of sentiment analysis, including handling domain-specific language, contextual nuances, and accurately classifying neutral sentiments. To address these issues, the study compares the performance of different word embedding models using intrinsic metrics (e.g., semantic similarity) and extrinsic measures (e.g., sentiment classification accuracy). The findings demonstrate that FastText excels in handling rare words and morphologically rich languages, while GloVe provides strong semantic representations. However, a custom embedding model incorporating BiLSTM and self-attention mechanisms outperforms others in domain-specific sentiment analysis tasks with limited labeled data. This research underscores the importance of selecting appropriate embedding techniques based on task requirements and resource constraints. It also proposes potential solutions for improving neutral sentiment classification, such as enhanced dataset labeling and domain-specific embeddings. The study's findings have significant implications for businesses seeking to leverage sentiment analysis for product development, marketing strategies, customer engagement, and competitive positioning. By bridging advanced NLP techniques with practical applications, this research contributes to democratizing access to market sentiment analysis tools and enabling more responsive, consumer-centric business strategies. Index Terms – Natural language processing (NLP), Market Sentiment Analysis (MSA), Word Embedding Models (WEM), Word2Vec, FastText, GloVe