Research on topic mining of online reviews of psychological self-help books based on LDA and word2vec model

In order to better grasp the needs of readers, provide valuable data support for the publication of psychological self-help books, and provide decision-making references for the improvement of mental health services. Using the review texts of the top 500 books on the spiritual healing book list of China's Dangdang website in 2023 as the data source, after preprocessing the review texts, this paper constructed an LDA model to mine the review data for topics and extract their topic-feature word distributions. Secondly, to compensate for the LDA model's limitation in capturing contextual semantics, a word2vec model is constructed based on the review text to train word vectors, and the extracted feature words are expanded through vector similarity calculation. Finally, five main topics were extracted, including product function, physical attributes of books, shopping service experience, promotional activities, and content value. This paper found that psychological self-help books have a prominent role in parent-child education and emotional management, readers attach great importance to the quality of the physical attributes of books, and the practicality and classic nature of content promote the word-of-mouth spread of books.