A Study on Sentiment Analysis of Tourism Texts Based on Word2vec Optimization Weights

Sentiment analysis of tourism reviews on online tourism platforms can effectively track the emotional changes of tourists. It also provides valuable improvement suggestions for the management team of scenic spots, which has a wide range of practical application potential. Aiming at the text of tourism reviews, which is characterized by rich vocabulary, high emotional content, and insufficient, and insufficient specification, a textual representation of word vectors is proposed to optimize the weights and integrate the emotional information. First, the Word2vec model is used to vectorize the unstructured text data, and the word vectors of all texts are trained. Then, considering the text lexicality and the effect of modifiers on sentiment words, the sentiment information is combined with the TF-IDF algorithm. Finally, a feature-weighted word vector model is established, and a machine learning method is used to classify the text and analyze the sentiment polarity of the attraction reviews by taking Yibin Shunan Bamboo Sea as an example. Multiple sets of comparison experiments were conducted, and the experimental results show that, compared to the traditional use of methods based on distributed word vector representations, the best classification performance is achieved by using this model with SVM, with an accuracy rate of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 0. 6 1 \%}$</tex> and a precision rate of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 1. 1 1 \%}$</tex>.