Sentiment Analysis Method Based on Improved Feature Vector

Sentiment analysis is an important task in the field of natural language processing. The purpose of this article is to combine manually organized sentiment dictionaries with deep learning techniques to improve the accuracy of sentiment analysis. This article proposes a method based on improved feature vector for text sentiment analysis, which combines the semantic vector outputted by the deep learning model with the sentiment vector outputted by the emotional dictionary algorithm. Additionally, this article provides two combination schemes. The first solution is to combine in the sentence dimension. It involves calculating the sentiment score of the text through the emotional dictionary algorithm, converting it into a One-Hot vector, multiplying it with an Embedding matrix, and then converting it into a sentiment vector. The output of the deep learning model was concatenated with the semantic vector and then input into the decision layer. The second solution is to combine in the word dimension, filter out emotional words in the text, encode and preprocess emotional words, and assign a marker vector based on the type of emotional words. After combining the two vectors, input them into a deep learning model to generate an sentiment vector, and then concatenate them with the output vector of deep learning for decision-making. After validation through dataset experiments, the two models proposed in this article have demonstrated varying degrees of improvement across four key performance metrics: accuracy, precision, recall, and F1-score.