SentiVec: Learning Sentiment-Context Vector via Kernel Optimization Function for Sentiment Analysis.

Area of Science:

• Natural Language Processing
• Machine Learning
• Artificial Intelligence

Background:

• Deep learning methods for sentiment analysis (SA) are increasingly popular.
• There is a need for more precise word embedding techniques to improve SA performance.

Purpose of the Study:

• To propose SentiVec, a kernel optimization function system for sentiment word embedding.
• To integrate statistical information and sentiment orientation into word vectors.
• To propagate and update semantic information across all word representations in a corpus.

Main Methods:

• SentiVec employs a two-phase approach: a supervised learning phase followed by two unsupervised updating models.
• The unsupervised models include the object-word-to-surrounding-words reward model (O2SR) and the context-to-object-word reward model (C2OR).

Main Results:

• SentiVec successfully extracts features related to semantic and sentiment information.
• The optimized sentiment vectors demonstrate superior performance compared to baseline methods.
• Outperformance was observed in word similarity, word analogy, and sentiment analysis tasks.

Conclusions:

• SentiVec provides an effective method for generating high-quality sentiment word embeddings.
• The proposed system enhances the precision of word representations for sentiment analysis applications.
• SentiVec offers a significant advancement over existing word embedding techniques in the field of sentiment analysis.