Word-Level Contextual Sentiment Analysis with Interpretability

• DOI: 10.1609/aaai.v34i04.5845
• Publication Date: 2020-06-29T21:26:55Z
• Authors: Tomoki Ito; Kota Tsubouchi; Hiroki Sakaji; Tatsuo Yamashita; Kiyoshi Izumi
• Keywords: Interpretability; Sentiment Analysis; Initialization

Word-level contextual sentiment analysis (WCSA) is an important task for mining reviews or opinions. When analyzing this type of in the industry, both interpretability and practicality are often required. However, such a WCSA method has not been established. This study aims to develop with practicality. To achieve aim, we propose novel neural network architecture called Sentiment Interpretable Neural Network (SINN). realize SINN practically, learning strategy Lexical Initialization Learning (LEXIL). interpretable because it can extract word-level through extracting original its local global contexts. Moreover, LEXIL without any specific knowledge context; therefore, practical. Using real textual datasets, experimentally demonstrate that proposed effective improving features high ability interpretability.