Adaptive Local Context and Syntactic Feature Modeling for Aspect-Based Sentiment Analysis

Aspect-based sentiment analysis is a fine-grained task that consists of two types subtasks: aspect term extraction and classification. In the task, current methods suffer from lack information in difficulty identifying boundaries. classification classifier cannot adapt itself to text determine local context. To address these challenges, this work proposes an adaptive semantic relative distance approach based on dependent syntactic analysis, which uses appropriate context for each increase accuracy analysis. Meanwhile, study also predicts word labels by combining features extracted convolutional neural networks global precisely locate labels. subtasks, our proposed model improves F1 scores SemEval-2014 Task 4 Restaurant Laptop datasets compared state-to-the-art approaches, especially subtask.