This paper is published by ACM and requires payment for access.

Sentiment Analysis with Incremental Human-in-the-Loop Learning and Lexical Resource Customization


The adjustment of probabilistic models for sentiment analysis to changes in language use and the perception of products can be realized via incremental learning techniques. We provide a free, open and GUI-based sentiment analysis tool that allows for a) relabeling predictions and/or adding labeled instances to retrain the weights of a given model, and b) customizing lexical resources to account for false positives and false negatives in sentiment lexicons. Our results show that incrementally updating a model with information from new and labeled instances can substantially increase accuracy. The provided solution can be particularly helpful for gradually refining or enhancing models in an easily accessible fashion while avoiding a) the costs for training a new model from scratch and b) the deterioration of prediction accuracy over time.

DOI: 10.1145/2700171.2791022

Extracted Key Phrases

3 Figures and Tables

Sorry, there's nothing here.