Identifying complaints based on semi-supervised mincuts

In today's competitive business world, customer service is often at the heart of businesses that can help strengthen their brands. Customer complaint resolution in a timely and efficient manner is key to improving customer satisfaction. Moreover, customers’ complaints play important roles in identifying their requirements which offer a starting point for effective and efficient planning of the company's overall R&D and new product or service development activities. That said, businesses face challenges towards automatically identifying complaints buried deep in massive online content. In this paper, we propose a graph-based semi-supervised learning paradigm leveraging syntactic and semantic representations of tweets. Intrinsic evaluation results on a benchmark dataset illustrate that the proposed approach outperforms state-of-the-art supervised non-graph based classification models for solving the complaints identification task, and confirms the efficacy of the proposed approach. Experimental results also show that the performance of the state-of-the-art supervised complaint classification model trained over hand-crafted features extracted from several linguistic resources can be reached with less than 50% of the training data with the proposed approach.