Design and Implementation of a Toolkit for the Aspect-Based Sentiment Analysis of Tweets

The microblogging service Twitter is a lively social media channel where people report on almost everything imaginable. Using tweets to analyze consumers' opinions about individual products and services is a valuable source of information used by brands and companies to learn about their public perception. We present an aspect-based sentiment analysis approach for extracting positive, negative, and neutral aspects from tweets by leveraging part-of-speech tagging and dependency parsing from natural language processing. The approach manifests in the design of a software toolkit that facilitates the sentiment analysis, starting from the extraction of tweets, filtering, over the automated analysis, up to the interactive display of the results. By applying dependency patterns between nouns, adjectives, adverbs, and negotiations to the individual sentences, the toolkit identifies aspects and sentiment expressions, for which the polarities are determined subsequently. We demonstrate the developed solution by looking at the case of Airbnb, where people can rent out their homes to others over a Web platform. It is important to note that Airbnb only provides the platform and has little influence on the actual lodging service. Thus, looking at word-of-mouth regarding individual service experiences yields evidence of praise and criticism, which can help to seize improvement potential.