Universal, Unsupervised (Rule-Based), Uncovered Sentiment Analysis

TL;DR

This paper introduces a universal, unsupervised, rule-based approach for multilingual sentiment analysis that leverages syntactic structures to improve robustness and transferability across languages and domains.

Contribution

It proposes a novel compositional, syntax-based rule system for unsupervised sentiment analysis, enhancing cross-lingual and cross-domain performance.

Findings

Outperforms existing unsupervised methods

Surpasses state-of-the-art supervised models outside their training corpus

Shared compositional operations across multiple languages

Abstract

We present a novel unsupervised approach for multilingual sentiment analysis driven by compositional syntax-based rules. On the one hand, we exploit some of the main advantages of unsupervised algorithms: (1) the interpretability of their output, in contrast with most supervised models, which behave as a black box and (2) their robustness across different corpora and domains. On the other hand, by introducing the concept of compositional operations and exploiting syntactic information in the form of universal dependencies, we tackle one of their main drawbacks: their rigidity on data that are structured differently depending on the language concerned. Experiments show an improvement both over existing unsupervised methods, and over state-of-the-art supervised models when evaluating outside their corpus of origin. Experiments also show how the same compositional operations can be shared across languages. The system is available at http://www.grupolys.org/software/UUUSA/