Multi-domain Multilingual Sentiment Analysis in Industry: Predicting Aspect-based Opinion Quadruples

TL;DR

This paper presents a multi-domain, multilingual aspect-based sentiment analysis system using large language models that efficiently extracts opinion quadruples across different industries and languages, maintaining high performance.

Contribution

It introduces a unified LLM-based approach capable of handling multiple domains and languages simultaneously for structured opinion extraction, reducing complexity.

Findings

Multi-domain model performs comparably to single-domain models

Unified model reduces operational complexity

Insights on handling non-extractive predictions and failure modes

Abstract

This paper explores the design of an aspect-based sentiment analysis system using large language models (LLMs) for real-world use. We focus on quadruple opinion extraction -- identifying aspect categories, sentiment polarity, targets, and opinion expressions from text data across different domains and languages. We investigate whether a single fine-tuned model can effectively handle multiple domain-specific taxonomies simultaneously. We demonstrate that a combined multi-domain model achieves performance comparable to specialized single-domain models while reducing operational complexity. We also share lessons learned for handling non-extractive predictions and evaluating various failure modes when developing LLM-based systems for structured prediction tasks.