IGT system improves multilingual financial question answering accuracy
IGT @ FinMMEval 2026 Task 2: Question-Type Prompting with Targeted Extraction for Multilingual Financial QA
Computation and LanguageArtificial Intelligence
Summary
Answering financial questions from documents in different languages is hard because some questions need exact numbers while others need more complex explanations. The authors noticed two main types of questions and designed different ways to answer each. For number-based questions, their system picks out key words from financial filings. For explanation questions, it uses rules to find relevant news passages. Their method worked better than a general-purpose baseline and ranked third among competitors.
What this means in practice
- •For financial data analysts: Extract precise numeric metrics from multilingual financial filings using keyword-based methods tuned for structured data.
- •For multilingual news monitoring teams: Identify relevant news passages to answer complex financial questions spanning several languages with rule-based filtering.
Authors
Yuwen Chiu
Abstract
We present the IGT system for PolyFiQA Task 2 of the FinMMEval Lab at CLEF 2026, a multilingual financial question answering task over English SEC filings and multilingual news articles (English, Chinese, Japanese, Spanish, Greek) for four companies. Our central observation is that the 344 development questions divide into two families requiring fundamentally different approaches: structured numeric types (R&D ratio, cash flow, capital expenditure) are best answered by direct keyword extraction on filing text, while synthesis types (investment strategy, capital allocation, top-three revenue focuses) require rule-based multilingual news passage selection. A dataset analysis reveals that 17-18 of 19 ground-truth reference answers per synthesis type share an exact evidence label prefix, whose unigram tokens contribute directly to ROUGE-1 overlap. The final system achieves development ROUGE-1 approximately 0.395, a 60% relative improvement over a generic RAG baseline (approximately 0.247), and ranks 3rd of 12 teams on the official test set with ROUGE-1 = 0.3071, Precision = 0.2821, and Recall = 0.4044.