CLARA: Clarification of Language Ambiguity through Result Analysis for Natural-Language Cancer Genomics Queries

2026-08-03

AI Summary

This work addresses the challenge of semantic ambiguity in natural language queries over cancer genomics databases, which often leads to scientifically ambiguous interpretations. The authors propose a novel method that translates user questions into typed scientific query specifications, generating and executing multiple plausible interpretations in parallel. When execution results from dual engines (SQLite and pandas) diverge beyond a predefined threshold, the system proactively prompts the user for clarification. This approach uniquely leverages downstream query result discrepancies to identify critical ambiguities, achieving a quantifiable trade-off between safety and user interaction burden. Evaluated on 330 real-world queries, the method achieves 100% recall of all 115 result-sensitive ambiguities and an overall clarification accuracy of 89.2%, demonstrating its effectiveness and reliability.

Abstract

A natural language interface can be used to make cancer genomics databases easier to use, but even if a question is perfectly fluent, its scientific meaning can be ambiguous. We propose CLARA, a framework that represents a question as a typed scientific query specification, considers a few possible interpretations, executes them, and asks for clarification when the estimates diverge. CLARA was assessed on mutation-prevalence contrasts among eight TCGA PanCancer Atlas cohorts and a 30-gene panel. This benchmark consisted of 330 unique executable contrasts varying in mutation scope, assay denominator, and sample context; 115 contrasts were result-sensitive and 215 were result-stable, per the preregistered definition of relative divergence greater than 0.10 or absolute divergence greater than 5 percentage points. An independently implemented pandas execution engine perfectly replicated all 660 results from the SQLite engine. In a separate 120-question LLM-generated, manually vetted language stress test, CLARA recognized all 60 result-sensitive contrasts and needlessly clarified 13 of 60 stable contrasts (accuracy 89.2%, sensitivity/recall 100%, specificity 78.3%). Standalone machine learning had superior overall accuracy (97.5%) but missed one critical contrast. This demonstrates that downstream execution can distinguish consequential from inconsequential ambiguity and reveal an explicit trade-off between safety and burden.

Problem

• natural language ambiguity
• cancer genomics
• query interpretation
• semantic uncertainty
• result divergence

Innovation

• natural language interface
• ambiguity resolution
• query execution feedback
• cancer genomics
• result-sensitive clarification