ExplainabilityTransparencyAccountabilitycs.AI
An xAI Approach for Data-to-Text Processing with ASP
Alessandro Dal Palù, Agostino Dovier, Andrea Formisano
August 30, 2023
International Conference on Logic Programming
4 citations
Key Takeaways
-
Explainability through logic-based rules: Unlike black-box machine learning approaches, this framework uses explicit logic rules (Answer Set Programming) to generate text from data, making it possible to understand and verify exactly why the system produces specific descriptions—a key requirement for responsible AI systems.
-
Guaranteed accuracy and control: The approach provides "proven optimal solutions" that explicitly control for accuracy errors and the level of detail in summaries, addressing a critical challenge in data-to-text systems where generated descriptions may contradict or misrepresent the underlying data.
-
Transparency in decision-making: By organizing text hierarchically through interpretable logic rules rather than learned patterns, stakeholders can audit how the system decides "what to say" (which data elements matter) and "how to say it" (phrasing and synthesis level), enabling better oversight of AI-generated content.
-
Reduced risk of contradictions and hallucinations: The rule-based structure systematically prevents the generation of contradictory or redundant statements that can occur in neural approaches, improving reliability for applications where accuracy between data and generated text is critical (e.g., financial reports, medical summaries).
Abstract
The generation of natural language text from data series gained renewed interest among AI research goals. Not surprisingly, the few proposals in the state of the art are based on training some system, in order to produce a text that describes and that is coherent to the data provided as input. Main challenges of such approaches are the proper identification of "what" to say (the key descriptive elements to be addressed in the data) and "how" to say: the correspondence and accuracy between data and text, the presence of contradictions/redundancy in the text, the control of the amount of synthesis. This paper presents a framework that is compliant with xAI requirements. In particular we model ASP/Python programs that enable an explicit control of accuracy errors and amount of synthesis, with proven optimal solutions. The text description is hierarchically organized, in a top-down structure where text is enriched with further details, according to logic rules. The generation of natural language descriptions' structure is also managed by logic rules.