Improving Chess Commentaries by Combining Language Models with Symbolic Reasoning Engines
This addresses the challenge of grounding language models in complex reasoning tasks for applications like chess analysis, though it is incremental in bridging symbolic and neural approaches.
The paper tackled the problem of generating chess commentaries by combining symbolic reasoning engines with controllable language models, resulting in commentaries preferred by human judges over previous baselines.
Despite many recent advancements in language modeling, state-of-the-art language models lack grounding in the real world and struggle with tasks involving complex reasoning. Meanwhile, advances in the symbolic reasoning capabilities of AI have led to systems that outperform humans in games like chess and Go (Silver et al., 2018). Chess commentary provides an interesting domain for bridging these two fields of research, as it requires reasoning over a complex board state and providing analyses in natural language. In this work we demonstrate how to combine symbolic reasoning engines with controllable language models to generate chess commentaries. We conduct experiments to demonstrate that our approach generates commentaries that are preferred by human judges over previous baselines.