Answer Candidate Type Selection: Text-to-Text Language Model for Closed Book Question Answering Meets Knowledge Graphs
This work addresses a specific bottleneck in knowledge graph question answering for users needing accurate answers about less common entities, but it is incremental as it builds on existing pre-trained models.
The paper tackles the problem of limited capacity and reduced quality in pre-trained text-to-text language models for knowledge graph question answering, particularly for questions with less popular entities, by introducing a method that filters and re-ranks answer candidates based on entity types from Wikidata, resulting in improved performance.
Pre-trained Text-to-Text Language Models (LMs), such as T5 or BART yield promising results in the Knowledge Graph Question Answering (KGQA) task. However, the capacity of the models is limited and the quality decreases for questions with less popular entities. In this paper, we present a novel approach which works on top of the pre-trained Text-to-Text QA system to address this issue. Our simple yet effective method performs filtering and re-ranking of generated candidates based on their types derived from Wikidata "instance_of" property.