Scholarlink

<p>Debugging an ontology stated in description logic is a cognitively straining task for humans and there has been a recent increase in literature discussing various methods of facilitating this task. While some of these methods involve selected fragments of cognitive theory, none systematically integrate a cognitive theory. This paper aims to fill that gap by using the cognitive architecture ACT-R to model the ABox consistency task; a task which is essential for debugging ontologies. The resulting model is called SHARP (Simulating Human ABox Reasoning Performance) which simulates the ABox consistency algorithm as if it were run by a human mind. We evaluate the model by comparing its predicted inference times with empirical data gathered from an experiment with 70 participants. The complexity measures based on SHARP prove very accurate in predicting the complexity ordering of the set of ABoxes used in the experiment and outperform other measures found in the literature. These measures can thus be considered cognitively adequate, which opens up opportunities for them being utilised in cognitively optimised debugging of ontologies. Besides the complexity ordering, several other hypotheses are tested, which demonstrate that SHARP displays certain inaccurate consequences on a finer scale. These are considered points of improvement.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Estimating the cognitive complexity of description logic entailments with a cognitive architecture

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月14日

查看全文
  • Jelle Tjeerd Fokkens,
  • Fredrik Engström

摘要

Debugging an ontology stated in description logic is a cognitively straining task for humans and there has been a recent increase in literature discussing various methods of facilitating this task. While some of these methods involve selected fragments of cognitive theory, none systematically integrate a cognitive theory. This paper aims to fill that gap by using the cognitive architecture ACT-R to model the ABox consistency task; a task which is essential for debugging ontologies. The resulting model is called SHARP (Simulating Human ABox Reasoning Performance) which simulates the ABox consistency algorithm as if it were run by a human mind. We evaluate the model by comparing its predicted inference times with empirical data gathered from an experiment with 70 participants. The complexity measures based on SHARP prove very accurate in predicting the complexity ordering of the set of ABoxes used in the experiment and outperform other measures found in the literature. These measures can thus be considered cognitively adequate, which opens up opportunities for them being utilised in cognitively optimised debugging of ontologies. Besides the complexity ordering, several other hypotheses are tested, which demonstrate that SHARP displays certain inaccurate consequences on a finer scale. These are considered points of improvement.