A Brief Introduction on Granular Reasoning in LarKC
In LarKC, we think that if knowledge represented in RDF triples can be represented in multiple levels to meet users' needs in different grain sizes (granularities) or can be represented in multiple views to meet users' needs in different perspectives, it might be a step towards scalable knowledge processing (search, reasoning, etc.).
Granular Reasoning is a notion for multi-level and multi-view reasoning. From the multi-level reasoning point of view, it is very relevant to variable precision logic. Variable precision can be divided into two types, namely, variable specificity and variable certainty [Michalski 1985]. The reason why it is very relevant to LarKC is that: concerning time constraint, given more time, variable specificity system can provide a more specific answer, while variable certainty system can provide a more certain answer [Michalski 1985]. Based on this idea, LarKC will provide knowledge query answers with multi-level of specificity and multi-level of certainty. From the multi-view reasoning point of view, since the same knowledge source can be represented in different views [Minsky 2006], LarKC investigates on different viewpoints of the RDF triple organizations.
For LarKC, the study is from two aspects. One aspect is the study of granular reasoning from human perspective, an in progress work is that we explore the granular structure (tree style, network style, etc.) in human information processing system by Cognitive Psychology and Cognitive Neuroscience experiments (such as experiments on human heuristic search strategies). The other aspect is to apply the human problem solving strategies observed in our experiments to reasoning systems and develop granular reasoner for LarKC reasoning plugins. Because ACT-R is a cognitive architecture that implements on computers and it can be used to imitate human problem solving strategies, we are currently developing our granular reasoner based on ACT-R.
Further Readings
Detailed discussion on the idea of granular reasoning can be found in following links:
[1] Unifying Web-scale Search and Reasoning from the Viewpoint of Granularity. Yi Zeng, Yan Wang, Zhisheng Huang, Ning Zhong. In: Proceedings of the 2009 International Conference on Active Media Technology, Lecture Notes in Computer Science 5820, 418-429, Springer, Beijing, China, October 22-24, 2009.
[2] Towards Granular Reasoning on the Web, Ning Zhong, Yiyu Yao, Yulin Qin, Shengfu Lu, Jia Hu, and Haiyan Zhou. Proceedings of the 2008 Workshop on New forms of Reasoning for the Semantic Web: scalable, tolerant and dynamic (NEFORS 2008), the 3rd Asian Semantic Web Conference (ASWC2008).
[3] Cognitive Memory Retention Based Starting Point for Query Extension and Granular Selection, Yi Zeng, Haiyan Zhou, Ning Zhong, Yulin Qin, Shengfu Lu, Yiyu Yao, Yang Gao. In: Cognitive Memory Component (v1), LarKC deliverable 2-3-1, Coordinated by Jose Quesada and Yi Zeng, March 30, 2009.
[4] Information Granulation and Switching --Through Basic-level to Accelerate Selection,Ning Zhong, Yulin Qin, Shengfu Lu, Jia Hu, Haiyan Zhou, Erzhong Zhou, LarKC General Assembly, 12th-14th Jan 2009 in Bled, Slovenia (restricted to LarKC members now).
[5] Accelerating LarKC by Reasoning through Granular Knowledge Structures, Yi Zeng, Ning Zhong, Shengfu Lu, Yulin Qin, Haiyan Zhou, LarKC General Assembly, 12th-14th Jan 2009 in Bled, Slovenia (restricted to LarKC members now).
Investigations of knowledge representation from the view point of granularity can be found in following links:
[1] Supporting Literature Exploration with Granular Knowledge Structures, Yiyu Yao, Yi Zeng, and Ning Zhong. In: Proceedings of the 11th International Conference on Rough Sets, Fuzzy Sets, Data Mining, and Granular Computing, Lecture Notes in Artificial Intelligence 4482, Springer, Toronto, Canada, May 14-16, 2007, 182-189.
[2] On Granular Knowledge Structures, Yi Zeng, Ning Zhong. In: Progress of Advanced Intelligence: Proceedings of 2008 International Conference on Advanced Intelligence, Beijing, China, October 18-22, 2008, 28-33.