Markov Chains for Robust Graph-Based Commonsense Information Extraction

TitleMarkov Chains for Robust Graph-Based Commonsense Information Extraction
Publication TypeMiscellaneous
Year of Publication2012
AuthorsTandon, N., Rajagopal D., & de Melo G.
Other Numbers3390
Abstract

Commonsense knowledge is useful for making Web search, local search, and mobileassistance behave in a way that the user perceives as “smart”. Most machine-readableknowledge bases, however, lack basic commonsense facts about the world, e.g. the propertyof ice cream being cold. This paper proposes a graph-based Markov chain approach to extractcommon-sense knowledge from Web-scale language models or other sources. Unlike previouswork on information extraction where the graph representation of factual knowledge is rathersparse, our Markov chain approach is geared towards the challenging nature of commonsenseknowledge when determining the accuracy of candidate facts. The experiments show thatour method results in more accurate and robust extractions. Based on our method, wedevelop an online system that provides commonsense property lookup for an object in realtime.

Acknowledgment

This work was partially funded by the Deutscher Akademischer Austausch Dienst (DAAD) through a postdoctoral fellowship.

URLhttp://www.icsi.berkeley.edu/pubs/ai/commonsenseinfoextraction12.pdf
Bibliographic Notes

Demo at the 24th International Conference on Computational Linguistics (COLING), Mumbai, India

Abbreviated Authors

N. Tandon, D. Rajagopal, and G. de Melo

ICSI Research Group

AI

ICSI Publication Type

None