Affiliation:
1. School of Electrical Engineering and Computer Science, Washington State University, Pullman, WA 99163, USA
Abstract
The Resource Description Framework (RDF) is the primary language to describe information on the Semantic Web. The deployment of semantic web search from Google and Microsoft, the Linked Open Data Community project along with the announcement of schema.org by Yahoo, Bing and Google have significantly fostered the generation of data available in RDF format. Yet the RDF is a computer representation of data and thus is hard for the non-expert user to understand. We propose a Natural Language Generation (NLG) engine to generate English text from a small RDF graph. The Natural Language Generation from Graphs (NLGG) system uses an ontology skeleton, which contains hierarchies of concepts, relationships and attributes, along with handcrafted template information as the knowledge base. We performed two experiments to evaluate NLGG. First, NLGG is tested with RDF graphs extracted from four ontologies in different domains. A Simple Verbalizer is used to compare the results. NLGG consistently outperforms the Simple Verbalizer in all the test cases. In the second experiment, we compare the effort spent to make NLGG and NaturalOWL work with the M-PIRO ontology. Results show that NLGG generates acceptable text with much smaller effort.
Publisher
World Scientific Pub Co Pte Lt
Subject
Artificial Intelligence,Computer Networks and Communications,Computer Science Applications,Linguistics and Language,Information Systems,Software
Reference40 articles.
1. J. Ellson, Graph Drawing Software (2003) pp. 127–148.
2. Building Natural Language Generation Systems
3. J. Coch, Natural Language Generation (1998) pp. 300–303.
4. Choosing words in computer-generated weather forecasts
5. S. Williams and E. Reiter, Corpus Linguistics (2003) pp. 899–908.
Cited by
3 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献