Abstract
In this article, we investigate generation of referring expressions from large knowledge bases. We discuss some of the issues that arise when using existing referring expression generation algorithms and introduce a corpus-based algorithm that aims to overcome these issues. The new algorithm is based on the idea of communal common ground and uses a search engine to estimate what properties of a referent are likely to be known by hearers. The algorithm was evaluated against the Incremental Algorithm and human-created descriptions in a hearer oriented experiment where hearers attempted to identify described people and provide judgements about the used descriptions.
The algorithm outperformed the Incremental Algorithm in terms of the number of correctly identified referents.
The algorithm outperformed the Incremental Algorithm in terms of the number of correctly identified referents.
Original language | English |
---|---|
Publication status | Published - 31 Jul 2013 |
Event | PRE-CogSci - Berlin, Germany Duration: 31 Jul 2013 → 31 Jul 2013 |
Workshop
Workshop | PRE-CogSci |
---|---|
Country/Territory | Germany |
City | Berlin |
Period | 31/07/13 → 31/07/13 |
Keywords
- Referring Expression Generation
- Common Ground
- Heuristic
- Search Engine