| EPSRC Reference: |
GR/S76434/01 |
| Title: |
Anaphora Resolution and Underspecification (ARRAU) |
| Principal Investigator: |
Poesio, Professor M |
| Other Investigators: |
|
| Researcher Co-Investigators: |
|
| Project Partners: |
|
| Department: |
Computer Sci and Electronic Engineering |
| Organisation: |
University of Essex |
| Scheme: |
Standard Research (Pre-FEC) |
| Starts: |
01 September 2004 |
Ends: |
30 November 2007 |
Value (£): |
233,241
|
| EPSRC Research Topic Classifications: |
| Human Communication in ICT |
|
|
| EPSRC Industrial Sector Classifications: |
|
| Related Grants: |
|
| Panel History: |
|
|
Summary on Grant Application Form |
|
The biggest current challenge for Natural Language Engineering (NLE) is to develop robust techniques for semantic interpretation. One of the areas of semantic interpretation in which most progress has taken place is anaphora resolution. The performance of the current algorithms for interpreting anaphoric expressions such as pronouns or definite descriptions is continuously increasing, and pronoun resolution algorithms can already resolve up to 80% of antecedents correctly, at least in certain genres (Lappin and Leass, 1994; Mitkov, 1998, 2002; Tetreault, 2001). As a result, these algorithms are beginning to be used in applications such as information extraction (Gaizauskas & Wilks, 1998), question answering (Morton 2000) and summarization (Boguraev and Kennedy, 1997); but so far the results have been mediocre, as the performance of the anaphora resolution algorithms is still not sufficient. The 20-30% of 'difficult' anaphoric references current algorithms cannot handle are mostly references to complex objects such as plurals, events, and abstract objects, which often are only mentioned implicitly. Part of the problem is that in these cases it is not always clear what the antecedent of the anaphoric expression should be, and whether identifying this antecedent is essential. The primary goal of this project is to carry out an empirical study of these 'difficult' cases, identifying the anaphoric expressions that have to be resolved, and indicating what their antecedent should be. These concerns are also related to an issue of more general theoretical interest, that of underspecification in semantics: the extent to which aspects of the interpretation of an utterance have to be recovered, and when. We expect the project to contribute to the understanding of this more general issue, as well. The second goal of the project is to produce an annotated corpus that can be used to develop and evaluate algorithms in this area. The third goal is to use the output of this empirical analysis to modify an existing anaphora resolution system (Vieira and Poesio, 2000) which has already been used for applications such as question-answering (Morton, 2000).
|
|
Key Findings |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Potential use in non-academic contexts |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
|
Impacts |
|
| Description |
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk |
| Summary |
|
| Date Materialised |
|
|
|
|
Sectors submitted by the Researcher |
|
This information can now be found on Gateway to Research (GtR) http://gtr.rcuk.ac.uk
|
| Project URL: |
|
| Further Information: |
|
| Organisation Website: |
http://www.sx.ac.uk |