ANAPHORA RESOLUTION for ENGLISH (by Richard Evans)

The anaphor resolver is dependent on output from preprocessing tools performing syntactic analysis (FDG Parser), and classification of instances of it.

When you have typed your sentences, containing at least one 3rd person pronoun (not reflexives or recipricals), press enter and click `Send the text.' The program should resolve each pronoun to a NP antecedent. MARS is not equipped to deal with hyphenated text.

Mark paragraph boundaries by leaving a blank line between paragraphs.

The resolver is intended to resolve 3rd person pronouns ONLY, singular and plural. Currently, there is no provision for gender agreement, except for the case of common English masculine and feminine proper names. (It also doesn't like some extended/non-ascii symbols)

If you don't get an answer from the server, either be stubborn and try again (click `reload' at `http://www.clg.wlv.ac.uk/MARS/'), or e-mail me.

The anaphor resolver outputs the names and positions of each 3rd person pronoun and it's indicated antecedent. There is also a table printed with all possible candidates and the scores that they obtained for each of 14 indicators. The first 12 indicators are presented in Mitkov, R. (1998) "Robust pronoun resolution with limited knowledge". The last two are my own additions, reflecting syntactic parallelism and a boost for pronoun candidates. The scoring scheme implemented here differs from the original, let me know if you want to know how and why. Please e-mail Richard Evans with your comments.

Home | About | People | Research | News | Papers | Seminars | Software | Links | Contact |