Anaphoric reference resolution

Ivana Kruijff-Korbayova

Hauptseminar. Computerlinguistik, 2. Studienabschnitt
Geb. 17.2, Konferenzraum 2.11, Freitag 11-13

Questions for session 20.12.2002: Centering Theory I.

• It was stated therein, that there are several possibilities how to handle informational incompleteness or ambiguity. It was among all proposed computing all possibilities and applying constraints after that, or choosing the most likely possibility and performing backtracking when needed. Given the second way of handling the problem - how would one detect the performed choice in the past was wrong and one should backtrack? (Vasek)
• It is pressuposed that the discourse segment consists of a sequence of utterances. Does considering complexer structures with embedded sentences cause any particular problems or does it make no difference at all (meaning, these cases can be dealt with using this approach aswell)? Moreover, one would expect role "binding" preferences in some kind of subordinate sentences (depending on the actual connective) - is this in compliance with the theory or is it a different point of view and one can't tell? (Vasek)
• The relation between the ranking and the surface word order in languages with free word order is presented as one of the open questions and inspiration for researchers at the moment. Are there any results for comparable phenomena (eg. topicalization) in English? (Vasek)
• It is said that you can have only one backward-looking center....But is this really a solution to avoid ambiguity? (Iris)
• What is the difference between Center Continuation and Center Retaining? (Iris)
• Why ist (13) more coherent than (14)? And if it is more coherent refering always to the same thing, what is the advantage of beening more coherent? One can understand (14) perfectly....(Iris)
• Hasn't "key man" (29) a value which is positive connected? It is said it can be interpreted either VF or VL.... (Iris)