List of questions suggested for consideration in each session

How might we find out such information? How important is psychological accuracy? What aspects and consequences of various formalisms lead to implausible cognitive models? 2. How would highly parallel hardware affect representations and processing? What evidence is there for the action of parallel agents (in Minsky's sense) in language understanding? How would such a model of cognition affect models of language comprehension? 3. How general are various formalisms? Are they really ad hoc solutions to relatively narrow domains (e.g stories, newspaper articles , data base question-answering, isolated sentences, etc.)? Which could be most successfully generalized? What problems are still unsolved by any formalism? Session 2. Language Representation and Reference It is the hypothesis of this session that entities which can be referred to provide prime evidence for the underlying representation necessary for extended passages of language (narrative, text or discourse). I. What functions do descriptions serve (e.g. inferential as well as referential)? How much inference is necessary to resolve reference? Are items which can be referenced "naturally" already appropriately organized or "indexed?" How close is representation to surface structure? Does representation depend on factors like attention or visualization on the part of a listener? 2. What things can be referred to anaphorically? What things can not? When (under what circumstances) can they be referenced? What intervening items can confuse reference or make it impossible? How are appropriate referring expressions constructed/understood? Are there differences in the answers for reference by pronoun and definite noun phrases? 3. Is the initial hypothesis above valid? What other methods can be used to find out about underlying passage representation? Session 3. Discourse: Speech Acts and Dialogue i. What sorts of models are necessary? (e.g. self-models, other-models, model of "contract," model of topic, etc.). 2. What should be included in a model? (e.g. beliefs, goals, current topic content, current topic constituents, etc.). 3. How domain dependent must the models be? What signals are used to cue model information? How are these signals understood? 4. How much information is lost in transcripts of dialogues (i.e. without intonation, body language, etc.)? Do we use different techniques in writing as substitutes? 5. What makes discourse coherent? How could we characterize and model what is conmmnicated in a coherent discourse? What mechanisms are used to relate utterances in a discourse? What relationships are there between production and comprehension, and how are the models of these processes used? 6. What extra …