250 likes | 385 Views
An information state approach to natural interactive dialogue. Staffan Larsson, Robin Cooper Department of linguistics G ö teborg University, Sweden. The information state approach – key concepts.
E N D
An information state approach to natural interactive dialogue Staffan Larsson, Robin Cooper Department of linguistics Göteborg University, Sweden
The information state approach – key concepts • Information states represent information available to dialogue participants, at any given stage of the dialogue • Dialogue moves trigger information state updates, formalised as information state update rules • TRINDIKIT: software package for implementing dialogue systems; based on the information state approach to dialogue management
GoDiS – a dialogue system • Implemented using the TRINDIKIT • Adapted for information-seeking dialogue, menu-based dialogue, and instructional dialogue
Information state in GoDiS • Based on Ginzburgs notion of QUD (Questions Under Discussion): a partially ordered set of questions which have been raised and are under discussion • Includes dialogue plan
Sample GoDiS information state findout(?x.month(x)) findout(?x.class(x)) respond(?x.price(x)) PLAN = PRIVATE = AGENDA = { findout(?return) } dest(paris) transport(plane) task(get_price_info) COM = SHARED = QUD = < x.origin(x) > LM = { ask(sys, x.origin(x)) }
Sample update rule • downdateQUD • Before an answer can be integrated by the system, it must be matched to a question on QUD in(SHARED.LM, answer(usr, A)) fst(SHARED.QUD, Q) relevant_answer(Q, A) pre: pop(SHARED.QUD) reduce(Q, A, P) add(SHARED.COM, P) eff:
Information-seeking dialogue • User needs to give information which enables the system to perform its task (booking a ticket, providing price information etc.) • Typical dialogue system behaviour: user must give information in the order determined by the system questions
Typical human-computer dialog S: Where do you want to go? U: Paris S: How do you want to travel? U: A flight please S: When do you want to travel U: April S: what class did you have in mind? … S: The price is $123
Dialogue plans for information-seeking dialogue • Ask how user wants to travel • Ask where user wants to go to • Ask where user wants to travel from • Ask when user wants to travel • … • Lookup database • Tell user the price
Typical human-human dialogue S(alesman), C(ustomer) S: hi C: flights to paris S: when do you want to travel? C: april, as cheap as possible ...
Accommodation • Lewis (1979): If someone says something at t which requires X to be in the conversa- tional scoreboard, and X is not in the scoreboard at t, then (under certain conditions) X will become part of the scoreboard at t • Has been applied to referents and propositions
Question accommodation • If questions are part of the information state, they too can be accommodated • If the latest move was an answer, and there is an action in the plan to ask a matching question, put that question on QUD
Update rule for question accommodation accommodateQuestion in(SHARED.LM, answer(usr, A)) in(PRIVATE.PLAN, findout(Q)) relevant_answer(Q, A) pre: delete(PRIVATE.PLAN, findout(Q)) push(SHARED.QUD, Q) eff:
Task accommdation • In some cases, the system may not even know what task the user wants the system to perform • If latest move was an answer, and there is currently no plan, find a task and corresponding plan containing a matching question; accommodate the task and load the appropriate plan • If there are several matching plans, ask clarification question
Question and task accommodation in information-seeking dialogue S: hi U: flights to paris • system finds plan containing appropriate questions, and loads it into the plan field in the information state • system accommodates questions: how does user want to travel + where does user want to go, and integrates the answers “flight” and “to paris” • system proceeds to next question on plan S: when do you want to travel?
Menus vs. dialogue • Menu-driven interaction is ubiquitous: automated cinema ticket booking, mobile phones, computers, video recorders… • Often tedious and frustrating; hard to find what you want; inflexible • Can be straightforwardly implemented as dialogue systems, but you still have to descend the menu structure one node at a time
Typical menu-based dialogue S: What do you want to do? U: Search the phonebook S: What name do you want to search for? U: John S: John’s number is 0312345566. Do you want to call John? U: Yes S: Calling John.
Plans derived from menu structure • Toplevel: ask what user wants to do (phonebook, messages etc); load corresponding plan • Phonebook: ask what user wants to do (search phonebook, add to phonebook etc); load corresponding plan • Search phonebook: ask for name; if name exists, inform of number; ask if user wants to call number; if yes, call number
Question accommodation in menu-based dialogue U: John • system finds several plans containing a request for a name, and asks the user which one is correct S: Do you want to search the phonebook for John? U: Yes, and call him up • system accommodates answer to the question whether user wants to call S: John’s number is 0312345566. Calling John.
Strategies for asking clarification questions • Ask a series of yes/no-questions, one for each alternative; OK if user can interrupt • Ask wh-question; if user does not provide answer, list alternatives • Ask alternative question
From manuals to instructional dialogue • Domain plan is extracted from manual • Domain plan is converted into dialogue plan, including dialogue moves • Surface realisation of moves based on manual • Manual can be reconstructed from domain plan, if system is run in monologue mode
From manual to dialogue plan • Plan conversion table
Advantages of dialogue mode for manuals • User does not have to look in manual, or keep track of the current point • System avoids irrelevant information when the action to be taken depends on a condition • User controls the level of detail; can skip parts already known or ask for more specific instructions if necessary • Grounding and question accommodation
Conclusion • Question and task accommodation support natural interactive dialogue, where user controls in which order information is presented • Information state approach enables easy implementation of question and task accommodation • (in paper: more on instructional dialogue and its relation to manuals) • Implemented in GoDiS using the TRINDIKIT software package (www.ling.gu.se/research/projects/trindi/trindikit.html)