22
Towards Conversational Human Computer Interaction Dept. of Computer Science University of Rochester Rochester, NY 14627 By: James F. Allen, Donna K. Byron, Myroslava Dzikovska George Ferguson, Lucian Galescu, Amanda Stent Present By: Nourhan Younis MOIS 466- Human Computer Interaction Mini-Project 1

Towards Conversational Human Computer Interaction

  • Upload
    teleri

  • View
    30

  • Download
    0

Embed Size (px)

DESCRIPTION

MOIS 466- Human Computer Interaction Mini-Project 1 . Dept. of Computer Science University of Rochester Rochester, NY 14627 By: James F. Allen, Donna K. Byron, Myroslava Dzikovska George Ferguson, Lucian Galescu, Amanda Stent Present By: Nourhan Younis . - PowerPoint PPT Presentation

Citation preview

Page 1: Towards Conversational Human Computer Interaction

Towards Conversational Human Computer Interaction

Dept. of Computer Science University of Rochester Rochester, NY 14627

By: James F. Allen, Donna K. Byron, Myroslava Dzikovska George Ferguson, Lucian Galescu, Amanda Stent

Present By: Nourhan Younis

MOIS 466- Human Computer Interaction Mini-Project 1

Page 2: Towards Conversational Human Computer Interaction

What is a dialogue system?

A computer intended to converse with a human, with a coherent structure.

Page 3: Towards Conversational Human Computer Interaction

Dialogue System

Limit communication Increasing efficiency

of service Making speech more

predictable Easier recognition

and language processing

Page 4: Towards Conversational Human Computer Interaction

Human Conversation

Basing HCI on human conversation

More interaction and complexity

BUT People claim it won’t

be as effective as graphic user interfaces (GUIs)

Page 5: Towards Conversational Human Computer Interaction

Conversational Interfaces Vs. GUIs

Limitations of Graphic User Interfaces:

The size of the device used Some tasks requires the use

of hands and eyes More flexible Reduces the training time The provide a set of

operations with many complex, sequential tasks

Adds high level commands with many options

Requires lots of training

Page 6: Towards Conversational Human Computer Interaction

Considerations

It can’t totally replace GUIs because: The user will have to remember a long

set of verbal commands to perform a task (ex. Menus)

The Conversational inter- faces, however, would provide the opportunity for the user to state what he/she wants to do in his/her own terms, just as he/she would do to another person, and the system takes care of the complexity.

Page 7: Towards Conversational Human Computer Interaction

Conversation Interfaces Conversational Interaction will not depend on a

series of commandsBut Interaction involves:

Defining and discussing tasks Exploring ways to perform the task Collaborating to get it done

All interactions are contextually interpreted with respect to the interactions performed to this point, allowing the system to anticipate the user’s needs and provide responses that best further the user’s goals

Page 8: Towards Conversational Human Computer Interaction

Dialogue Task Complexity

Page 9: Towards Conversational Human Computer Interaction

Hypotheses H1: Practical Dialogues

Conversational competence required for practical dialogues, although still complex, is significantly simpler to achieve than general human conversational competence.

H2: The Domain IndependenceWithin the genre of practical dialogue, the bulk of the complexity in the language interpretation and dialogue management are independent of the task being performed

Page 10: Towards Conversational Human Computer Interaction

A Fragment of Practical Dialogue

Page 11: Towards Conversational Human Computer Interaction

Challenges For Practical Dialogue

1. Parsing language in practical dialogues

2. Integrating dialogue and task performance

3. Intention recognition

4. Mixed initiative dialogue

Page 12: Towards Conversational Human Computer Interaction

1.Parsing Language In Practical Dialogues Need to produce a detailed semantic

representation the captures what the user meant by the utterance Put limitations on the different meanings of the words Encode restrictions to make sure words are in the

right context

Spoken language is not sentence based, rather a single utterance may realize a sequence of communicative acts called Speech Acts The grammar used is an act description rather than

sentence structures

Page 13: Towards Conversational Human Computer Interaction

Speech Acts

EXAMPLE: “OK, let’s do that then

send a truck to Avon” OK: an

acknowledgement Let’s do that: an

acceptance Send a truck to

Avon: a request

Page 14: Towards Conversational Human Computer Interaction

2. Integrating Dialogue and Task Performance We need dialogue systems that can be

adapted to any practical tasks Requires an AGENT BASED FRAMEWORK

Dialogue Systems:▪ Knows about the tasks that needs to be accomplished ▪ The generic system is specialized to the particular

domain by integrating domain specific information Broker : link between the dialogue system and

backend Backend: a set of agents providing services

Page 15: Towards Conversational Human Computer Interaction

The Agent Based Architecture

Page 16: Towards Conversational Human Computer Interaction

Abstract Problem Solving and Domain-Specific Task Models The key to this enterprise is the development of an

abstract problem solving model with: ▪ Objectives▪ Solutions▪ Resources▪ Situations

As utterances can mean different things in different domains:

▪ A domain specific task model: provides mapping from the abstract problem solving model to the operations in a particular domain by specifying what things count as objectives, solutions, resources, and situations

Page 17: Towards Conversational Human Computer Interaction

3. Intention Recognition

Determining what the user wants to say by saying the utterance

Example: Emergency System that will be used to evacuate people from an island in the face of a hurricane

Page 18: Towards Conversational Human Computer Interaction

Example

Page 19: Towards Conversational Human Computer Interaction

Intention Recognition (Cntd) That example shows

different interpretations of those utterances

Reasoning used, requires information about the task to identify which makes sense rationally in the current situation

The challenge here is to match the intentions with the context of the the question asked

Page 20: Towards Conversational Human Computer Interaction

4. Mixed Initiative Dialogue Practical human dialogue involves mixed

initiative interaction i.e. dynamic exchange of control of the dialogue flow, increasing efficiency, effectiveness; enabling both participants'’ needs to be met

For systems that ask questions it can work in simple tasks, like long distance dialing

For complex tasks: you may often need to get

through irrelevant interactions before getting the information you need.

Page 21: Towards Conversational Human Computer Interaction

Limitations of Frame Based System The user has to supply information for

the system, until there is sufficient information for the system to perform the task

BUT The user might not know what

information he might still needs to supplySo Currently system provide limited mixed

initiative interactions

Page 22: Towards Conversational Human Computer Interaction

Conclusion There are still lots of technical issues

remaining to overcome to increase the efficiency and effectiveness of the dialogue based user interfaces.

Thank You