34
University of Brescia Department of Information Engineering Knowledge Engineering and Human-Computer Interaction Research Group © 2010 Federico Cerutti <[email protected]> From User-Generated Tagging From User-Generated Tagging to User-Agreed Knowledge: to User-Agreed Knowledge: An Argumentation-Based An Argumentation-Based Approach Approach Pietro Baroni, Federico Cerutti, Daniela Fogli, Claudio Gandelli, Massimiliano Giacomin

Cerutti--NMR 2010

Embed Size (px)

Citation preview

Page 1: Cerutti--NMR 2010

University of BresciaDepartment of Information Engineering

Knowledge Engineering and Human-Computer Interaction Research Group

© 2010 Federico Cerutti <[email protected]>

From User­Generated Tagging From User­Generated Tagging to User­Agreed Knowledge:to User­Agreed Knowledge:An Argumentation­Based An Argumentation­Based 

Approach Approach 

Pietro Baroni, Federico Cerutti, Daniela Fogli,

Claudio Gandelli, Massimiliano Giacomin

Page 2: Cerutti--NMR 2010

Slide 2 © 2010 Federico Cerutti <[email protected]>

A Web 2.0 scenario

Page 3: Cerutti--NMR 2010

Slide 3 © 2010 Federico Cerutti <[email protected]>

MaryCarl John

A Web 2.0 scenario

I heard a slaty-backed gull

in Boulder

It's impossible!Those gulls are indigenous

of East Asia

You are wrong!Look at youtube!

I agree with Mary.

Look at local press.

All right... I'm wrong.

1

2

3

4

5

Page 4: Cerutti--NMR 2010

Slide 4 © 2010 Federico Cerutti <[email protected]>

A Web 2.0 scenario:Pros/Cons

Ambiguity, imprecision, lack of relationships,

poor content findability

Quick, close to users' experience Who says what?

Pro or against who/what?Who wins?

How this dialogue affect tags?Does this dialogue provide

new knowledge?

Page 5: Cerutti--NMR 2010

Slide 5 © 2010 Federico Cerutti <[email protected]>

Keep pros, overcome cons:a demonomy

A shared unambiguous dictionary Formal relationships among tags “Unquestionable” knowledge

An arena Opinion exchanging Formal dialogue, as close as possible to users' experience Defeasible knowledge Can also enrich the knowledge provided by the shared

dictionary

Page 6: Cerutti--NMR 2010

Slide 6 © 2010 Federico Cerutti <[email protected]>

The shared unambiguous dictionary: Gull

Wordnet graphical representation by visuwords.com

Legend

Page 7: Cerutti--NMR 2010

Slide 7 © 2010 Federico Cerutti <[email protected]>

The shared unambiguous dictionary: Slaty-backed Gull

Page 8: Cerutti--NMR 2010

Slide 8 © 2010 Federico Cerutti <[email protected]>

The defeasible knowledge

Object(sound, image, document…)

Tag(<gull>, <slaty-backed gull>…)

Tagrelationship

Dictionary term(<gull, seagull>…)

Semanticrelationship

Page 9: Cerutti--NMR 2010

Slide 9 © 2010 Federico Cerutti <[email protected]>

The defeasible knowledge in the example

Unquestionable knowledge There is the sound (an object) X There is a dictionary term <gull, seagull>

Questionable knowledge tag(X, <gull, seagull>) tag(X, <slaty-backed gull>) sem(<slaty-backed gull>, <gull, seagull>, hyperonomy)

How can we codify and question this knowledge?

Page 10: Cerutti--NMR 2010

Slide 10 © 2010 Federico Cerutti <[email protected]>

“Classic” Argumentation Scheme John's statements

Arg1 [From Appearance](Premise) This object looks like it could be classified under verbal category <gull> (C)(Conclusion) Therefore, this object can be classified under verbal category <gull> (C)

Arg3 [From Def. to verbal classification](Definition Premise) <slaty-backed gull> (a) fits definition <large gull...> (D)(Premise) For all x, if x fits definition D, then x can be classified as having property <hyperonomy of D> (G)(Conclusion) a has property G

(CQ1) Could the appearance of its looking like it could be classified under C be misleading for some reason?(CQ2) Although it may look like it can be classified under C, could there be grounds for indicating that it might be more justifiable to classify it under another category D?

tag(X, <gull, seagull>) sem(<slaty-backed gull>, <gull, seagull>, hyperonomy)

Arg2 [From Appearance](Premise) This object looks like it could be classified under verbal category <slaty-backed gull>(Conclusion) Therefore, this object can be classified under verbal category <slaty-backed gull>

tag(X, <slaty-backed gull>)

(CQ1) Could the appearance of its looking like it could be classified under C be misleading for some reason?(CQ2) Although it may look like it can be classified under C, could there be grounds for indicating that it might be more justifiable to classify it under another category D?

(CQ1) What evidence is there that D is an adequate definition, in light of other possible alternative definition that might exclude a's having G?(CQ2) Is the verbal classification in the classification premise based merely on a stipulative or biased definition that is subject to doubt?

Page 11: Cerutti--NMR 2010

Slide 11 © 2010 Federico Cerutti <[email protected]>

“Classic” Argumentation SchemeMary's critique

Arg2 [From Appearance](Premise) This object looks like it could be classified under verbal category <slaty-backed gull>(Conclusion) Therefore, this object can be classified under verbal category <slaty-backed gull>

(CQ1) Could the appearance of its looking like it could be classified under C be misleading for some reason?(CQ2) Although it may look like it can be classified under C, could there be grounds for indicating that it might be more justifiable to classify it under another category D?

Arg4 [From Position to Know](Major Premise) Source <Mary> (a) is in position to know about things in a certain subject domain <ornithology> (D) containing proposition <Slaty-backed gull is indigenous of East Asia, therefore a specimen of this species cannot be on North America> (A)(Minor Premise) a asserts that A is true(Conclusion) A is true

(CQ1) Is a in position to know whether A is true?(CQ2) Is a an honest (trustworthy, reliable) source?(CQ3) Dis a assert that A is true?

Page 12: Cerutti--NMR 2010

Slide 12 © 2010 Federico Cerutti <[email protected]>

“Classic” Argumentation SchemeJohn's answer

Arg5 [From Expert Opinion](Major Premise) Source <youtube...> (E) is an expert in subject domain S containing proposition A(Minor premise) E asserts that proposition A is true(Conclusion) A is true

(CQ1) How credible is E as an expert source?(CQ2) Is E an expert in the field that is in?(CQ3) What did E assert that implies A?(CQ4) Is E personally reliable as a source?(CQ5) Is A consistent with what other experts assert?(CQ6) Is E's assertion based on evidence?

(CQ1) Is a in position to know whether A is true?(CQ2) Is a an honest (trustworthy, reliable) source?(CQ3) Dis a assert that A is true?

Arg4 [From Position to Know](Major Premise) Source <Mary> (a) is in position to know about things in a certain subject domain <ornithology> (D) containing proposition <Slaty-backed gull is indigenous of East Asia, therefore a specimen of this species cannot be on North America> (A)(Minor Premise) a asserts that A is true(Conclusion) A is true

Page 13: Cerutti--NMR 2010

Slide 13 © 2010 Federico Cerutti <[email protected]>

“Classic” Argumentation SchemeCarl's claim

Arg5 [From Expert Opinion](Major Premise) Source <youtube...> (E) is an expert in subject domain S containing proposition A(Minor premise) E asserts that proposition A is true(Conclusion) A is true

(CQ1) How credible is E as an expert source?(CQ2) Is E an expert in the field that is in?(CQ3) What did E assert that implies A?(CQ4) Is E personally reliable as a source?(CQ5) Is A consistent with what other experts assert?(CQ6) Is E's assertion based on evidence?

Arg6 [From Expert Opinion](Major Premise) Source <local press> (E) is an expert in subject domain S containing proposition A(Minor premise) E asserts that proposition A is true(Conclusion) A is true

(CQ1) How credible is E as an expert source?(CQ2) Is E an expert in the field that is in?(CQ3) What did E assert that implies A?(CQ4) Is E personally reliable as a source?(CQ5) Is A consistent with what other experts assert?(CQ6) Is E's assertion based on evidence?

Page 14: Cerutti--NMR 2010

Slide 14 © 2010 Federico Cerutti <[email protected]>

Problematic issues due to the Web Naïve Users

Users do not know what they want to express Users do not know how they should express what

they want Users do not know how to critique other users

Analise the knowledge the users want to share Provide a simplified version of reasoning pattern Balance the burden of proof (avoid fallacious, but

likely in this context, arguments and attacks)

Page 15: Cerutti--NMR 2010

Slide 15 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationRequirements

1 Easy and close to user' experience2 The less knowledge is required, the more it is

appreciated3 Arguments and critiques should be generated in

an obfuscated (for the users) way4 Every user is committed to justify his/her

counterargument5 Automatic evaluation of the acceptability of

arguments through semantics analysis

Page 16: Cerutti--NMR 2010

Slide 16 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationThe Goal

Arg2...

Arg4...

Arg2 Arg4GroundedExtension

Page 17: Cerutti--NMR 2010

Slide 17 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationArguments (1)

Arg1 [AUT]User <john> says that the tag <gull, seagull> (T) is related to the content of the sound <seagull> (S), so, the tag T has been added to S sound.

Arg2 [AUT]User <john> says that the tag <slaty-backed gull> (T) is related to the content of the sound <seagull> (S), so, the tag T has been added to S sound.

Page 18: Cerutti--NMR 2010

Slide 18 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationArguments (2)

Arg3 [AUDR]User <john> says that the word <slaty-backed gull> (W) is related to the synset <gull, seagull> (S) according to the <hyponomy> (R) relation, so, the relation R has been added between W and S.

Page 19: Cerutti--NMR 2010

Slide 19 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationCounterarguments (1)

AUT

Why do you say IC1?

Your ownjustification?

Have you got somesort of proof?

Do you know an external informationsource that contradicts user’s claim?

AUDR

Why do you say IC1?

Your ownjustification?

Have you got somesort of proof?

Do you know an external informationsource that contradicts user’s claim?

Page 20: Cerutti--NMR 2010

Slide 20 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationCounterarguments (2)

Arg4 [APK]User <mary> (u) knows that <slaty-backed gull is indigenous of East Asia> (C) holds, and, from C, u derives that <”slaty-backed gull” is not related to “seagull” sound> (a) holds. So it should be the case that a.

Page 21: Cerutti--NMR 2010

Slide 21 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationCounterarguments (3)

Arg5 [ASK]The user <john> says that, acording to the source of knowledge <youtube> (S), it holds that <fromthe fact that the slaty-backed gull is indigenous to East Asia we cannot know that “slaty-backed gull” tag is not related to “seagull” sound> (a). So, it should be the case that a.

Arg6 [ASK]The user <carl> says that, acording to the source of knowledge <local press> (S), it holds that <”youtube” is not credible as a source of knowledge> (a). So, it should be the case that a.

Page 22: Cerutti--NMR 2010

Slide 22 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationThe graph for (counter)arguments (1)

Arg2

Arg1 Arg3`

`

`

Page 23: Cerutti--NMR 2010

Slide 23 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationThe graph for (counter)arguments (2)

Arg4

Arg2

Page 24: Cerutti--NMR 2010

Slide 24 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationThe graph for (counter)arguments (3)

Arg4Arg5

Page 25: Cerutti--NMR 2010

Slide 25 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationThe graph for counterarguments (4)

Arg6 Arg5

Page 26: Cerutti--NMR 2010

Slide 26 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationDialogue as arguments exchange

A dialogue as a sequence of commitments: Propositional commitments

Critique commitments

Page 27: Cerutti--NMR 2010

Slide 27 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationDialogue protocol

Page 28: Cerutti--NMR 2010

Slide 28 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationThe dialogue in the example

Page 29: Cerutti--NMR 2010

Slide 29 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationDetermine the dialogue outcome

Page 30: Cerutti--NMR 2010

Slide 30 © 2010 Federico Cerutti <[email protected]>

User-centred argumentationThe dialogue outcome

Arg1 [AUT]User <john> says that the tag <gull, seagull> (T) is related to the content of the sound <seagull> (S), so, the tag T has been added to S sound.

Arg2 [AUT]User <john> says that the tag <slaty-backed gull> (T) is related to the content of the sound <seagull> (S), so, the tag T has been added to S sound.

Arg3 [AUDR]User <john> says that the word <slaty-backed gull> (W) is related to the synset <gull, seagull> (S) according to the <hyponomy> (R) relation, so, the relation R has been added between W and S.

Arg4 [APK]User <mary> (u) knows that <slaty-backed gull is indigenous of East Asia> (C) holds, and, from C, u derives that <”slaty-backed gull” is not related to “seagull” sound> (a) holds. So it should be the case that a.

Arg5 [ASK]The user <john> says that, acording to the sourche of knowledge <youtube> (S), it holds that <fromthe fact that the slaty-backed gull is indigenous to East Asia we cannot know that “slaty-backed gull” tag is not related to “seagull” sound> (a). So, it should be the case that a.

Arg6 [ASK]The user <carl> says that, acording to the sourche of knowledge <local press> (S), it holds that <”youtube” is not credible as a source of knowledge> (a). So, it should be the case that a.

Page 31: Cerutti--NMR 2010

Slide 31 © 2010 Federico Cerutti <[email protected]>

Result for the user

Page 32: Cerutti--NMR 2010

Slide 32 © 2010 Federico Cerutti <[email protected]>

Conclusion

“Demonomy” Shared unambiguous dictionary An arena

User-centred argumentation A set of innovative reasoning patterns specifically

designed, in this case, for Web 2.0 Users formalised as argument schemes

The less knowledge required, the more appreciated Every user is committed to justify his/her counterargument Automatic evaluation of the justification status of

arguments through semantics analysis

Page 33: Cerutti--NMR 2010

Slide 33 © 2010 Federico Cerutti <[email protected]>

Future works

More (and more general) argument schemes More (and more general) invalidity conditions Study and encompass more Web 2.0 users' needs Voting mechanism Running prototype

Users' evaluation Feedback about users' experience

Page 34: Cerutti--NMR 2010

© 2010 Federico Cerutti <[email protected]>

From User-Generated Tagging to From User-Generated Tagging to User-Agreed Knowledge:User-Agreed Knowledge:

An Argumentation-Based Approach An Argumentation-Based Approach

Thank you