7. Sex impact on Silliness Prosody (Results) Representing sex contrasts with 2-way ANOVA The test demonstrates –.

Uploaded on:
Silliness: Prosody Examination and Programmed Acknowledgment For F.R.I.E.N.D.S Amruta Purandare and Diane Litman College of Pittsburgh EMNLP 2006, Sydney, Australia 1. Inspiration Need social knowledge in PCs Approaches: Influence, Identity, Funniness?
Slide 1

Diversion: Prosody Analysis and Automatic Recognition For F.R.I.E.N.D.S Amruta Purandare and Diane Litman University of Pittsburgh EMNLP 2006, Sydney, Australia 1. Inspiration Need social knowledge in PCs Approaches: Affect, Personality, Humor? Best in class in Computational Humor: Humorous Text (Acronyms, One-liners, Wordplays) Lexical signs (Alliteration, Slang, Antonymy) Our commitment: Humor Detection in Spoken Conversations Do Prosodic prompts (e.g. Pitch, Intensity, Tempo) help? 6. Funniness Prosody Analysis (Results) Most prosodic elements show critical (p<=0.05 for a t-test) contrasts in the middle of Humor and Non-Humor bunches Humorous turns show higher Max, Range, Std-Dev in Pitch and Energy, higher Tempo and littler Internal Silence 7. Sexual orientation impact on Humor-Prosody (Results) Accounting for sex contrasts with 2-way ANOVA The test shows – Humor impact on prosody balanced for Gender impact on prosody balanced for Humor Interaction impact in the middle of Gender and Humor i.e. in the event that the prosodic style of communicating silliness is distinctive for Males and Females Findings: Significant impact of Humor notwithstanding when balanced for Gender 2) Significant impact of Gender, however just Pitch components demonstrate the Interaction Effect. i.e. guys and females use diverse Pitch varieties while communicating Humor 2. Companions Corpus 75 Dialogs from an excellent TV-satire: FRIENDS 2hrs of Audio Text transcripts from: http://www.friendscafe.org/scripts Humorous turns are trailed by snickers Automatic naming utilizing giggles Corpus size = 1629 turns 714 Humorous, 915 Non-Humorous 6 Main Actors (3 Male, 3 Female), 26 Guest Actors Y: noteworthy impact N: non-critical impact 3. Illustration Dialog Rachel: Guess what? [no] Ross: You landed a position? [no] Rachel: Are you joking? I am prepared in vain! [yes] <laugh> Rachel: I was giggled out of 12 meetings today. [no] Chandler: but you are shockingly perky! [no] Rachel: Well, you would be as well, in the event that you discovered John & David’s boots discounted, half off... [yes] <laugh> Chandler: Oh how well you know me! [yes] <laugh> Rachel: They are my new, I don’t need work, I don’t need my guardians, I got extraordinary Boots, Boots! [yes] <laugh> [yes] Humorous Turns [no] Non-Humorous Turns 8. Silliness Recognition (Results) Supervised 2-way order Results above benchmark (56.2%) Results steady for sexual orientations Marginal change higher for guys Decision tree demonstrates that the calculation picked generally prosodic and speaker highlights in the initial 10 emphasess 4. Elements Borrowed from passionate discourse writing Prosodic (13) Pitch (F0): Mean, Max, Range, Std-Dev Energy (RMS): Mean, Max, Range, Std-Dev Temporal: Duration, Internal Silence, Tempo Lexical (2011) all Words Turn Length (#words in the turn) Speaker ID (1) 9. Conclusions & Future Work Humor acknowledgment in talked discussions Data Dialogs from a fantastic satire TV show, FRIENDS Used snickers for consequently naming hilarious turns Humor-Prosody Analysis Humorous turns show higher tops and varieties in pitch and vitality, and higher rhythm, contrasted with non-funny turns Gender Effect Most elements show amusingness impact notwithstanding when balanced for sexual orientation Only pitch components demonstrate the connection impact Results Promising, 8% over the gauge with all elements Humor identification simpler for male speakers than for females Future Pragmatic elements e.g. Equivocalness, Incongruity, Expectation-Violation and so on 5. Highlight Extraction utilizing Wavesurfer

Slide 2


View more...