/
OMG UR OMG UR

OMG UR - PowerPoint Presentation

marina-yarberry
marina-yarberry . @marina-yarberry
Follow
374 views
Uploaded On 2017-11-12

OMG UR - PPT Presentation

  Funny ComputerAided Humor with an Application to Chat miaomiao Wen Nancy Baym Omer Tamuz Jaime Teevan Susan Dumais Adam Kalai Outline Introduction System Design Image Suggestion Strategies ID: 604880

computer humor aided chat humor computer chat aided application images image funny retrieval suggestions meme amp based introductionsystem insightsconclusions

Share:

Link:

Embed:

Download Presentation from below link

Download Presentation The PPT/PDF document "OMG UR" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.


Presentation Transcript

Slide1

OMG UR Funny! Computer-Aided Humor with an Application to Chat

miaomiao Wen

Nancy

Baym

Omer

Tamuz

Jaime

TeevanSusan DumaisAdam KalaiSlide2

OutlineIntroductionSystem DesignImage Suggestion Strategies

Experiments & ResultsQualitative InsightsConclusions

Computer-Aided Humor with an Application to Chat

2Slide3

OutlineIntroductionSystem Design

Image Suggestion StrategiesExperiments & ResultsQualitative InsightsConclusions

Computer-Aided Humor with an Application to Chat

3Slide4

Motivation Computer-Aided Humor with an Application to Chat

4

Can computers be funny?Slide5

Simsimi http://www.simsimi.com/

Computer-Aided Humor with an Application to Chat 5Slide6

Motivation Computer-Aided Humor with an Application to Chat

6

Can computers help people be funny?Slide7

OutlineIntroductionSystem Design

Image Suggestion StrategiesExperiments & ResultsQualitative InsightsConclusions

Computer-Aided Humor with an Application to Chat

7Slide8

Computer Generated Humor vs. Computer Aided Humor

Computer-Aided Humor with an Application to Chat 8Slide9

The Iterative Design ProcessWe started with text-based humor suggestions based on prior work [e.g.

Valitutti et al. 2009]PunsCanned jokes“that’s what she said”Etc.Negative user feedback from crowdsourced

workers“The jokes might be funny for a three year old”

“The suggestions are very silly” Suffered from lack of context

Most human jokes are produced within humorous frames and rarely communicate meanings outside it [Dynel 2009]

Computer-Aided Humor with an Application to Chat

9Slide10

Image vs. Text Suggestions

Success of adding a meme image into suggestionsImage vs. Textimages are often more open to interpretation than text

images are slower for users to provide on their own than entering text

images provide much more context on their own Computer-Aided Humor with an Application to Chat

10

We will focus on generating

funny images suggestions.Slide11

User Interface Computer-Aided Humor with an Application to Chat

11

Chat History

Input Box

Generated

SuggestionsSlide12

OutlineIntroduction

System DesignImage Suggestion StrategiesExperiments & ResultsQualitative InsightsConclusions

Computer-Aided Humor with an Application to Chat

12Slide13

Strategy 1. Emotional Reaction Images and gifsEmoticons can support nonverbal communications

[Walther and Kyle 2001]Emoticons are often used to display or support humor [Tossell et al 2012]Many chat apps provide emoticon libraries

Computer-Aided Humor with an Application to Chat

13Slide14

Emotional Reaction Images and gifs

Detected SentimentSuggestion Reaction ImageQuestion

“I don’t know”

AngerMad and Confused

DisgustGross

JoyLove and Win

SadnessSad and Face-palm

SurpriseOMG Computer-Aided Humor with an Application to Chat 14

http://

giphy.com

http:/

/

reacticons.comSlide15

Strategy 2. Image RetrievalImage retrieval

using Bing image search API“funny keyword(s)” Computer-Aided Humor with an Application to Chat

15

Mturk

is like a desert today. Not a ton of great HITs.Slide16

Image RetrievalImage retrieval using Bing

image search API“funny keyword(s)”Keywords chosen based on tf-idf Computer-Aided Humor with an Application to Chat

16

Mturk

is like a

desert

today. Not a ton of great HITs.Slide17

Image RetrievalImage retrieval using Bing image search API

“funny keyword(s)”keywords chosen based on tf-idf Computer-Aided Humor with an Application to Chat

17Slide18

Image RetrievalImage retrieval using Bing

image search API“funny keyword(s)”keywords chosen based on tf-idf Computer-Aided Humor with an Application to Chat

18Slide19

Image RetrievalImage retrieval using Bing

image search API“funny keyword(s)”keywords chosen based on tf-idf Computer-Aided Humor with an Application to Chat

19Slide20

Strategy 3. Meme Images“Viral images, videos, and catchphrases under constant modification by users” [Coleman 2012

]a meme template is an image of a meme character without the captionsUsers generate their own memes pictures through meme generation websitewrites the last utterance on the top and bottom of a popular meme

template

Computer-Aided Humor with an Application to Chat

20Slide21

Meme ImagesMachine-learning trained classifier picks the most suitable meme template image based on the last utteranceTop 50 popular meme templates and their instances from

memegenerator.netBag-of-word featuresAccuracy 51% Computer-Aided Humor with an Application to Chat

21Slide22

Strategy 4. Responding to Images with ImagesUsers often responding to images with similar images

Computer-Aided Humor with an Application to Chat 22Slide23

Responding to Images with ImagesRespond-in-kind strategyan image chosen from a search for “funny keyword

” with a second image from the same searchRule-of-threea common joke structure [Quijano 2012]comedians use the first two points to establish a pattern throw

the audience off with the third element

“funny keyword” -> “no

keyword” -> “many keyword”

Computer-Aided Humor with an Application to Chat

23Slide24

Responding to Images with Images

Computer-Aided Humor with an Application to Chat 24

“funny dog”

“no dog”

“many dog”Slide25

Responding to Images with ImagesResponse to images with emoticons“LOL”

“not-amused” Computer-Aided Humor with an Application to Chat

25Slide26

Ranking Suggestions using Reinforcement Learning Rank images based on how often it was chosen before

Back-off: how often the strategy was chosen before Computer-Aided Humor with an Application to Chat

26Slide27

OutlineIntroduction

System DesignImage Suggestion StrategiesExperiments & ResultsQualitative InsightsConclusions

Computer-Aided Humor with an Application to Chat

27Slide28

Experiments Three SystemsComputer-aided Humor (with 1,2,3,6,10 suggestions)

Computer-generated Humorautomatically insert the top-ranking imagePlain chatPair of crowdsourced workers

chatted for 10 minutes,

no topic

restrictions738 participantsAt least 100 participants using each system

Evaluation: Post chat survey [Jiang et al. 2011]The

conversation was fun. I was able to express my sense of humor in this conversation. I felt pretty close to my partner during the conversation. I was involved in the conversation. Computer-Aided Humor with an Application to Chat

28Slide29

Experimental Results:Number of Suggestions

Computer-Aided Humor with an Application to Chat 29Slide30

Experimental Results

Computer-Aided Humor with an Application to Chat 30Slide31

Experimental Results:Computer Generated Humor vs. Computer Aided Humor

Computer-Aided Humor with an Application to Chat 31Slide32

Experimental Results:Computer Aided Humor v.s. Plain Chat

Computer-Aided Humor with an Application to Chat 32Slide33

Experimental Results:Which strategies are more successful?

#suggestions% chosenBing Images

44,710

10%Reaction Images & gifs

4,37519%

Meme709

13%Rule-of-three69824%

Computer-Aided Humor with an Application to Chat 33Slide34

OutlineIntroduction

System DesignImage Suggestion StrategiesExperiments & ResultsQualitative InsightsConclusions

Computer-Aided Humor with an Application to Chat

34Slide35

Qualitative Insights (1)Humorous Images Bring New Topics to the Conversation

Computer-Aided Humor with an Application to Chat 35Slide36

Qualitative Insights (2)Yes, and… first accept one another’s suggestions and try to build them into something even funnier

[Moshavi 2001] Our strategies can lead to yes-and behaviors

Computer-Aided Humor with an Application to Chat

36Slide37

Qualitative Insights (3)Image Humor is Robust

Participants had different backgrounds and thus understand images differentlysome participants did not recognize that the images were memesstill find the images amusingEndless funny images online

Positive feedback“It should be used for online speed dating!”

“When will this app be available for phones and whatnot? I want to use it ”

Computer-Aided Humor with an Application to Chat

37Slide38

OutlineIntroduction

System DesignImage Suggestion StrategiesExperiments & ResultsQualitative InsightsConclusions

Computer-Aided Humor with an Application to Chat

38Slide39

ConclusionsA chat system that builds on the relative strengths of people and computers to generate humor by suggesting images

Compared to plain chat and a computer generated humor system more fun enabled them to express their sense of humor

Computer-Aided Humor with an Application to Chat

39Slide40

Q & AThanks for your attentions!

!Miaomiao Wenmwen@cs.cmu.edu

Computer-Aided Humor with an Application to Chat

40

Related Contents


Next Show more