emotional embodied conversational agent name : ranjeet singh fan : sing0258 student-id : 2111524
TRANSCRIPT
Emotional Embodied Conversational Agent
Name : Ranjeet Singh
FAN : sing0258
Student-Id : 2111524
Introduction
Aim
Application
Methodology
Emotion detection
Senti-Wordnet
Continuing work
Introduction
HeadX is a talking Head which converts text into speech and
facial expressions.
HeadX takes text as input from the user and give the appropriate spoken textual response synhronized with emotional facial response.
Introduction(Continued…)
HeadX makes different emotional expressions such as anger, sadness, frustration, joy, surprise.
It uses turn-taking mechanism for managing dialogue.
HeadX has alicebot as its component for managing dialogue.
Introduction(Contd…)
HeadX interacts directly with program called synapse
Synapse performs interprocess communication
A channel for interaction.
Synapse Synapse
Synapse
User HeadX AliceBot
ProgramD
Aim
To mix emotion in the interaction between user and HeadX(ChatBot)
To make the interaction more socialistic rather just talking to animated character.
To detect emotion from the text input and make HeadX give appropriate emotional expression in return.
Application
The product could be used in social networking environment such as twitter and facebook.
It could help manage communication between users
Also useful in educational environment.
Aim(Contd…)
Objective is to immerse emotions in the dialogue.
To detect the emotion from user input(text)
Depending upon detected emotion, HeadX will make relevant facial expression.
To make the interaction more real rather just talking to animated character.
HeadX is being used as interface agent to work upon.
Methodology(Contd…)
Text-based emotion detection
To detect emotion from user input requires pre-processing of user input.
After preprocessing, emotion is detected
According to detected emotion, HeadX display its facial response.
Emotion Detection
Emotion could be detected through different approaches
Such as keyword spotting technique.
Keyword spotting technique is being used to detect emotion.
Words conveying emotion are being spotted in the user text
eg- in sentence “ I am wonderful” word “wonderful” conveys joy emotions.
Joy Sad Anger Fear Surprise
FascinatingMerry Goodglad
FearfulHorrorAnxiousAwful
AngryHateful envy Jealousy
Poorbad Sadnesssorry
Fantastic Amazing Astonish wonderful
Emotion detection(Contd…)
User text is being pre-processed to generate emotional output.
Detected Emotion
Tokenization
Identify emotion words
Checking intensifier
s
Checking NegativityUser Input
VeryToo
Quite
Emotion Detection(Contd..)
To detect emotion, the system look for this word in list of emotional words.
What if the emotional word being entered by user is not in the list.
For detecting emotion for words out of range, senti-wordnet is being used.
Senti-wordnet
A lexical resource containing words with their polarity.
Senti-wordnet assigns polarity:- positivity, negativity to a word.
Polarity indicates the sentiment value of particular word.
If words has –ve polarity, then sad emotion and joy for +ve polarity.
Continuing work
Emotion detection from text is complete.
Next step to manage the dialogue between user and HeadX.
User
HeadX
AliceBot(Dialogu
e Manage
r)
ProgramD
Rbot(DialogueMana
ger)
References
M. Shelke 2013, Approaches towards Emotion Extraction from Text, National Conference on Innovative Paradigms in Engineering & Technology .
Erik Boiy; Pieter Hens; Koen Deschacht; Marie-Francine Moens 2007, Automatic Sentiment Analysis in On-line Text.
Ameeta Agrawal, Aijun An 2012, Unsupervised Emotion Detection from Text using Semantic and Syntactic Relations.
Swati D. Bhutekar, Manoj. B. Chandak, A..J.Agrawal 2012, Emotion Extraction: machine learning for text-based emotion.
Jianhua Tao, Context Based Emotion Detection from Text Input, National Laboratory of Pattern Recognition Institute of Automation, Chinese Academy of Sciences, Beijing, China.