Predicting Readers’ Sarcasm Understandability by Modeling Gaze Behavior

Authors: Abhijit Mishra, Diptesh Kanojia, Pushpak Bhattacharyya

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental By recording and analyzing the eye-gaze data, we show that eyemovement patterns vary when sarcasm is understood visa-vis when it is not. Motivated by our observations, we propose a system for sarcasm understandability prediction using supervised machine learning. Our system relies on readers eyemovement parameters and a few textual features, thence, is able to predict sarcasm understandability with an F-score of 93%, which demonstrates its efficacy.
Researcher Affiliation Academia Abhijit Mishra, Diptesh Kanojia, Pushpak Bhattacharyya Center for Indian Language Technology Department of Computer Science and Engineering Indian Institute of Technology Bombay, India, {abhijitmishra, diptesh, pb}@cse.iitb.ac.in
Pseudocode No The paper describes the proposed system and its components in narrative text but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide an explicit statement or a link indicating that the source code for their methodology is publicly available.
Open Datasets Yes Our database can be freely downloaded3 for academic purposes. [Footnote 3]: http://www.cfilt.iitb.ac.in/cognitive-nlp/. Also, '76 sarcastic short movie reviews were manually extracted from the Amazon Movie Corpus (Pang and Lee )'.
Dataset Splits Yes After transforming our dataset to a multi-instance dataset, we performed a 5-fold cross validation. Each fold has a train-test split of 80%-20% and each split contains examples from all seven participants.
Hardware Specification No The paper mentions 'an SR-Research Eyelink-1000 eye-tracker (monocular remote mode, sampling rate 500Hz)' for data collection, but does not provide specific hardware details (like GPU/CPU models or memory) used for running the supervised learning experiments.
Software Dependencies No The paper states 'We apply Multi-instance Logistic Regression (MILR) ... implemented using the Weka API (Hall et al. 2009)' but does not specify a version number for the Weka API or any other software dependencies.
Experiment Setup No The paper describes the use of Multi-instance Logistic Regression and a 5-fold cross-validation setup, but does not provide specific hyperparameters or detailed training configurations (e.g., learning rates, batch sizes, number of epochs, or optimizer settings) for the model.