Prediction of Helpful Reviews Using Emotions Extraction

Authors: Lionel Martin, Pearl Pu

AAAI 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Using GALC, a general lexicon of emotional words associated with a model representing 20 different categories, we extracted the emotionality from the review text and applied supervised classification method to derive the emotion-based helpful review prediction. As the second contribution, we propose an evaluation framework comparing three different real-world datasets extracted from the most well-known product review websites. This framework shows that emotion-based methods are outperforming the structure-based approach, by up to 9%.
Researcher Affiliation Academia Lionel Martin and Pearl Pu Human Computer Interactions Group School of Computer and Communication Sciences Swiss Federal Institute of Technology (EPFL) CH 1015, Lausanne, Switzerland {lionel.martin, pearl.pu}@epfl.ch
Pseudocode No No pseudocode or algorithm blocks are explicitly presented in the paper.
Open Source Code No The paper does not provide any concrete access information (specific repository link, explicit code release statement, or code in supplementary materials) for the source code of the methodology described.
Open Datasets Yes We based our research on three datasets extracted from different product review websites. A first dataset of 68,049 reviews has been crawled from Trip Advisor... A second one contains reviews from Yelp... for the 2013 Recsys Challenge.4 Finally our last dataset contains reviews from Amazon... 1www.amazon.com 2www.yelp.com 3www.tripadvisor.com 4www.kaggle.com/c/yelp-recsys-2013
Dataset Splits Yes In all cases the evaluation is performed over a 10-fold stratified cross-validation: the dataset is separated into 10 chunks of equal size and class distribution, and 10 rounds are performed where each time a tenth of the dataset constitute the testing set while the rest is the training set; the outputted results are the averages of the ten runs.
Hardware Specification No The paper does not provide any specific hardware details (GPU/CPU models, processor types, memory amounts, or detailed computer specifications) used for running its experiments.
Software Dependencies No The paper mentions machine learning algorithms such as Support Vector Machine, Random Forest, and Naive Bayes, and refers to the GALC lexicon, but it does not specify any version numbers for these or other software dependencies.
Experiment Setup Yes First, we tune the parameters of the algorithms such as the kernel function (radial basis functions are performing the best), the penalty parameter C, and the kernel coefficient γ for SVM following (Hsu, Chang, and Lin 2003) s recommendation for grid-searching over 110 combinations.