Acquiring Commonsense Knowledge for Sentiment Analysis through Human Computation

Authors: Marina Boia, Claudiu Musat, Boi Faltings

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

Reproducibility Variable Result LLM Response
Research Type Experimental Section 7 presents our experiments. 7 Experiments and Results We first assessed the performance of our context-dependent lexicon on the source vacuum and camera domains, using the corpus Testvac-cam game . As baselines, we took several context-independent lexicons and a machine learning model. Our lexicon CLgame gave an accuracy of 85.17% (Figure 2).
Researcher Affiliation Academia Marina Boia and Claudiu Cristian Musat and Boi Faltings Ecole Polytechnique F ed erale de Lausanne Artificial Intelligence Laboratory CH-1015 Lausanne, Switzerland {marina.boia, claudiu-cristian.musat, boi.faltings}@epfl.ch
Pseudocode No The paper describes the system and process in text but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide an explicit statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes To create the game rounds, we used Amazon3 product reviews for four categories of vacuum cleaners and digital cameras, respectively. (footnote 3: www.amazon.com)
Dataset Splits Yes randomly split the reviews from each category into train and test data, using a ratio of roughly 2:1.
Hardware Specification No The paper does not specify the exact hardware (e.g., CPU, GPU models, memory) used for running the experiments, only general statements like 'We deployed the game on Amazon Mechanical Turk' which is a platform, not hardware specs.
Software Dependencies No The paper does not provide specific software dependencies with version numbers, such as programming languages, libraries, or frameworks used for the implementation.
Experiment Setup Yes We launched several HITs, mainly using a base payment of $0.25, approved with the first 100 points earned, and bonuses established based on game activity: $0.05 for every additional 100 points earned, and $0.05 for every 500 points milestone reached. We obtained the statistical model Lsvm by training a support vector machine on a subset of 2,000 reviews from Trainvac-cam, thus comparable in size with the corpus in the game (we used a linear kernel and 1,000 unigram presence features).