Learning Latent Opinions for Aspect-level Sentiment Classification

Authors: Bailin Wang, Wei Lu

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Extensive experiments show that our model achieves the state-of-the-art performance while extracting interpretable sentiment expressions.
Researcher Affiliation Academia Bailin Wang College of Information and Computer Sciences University of Massachusetts Amherst, MA 01003, USA bailinwang@cs.umass.edu Wei Lu Singapore University of Technology and Design 8 Somapah Road Singapore, 487372 luwei@sutd.edu.sg
Pseudocode No The paper does not contain structured pseudocode or algorithm blocks.
Open Source Code Yes Our implementation is available at https://github.com/berlino/SA-Sent
Open Datasets Yes We perform sentiment classification tasks on publicly available datasets on online reviews across different languages from Sem Eval tasks and social comments from Twitter. The first group consists of review datasets from Sem Eval2014 task 4 (Pontiki et al. 2014) and Twitter comments collected by (Dong et al. 2014).
Dataset Splits Yes One-sixth of training data is left out as the validation set for tuning hyperparameters and doing model selection.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper mentions software components like "Glo Ve" and "Adam" but does not specify their version numbers.
Experiment Setup Yes The dimension of target s binary indicator embedding is 30. We fixed the word embeddings in all the experiments. Dropout is also used after the input layer and it is tuned for each dataset. λ1 is tuned between 0 and 1, λ2 is chosen from [0, 0.2] with step size 0.04. For LSTM, we set the hidden dimension size to 300.