/MG-LDA

NLP-Implementation and Evaluation of Multi-Grained LDA on the SentiPers dataset

Primary LanguagePython

MG-LDA

NLP Persian Implicit Aspect Extraction from the SentiPers dataset using Multi-Grained LDA

This repository hosts an implementation of the MG-LDA algorithm, as stated in the original paper. It is used to extract aspects from the SentiPers dataset in an unsupervised manner, which can be further exploited in Persian Sentiment Analysis.
SentiPers consists of numerous reviews, each having their opinions and targets labeled, thus being very beneficial for Sentiment Analysis.
In this project, a rigorous preprocessing procedure has been applied (Stopwords, rephrasing slangs, etc.) to produce the best input for our model. Nevertheless, there is still room for improvement, so feel free to modify the data or contact me in case of any questions. The elicited aspects are stored in the results directory.