/digikala-sentiment-analysis

Aspect based sentiment analysis on digikala products comment.

Primary LanguageJupyter Notebook

digikala-sentiment-analysis

Aspect based sentiment analysis on digikala products comment. This project was a university group project. only a portion of this code is written by me.

Methods:

  1. Multilingual PyAbsa model.
  2. English PyAbsa model.
  3. Bert model fine tuned using ParsABSA dataset.
  4. Using LLM for the whole thing.(Llama,phi-3, Gemma)
  5. Using an LLM to translate then use english models to extract aspects and sentiments.

Installation

git clone <url>
cd <project>
pip install virtualenv # (if you don't already have virtualenv installed)
virtualenv venv # to create your new environment (called 'venv' here)
source venv/bin/activate # to enter the virtual environment
pip install -r requirements.txt # to install the requirements in the current environment

Refrences:

Github: Aspect-term extraction and Aspect-based sentiment analysis
ABSA datasets for PyABSA
Aspect-based Sentiment Analysis
Quick intro to Aspect-Based Sentiment Analysis
ParsBERT (v3.0)