https://github.com/shreyansh26/revopid
Review Opinion Diversification - Shared task in IJCNLP 2017, Taipei, Taiwan
https://github.com/shreyansh26/revopid
clustering doc2vec infersent liu mining-opinion-features review-opinion-diversification
Last synced: 8 months ago
JSON representation
Review Opinion Diversification - Shared task in IJCNLP 2017, Taipei, Taiwan
- Host: GitHub
- URL: https://github.com/shreyansh26/revopid
- Owner: shreyansh26
- License: apache-2.0
- Created: 2017-10-27T09:40:18.000Z (over 8 years ago)
- Default Branch: master
- Last Pushed: 2018-11-13T07:35:32.000Z (over 7 years ago)
- Last Synced: 2025-03-24T18:52:36.264Z (about 1 year ago)
- Topics: clustering, doc2vec, infersent, liu, mining-opinion-features, review-opinion-diversification
- Language: Jupyter Notebook
- Homepage: https://shreyansh26.github.io/RevOpiD/
- Size: 79.5 MB
- Stars: 6
- Watchers: 2
- Forks: 1
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
Review Opinion Diversification
==============================
A project I worked on in my 3rd Semester. [RevOpiD](https://sites.google.com/itbhu.ac.in/revopid-2017) was a shared task in IJCNLP 2017, Taipei, Taiwan.
I implemented three approaches of solving the problem -
1. Opinion Feature Mining (**Mining Opinion Features in Customer Reviews** (Liu et al.))
2. Doc2Vec Model and Clustering
3. Facebook Research's [InferSent](https://github.com/facebookresearch/InferSent) model for sebtence embeddings with Clustering
Three clustering techniques were tried -
* K-Means Clustering
* Spectral Clustering
* Agglomerative Clustering
Of the three, Spectral Clustering gave the best results on manually checking the contents of the clusters. So in all the description of the models, clustering implies Spectral Clustering.
This project was done under the guidance of [Dr. A.K. Singh](https://www.iitbhu.ac.in/dept/cse/people/aksinghcse), Associate Professor, Department of Computer Science and Engineering, IIT (BHU) Varanasi.
Also, special thanks to [Avijit Thawani](https://avi-jit.github.io/) for his mentorship and the intersting ideas he provided to solve the problems I faced in the way.