Project assignment for the ML and SII courses at Roma Tre university.
The work is based mainly on the approach proposed by Poria et al. in "Aspect extraction for opinion mining with a deep convolutional neural network". A Convolutional Neural Network has been implemented, along with Word Embedding using pre-trained embeddings from GloVe and POS tagging using the Stanford POS tagger.
Files description:
- AspectExtraction_CNN.ipynb: the notebook containing all code, models, word embedding procedures etc.
- glove.6B.300d.txt: the pre-trained embeddings from GloVe
- Restaurants.xml and Laptop.xml: the datasets used for Training and Validation
- RapportoAttività.pdf: the project report
[1] Poria et al., 2016 "Aspect extraction for opinion mining with a deep convolutional neural network" http://www.sentic.net/aspect-extraction-for-opinion-mining.pdf