coldfrenzy / word-sense-disambiguation

Natual Language Processing Project

The details about the project are described in the pdf file.
This readme describes the steps needed to follow in order to succesfully test the various scripts.

Requirements

This project requires tensorflow 1.13, python 3.6
and other packages that can be installed through the following commands:

• pip install lxml
• pip install nltk
• pip install tensorflow_hub
• pip install tqdm
• pip install json

In order to run the predict_multilingual function, you need to install the ElmoForManyLangs package that can be found at the following link:


ElmoForManyLangs


The main ElmoForManyLangs repository should be placed in the "Dataset/Multilingual/" directory.
In the same page you can find download the different model (it,de,es,fr). Successively you should unzip these model in the respective "Dataset/Multilingual/LANG" directory.
You should also insert the file "lemma2synsets4.0.xx.wn.ALL.txt" in the Multilingual directory, since it was too big to be uploaded here.
In the "Dataset/Eval" directory you should place the evaluation directories (semeval2007,semeval2013,semval2015,senseval2,senseval3) available in the Raganato's framework.
In the "Dataset/Train" directory you should place the semcor.data.xml and the semcor.gold.key.txt files available in the Raganato's framework.
In order to use the different vocabularies you should first run the vocab_parser script.
In order to run the different Networks, you need to download the weights and place them in the "Dataset/Model/NAME_model". You can find the weights at the following google drive link:

MODEL WEIGHTS