📺 Fact Checking through Neural Language Inference 💡

This repository contains the second assignment part of the Natural Language Processing course of the Master's degree in Artificial Intelligence, University of Bologna.

Project Details

The main objective of such assignment is to solve a Fact Checking taks by using Neural Language Inference (NLI) on a pre-processed version of the FEVER dataset using pre-trained GloVe embeddings.

We provided four different configurations and their implementation to test if any sentence embedding technique is sensibly better than another:

1. a RNN model where we take the last state obtained,
2. a RNN model where we take the mean of the states obtained,
3. a MLP model,
4. a Bag of Vectors (BoV) model.

For more details on the pipeline followed please read the report present in the report fact_checking_nli.pdf.

Technologies and Frameworks

Frameworks:

• Tensorflow (v2.3.0)

Platforms

• Google Colaboratory

Configurations and enviroments

The environment could be loaded by using conda by launching the command:

$ conda create --name <env> --file requirements.txt

Versioning

We used Git for versioning.

License

This project is licensed under the MIT License - see the LICENSE file for details.