Official MLMI13 Coursework Code

Handout Link: link to moodle
Contact Details: Dr. Kate Knill, Adian Liusie (kmk1001@cam.ac.uk, al826@cam.ac.uk)

Overview

This repository will provide the code framework for the MLMI coursework. The code base has been partialy completed where code sections marked with #TODO are left for the student to compelte. The task considered for the coursework is sentiment classifcation on the IMDB dataset, where the methods proposed in the original paper will be considered.

The code base has a very modular set up

• practical.py is the main python file to run all the experiments (an identical ipython version is also provided)
• Corpora.py is a class that reads and processes the IMDB data.
• Evaluation.py is a class which many other classes inherits for automatic evaluation capabilities.
• There are several other files to be completed to assess different NLP techniques on IMDB sentiment classification

This repository contains all the data needed for the first section of the handout:

• The IMDB reviews (1000 positive and 1000 negative) are saved in data/reviews (.zip and .tar.gz versions of the data are provided which have to first be extracted).
• The Lexicon sentiment file is saved in data/sent_lexicon

Requirements

This Coursework is based in python3 and is compatabile with any operating system. The required packages are:

• scipy
• scikit
• gensim
• nltk

Which are easily available through pip

Acknowledgements

This coursework is based on the MLSALT13 practical developed by Milica Gašić and Kevin Heffernan. The python code closely follows the previous framework but has been updated to interface better with python3.