peiyong-addwater commented 2 years ago

Abstract

Classic beginner's task for machine learning/natural language processing, but Quantum!

If you are doing NLP research or would like to dive into the world of NLP, this is your chance to experience something totally new.

do you guys just put quantum in front of everything

Description

Maybe you have done this multiple times, but I promise you, this time it will be different!

We will need to find a way to load the Twitter texts onto the quantum computer/simulator, and then develop a (simple) model to classify the sentiments expressed in the tweets. We'll be using a small subset of the open-source dataset, like the Sentiment140 dataset (or any other opensource dataset on Kaggle), and a good example of how to preprocess the data and analyse the sentiment in the traditional way can be found here. For modelling language with quantum computers, we'll try to follow a series of techniques developed by Bob Coecke and friends, see the tutorial of the lambeq Python package here. An interesting talk about quantum natural language processing by Bob can be found here. An example of how Bob's QNLP framework process the sentence can be found at a Jupyter Notebook modified from the Udemy QNLP course material. Anonther example of converting sentence to diagrams can be seen here .

Environment/Installation

Although you can just pip install the required packages (lambeq, spacy, allennlp, pytket, ...), I prepared docker images for everything we need. If you are familiar with/want to try docker, go to https://hub.docker.com/repository/docker/addwater0315/quantum to pull the image tagged with qnlp or qnlp_gpu for GPU-enabled machines that have NVdia container toolkit installed.

Goals

Through this project, we would like to learn the techniques of quantum natural language processing together, and show you how quantum can transform other areas, in this case, linguistics and natural language processing.

Develop a basic understanding of how to handle basic NLP tasks on a quantum computer;
Familiar with the basics of the DisCoCat algorithm;
Discover new possibilities of combining current NLP research with QNLP techniques.