Second Year Project (Spring 2019)

Abstract

After completing this course, students are able to tackle a challenge in Natural Language Processing, implement a solution, and critically assess their solution in the light of robustness and state-of-the art.

Description

An introduction to major topics in Natural Language Processing and relevant related topics (traditional and neural-network based methods to, e.g., language modeling, classification, sequence processing).
An introduction to deep neural networks for Natural Language Processing, including representation learning, and an implementation in a corresponding Python-based framework.

In this project-based course, students are going to work on a real-world problem using natural language processing technology. The focus of the theoretical part of this class is on natural language processing and deep learning. Social implications of data handling will be discussed as well.

The course has two parts: a series of lectures and exercises with a possibility for guest lecturers followed by project-based work in groups.

Amongst others, the following topics will be covered:

• Introduction to NLP and DL, what makes language so difficult; traditional versus neural approaches
• Data Science at the command line; obtaining and preprocessing data
• Language Models, Neural LMs
• Classification for NLP, Embeddings
• Sequence predictions for NLP, RNNs and LSTMs
• Ethical considerations of machine learning

Formal prerequisites

Intended learning outcomes

After the course, the student should be able to:

• Discuss, clearly explain, and reflect upon central concepts, algorithms, and challenges in natural language processing (NLP) and deep learning (DL).
• Organize, plan, and carry out collaborative work in a smaller project group.
• Obtain, scrub, explore and preprocess a wide range of relevant raw data for a given problem. Identify and analyze the relevant options for data collection and preprocessing and select the most suitable ones.
• Design and implement a sound experiment in NLP
• Distinguish and evaluate the advantages of different design choices or approaches to the same task (e.g., traditional versus deep-learning based solutions)
• Evaluate the achieved solution and carry out a detailed error analysis, relating the findings back to the overall problem domain
• Explain in writing (project group report) adhering to academic standards in writing
• Succinctly present the results of the project, discuss findings and limitations
• Understand and reflect upon ethical considerations that arise in the deployment of language technology