This module introduces students to the field of Natural Language Understanding and related topics including sentiment analysis, relation extraction, natural language inference, semantic parsing, question answering, language generation, and conversational agents. This module builds on the advanced NLP experience of students.

This module builds on users' experience of NLP and introduces the concepts on Natural Language Understanding through the 6 topics outlined below. Each topic 1. Supervised Sentiment Analysis: such as Conceptual challenges (Affective dimensions, relations, and transitions), Sentiment datasets, Sentiment Lexica, Sentiment-aware Tokenizing, Dangers of stemming and POS tagging, negation marking, Stanford Sentiment Treebank, Feature representation and selection, RNN classifiers, TreeNNs. 2. Relation Extraction: including Hand-built patterns, Supervised learning, Distant supervision. 3. Natural Language Inference: such as NLI task formulation, NLI datasets (e.g., SemEval, GLUE, SNLI, MultiNLI), Hand-built features, Sentence-encoding models, Chained models, Attention (global, local, word-by-word). 4. Grounded language understanding and Semantic Parsing. 5. Question Answering: challenges of machine reading comprehension and conversational QA, LCC's QA system, Stanford Question Answering Dataset (SQuAD), Stanford Attentive Reader, Bi-Directional Attention Flow for Machine Comprehension (BiDAF). 6. Conversational Agents: Automatic speech recognition (ASR), Language Understanding (NER, intent detection, slot filling), Dialogue Management (dialogue state tracking, dialogue policy), Natural Language Generation, Text-to-Speech (TTS).

On successful completion of this module, students will be able to: 1. Implement a supervised sentiment analysis system and evaluate its performance using the Stanford sentiment treebank. 2. Implement a relation extraction system using the distant supervision approach. 3. Implement an RNN-based NLI model and evaluate its performance on the GLUE benchmark. 4. Implement a simple question answering system and evaluate its performance on the SQuAD dataset. 5. Build dialog agents with open-source frameworks such as Rasa, and cloud-based platforms such as Google Dialogflow and Amazon's Alexa.

On successful completion of this module, students will be able to: 1. Recognize and be mindful of ethical issues in modern NLP systems such as algorithmic bias (gender, race) and disinformation (fake news).

On successful completion of this module, students will be able to:

The module will be delivered using fully online using on-line lectures, labs and tutorials