Details

Slides and code are available on GitHub [here]. Full details about the course are available in the syllabus [here]. Class schedule is at the bottom of the page.

Bulletin Description

 This will be an advanced course on computer-based processing of communication between people (languages) with a focus on text (NLP), and appreciation of multi-modal communication involving speech and images. The course will have lessons on parsing of input, representation of syntax and semantics, and analysis to derive insights. It will explain learning and reasoning based methods for extracting entities, disambiguating and linking them, and applying them for real-world problems. We will also discuss issues in building advanced systems for conversation (“chatbots”) and machine translation, and ethical concerns while testing and fielding them with people. 

Prerequisites

Experience with a first course on Artificial Intelligence is desirable, example: CSCE 580. Students without this course may still enroll knowing that necessary AI background can be found in the standard AI book, “AI – A Modern Approach”; http://aima.cs.berkeley.edu/. Necessary material is in chapters 3,4,5 and 6.

 Learning outcomes

•  L1: Appreciate diversity and similarity in natural languages – text, speech and visual; focus of course will, however, be text (NLP) and English

• L2: Understand issues related to data and tools. Experiment design, Metrics for evaluation and to detect bias, Methods to build trust in processing – transparent assessment, Providing explanations for output

• L3: Data processing: (a) Structured data representation from unstructured text; (b) Extract entities and relationships; (c) Extract contexts; (d) representation learning – word embedding

• L4: AI methods in NLP: (a) Learning methods – including language models, (b) Reasoning, (c) Representation – knowledge graphs/ ontology

• L5: NLP applications – (a) Document intelligence: sentiment, translation; (b) collaborative assistants

Grading Schema

1.  Project: 50% for project report and code, plus 10% for elevator presentation to class

2. Quiz: 20%  (3 out of 4)

3. Exam: Paper reading and presentation to class: 20%