COMP34412 Natural Language Systems syllabus 2019-2020
Enabling computers to use 'natural language' (the kind of language that people use to communicate with one another) is becoming more and more important. It allows people to communicate with them without having to use strange artificial languages and awkward devices like keyboards and mice. It also allows the computer to access the enormous amount of material that is stored as natural language text on the web or in document repositories.
This course provides an introduction to the area of natural language processing (NLP), mixing theory (if you don't understand the theory of how language works you cannot possibly write programs that understand it) with practice (if you haven't written or played with tools that embody the theory, you can't get a concrete handle on what the theory means).
The course unit aims to teach the techniques required to extend the theoretical principles of NLP to applications in a number of critical areas.
- To demonstrate how the essential components of practical NLP systems are built and modified.
- To introduce the principal applications of NLP, including information retrieval & extraction, spoken language access to software services, and machine translation.
- To explain the major challenges in processing large-scale, real-world natural language.
- To give students an understanding of the issues involved in evaluating NLP systems.
· Introduction, motivation, review of NLP principles
· Essential steps for NLP algorithms
o Part-of-speech tagging: probabilistic tagging, transformation-based learning
o Parsing: chunking, shallow parsing, statistical parsing
o Lexical semantics: lexical resources, word sense disambiguation algorithms
· Evaluation of natural language systems
o Crowdsourcing and inter-annotator agreement
· Information retrieval and extraction
o Document matching
o Named-entity recognition
o Template-filling, free text question answering systems
o Summarisation algorithms
· Natural language generation
o Surface realisation
o Discourse planning
· Machine translation
o Transfer-based approaches: the MT pyramid, transfer rules
o Statistical MT, memory-based MT
· Introduction to spoken language systems
o The nature of speech
o Speech synthesis
o Speech recognition
Lectures with some practical workshops and guest lecturers
11 x 2 hours
There is unassessed formative homework given to the students weekly; each lecture then starts with a 10 min feedback session where the students may review the homework from the previous week. Assessed coursework is due in weeks 6 and 11, with feedback in weeks 7 and 12 respectively.
- Lectures (22 hours)
- Analytical skills
- Problem solving
On successful completion of this unit, a student will be able to:
- describe the various standard levels of linguistic description (acoustic phonetics, phonology, morphology, syntax, semantics, pragmatics)
- describe the interfaces between the stages of a typical NLP pipeline and show the information that is passed from one to another for previously unseen examples
- discuss the use of underspecified representations to prevent errors propagating between stages
- outline a number of algorithms for carrying out the tasks at each level, illustrating their operation on previously unseen concrete examples, and discuss their limitations
- apply a range of numerical evaluation methods to NLP systems and discuss the issues involved in constructing `Gold Standard' test sets
No reading list found for COMP34412.
Course unit materials
Links to course unit teaching materials can be found on the School of Computer Science website for current students.