This course provides an overview of the area of automatic text summarisation. Summarisation is the process of shortening a text to its core information content. It is often seen as a general task to assess the level of text understanding achieved by a method, but the task is also very practically usable. The course assumes general knowledge of natural language processing. It enables students to understand research papers in the area of summarisation and to start designing their own summarisation-based research and implementation based on the information given in the course. Careful and objective evaluation of summarisation is particularly important, given the vague definition of the task. The course therefore pays special attention to evaluation. The class is a mixture of lectures, student presentations, and discussion of research methods. Class size is restricted to 10 students.
At the end of the course students should be able to
- digest and understand research papers on summarisation;
- explain main methods of summarisation and their challenges;
- understand evaluation methods for summarisation;
- build sample applications exemplifying the summarisers presented.
Specialist skills will be developed. Communication skills will be developed as well, because the course language is English and because student presentations are part of the assessment. Critical skills will also be developed, as there is no textbook, but research articles are given to students to deliver the background reading.
text summarisation, extractive summarisation, noisy channel model, integer linear programming, cohesion-based method, template-based method
✔ Specialist skills | ✔ Intercultural skills | Communication skills | Critical thinking skills | Practical and/or problem-solving skills |
Specified in the class
Course schedule | Required learning | |
---|---|---|
Class 1 | Introduction and Task Definition | Specified in the class |
Class 2 | Importance Indicators | |
Class 3 | Extractive Summarisation (Kupiec) and MMR | |
Class 4 | Summarisation by TextRank | |
Class 5 | Sentence Compression by Noisy Channel Model | |
Class 6 | Summarisation by Integer Linear Programming | |
Class 7 | Extraction-Template Based Summarisation | |
Class 8 | Cohesion-based Summarisation | |
Class 9 | Narrative Summarisation | |
Class 10 | Story Understanding and Summarisation | |
Class 11 | Memory-limitation Summarisation | |
Class 12 | Scientific Summarisation | |
Class 13 | Intrinsic and Extrinsic Summary Evaluation | |
Class 14 | Evaluation based on Meaning units | |
Class 15 | Wrap-up and Outlook |
None. Reading background (papers) will be announced ahead of each lecture on the website.
* Inderjeet Mani. Automatic Text Summarisation. 2001. John Benjamins Publishing.
* Juan-Manuel Torres-Moreno (editor). Automatic Text Summarisation. 2014. Wiley.
Note -- I will not follow these text books but use original materials (research papers).
Please don't buy these books; they are listed here just for your information.
10 % Contribution to Class discussion.
20 % Presentation.
70 % Literature-based Essay or Report on Student-Designed Summarisation Experiment
(to be discussed with lecturer beforehand)
Programming Ability
Completion of Course "Natural Language Processing" is preferable.