Lexikon, Syntax, Semantik II

Lexikon, Syntax, Semantik II




Please, fill in the form on the following link in order to register for the course:
http://diener.cis.uni-muenchen.de/max/nachrichten/lsf/nachmelden/message_index.html

Abstract   Semantic information is one of the building blocks for most NLP tasks. Adequate representation of semantic relations or the identification of the most appropriate meaning of words in discourse often requires extensive common sense and universal world knowledge. However, such information is not easy to extract and convey in useful data models.
This course will cover various levels of semantic relatedness between words as well as ways to represent or calculate such information. We will introduce multiple important natural language processing tasks that employ semantic knowledge and will discuss the state-of-the-art approaches to them.
Format of the course  
Instructors: Hinrich Schütze and Desislava Zhekova  
Syllabus  
#DateTopicSlides
111 April Introductory meeting (Zhekova)PDF
225 April Lexical Databases and Ontological Information for Word Sense Disambiguation(WSD), WSD (Zhekova)PDF
32 May cancelled
49 May Evaluation of WSD; Approaches to WSD (Zhekova)PDF
516 May cancelled
623 May Recent Advances in (Deep) Representation Learning (Part I) (Schütze)PDF
830 May Recent Advances in (Deep) Representation Learning (Part II) (Schütze)
96 June Semantic Similarity and Ontologies (Zhekova)PDF, Michaela's presentation
1013 June Words, Word Senses, Semantic Similarity, Word Space Models (Part I) (Schütze)
1120 June Words, Word Senses, Semantic Similarity, Word Space Models (Part II) (Schütze)
Project Abstracts due
1227 June Project Work
1304 July Project Work
1411 July Project Presentations
Group Work   The approximate number of participants in a group is 2-3. Please, make sure that all participants have an equal share of the work (this you will need to indicate clearly in your project documentation and presentation). In case you wish to work alone or in a group with more than 3 participants, please, discuss this with me beforehand. The size of the group may be varied based on the complexity and extent of the project.
Project Work   The last three sessions of the class will be devoted to your projects. Within that time you will need to complete the tasks you selected. You are free to choose the project topics yourself (the topic for each group needs to be approved before you submit the abstract) or select from a list of suggestions below:
  1. Term co-occurrence extraction (group size: 1-2) - This project should aim to provide a possibility for automatic extraction of term co-occurrence information. Various search spaces may be used for this purpose (e.g. corpora, online resources, the WWW).
  2. Definition extraction (group size: 1-2) - Extracting definitions from machine readable text (again the use of corpora, online resources or interactive web searches will be allowed).
  3. Ontology creation from extracted definitions (group size: 1-3) - This project should provide the possibility to construct an ontology given a predefined set of definitions. The extraction of definitions may be included in this task depending on the size of the group.
  4. A basic WSD pipeline (group size: 2-3) - Given the datasets and software provided from the Senseval 3 task (http://www.senseval.org) a WSD pipeline should be created.
  5. Mapping senses between machine readable dictionaries (group size: 2-3) - This project should provide the possibility for mapping various machine readable dictionaries independent of the granularity of senses they provide.
  6. Calculating semantic similarity on a cross-resource level (group size: 2-3) - For this project a semantic similarity score should be calculated via information extracted from more than one resource (e.g. WordNet, calculated term co-occurrence, etc.)
Project Abstract   After you have selected a topic, you will be required to provide a written abstract (1 page) of the problem at hand and how you plan to tackle it. The abstract should demonstrate a good understanding of the task and a proposed project work plan.
Project Presentations   The last session of the course will be devoted to the presentation and discussion of your project work. Each team will have the possibility to introduce the targeted task and the used approach to solve the problem. This session should mostly focus on a discussion of the problems you encountered during the implementation and on a proper evaluation of your results.
Project Paper   As part of your project work you will be required to write a paper (of at least 12 pages) that would both present and explain the linguistic background and NLP application of the topic you selected as well as provide an overview of your project work and discussion of the problems you encountered on the way. The paper should also aim to include a proper evaluation of the approach you used to tackle the task. Another possibility for the format of the paper is a scientific/conference paper (the length of this variant depends on the problem you have worked on and is within the limits of 4-8 pages). We will encourage you for the latter and will strongly promote the submission of such papers to international student workshops.
Assessment   The final assessment of the class is based on the following aspects:


File translated from TEX by TTH, version 4.03.
Last edited: 9 Apr 2014, 13:33.