Graph-based Methods for Natural Language Processing
Kursbeschreibung
| Studiengang | Modulkürzel | Leistungs- bewertung |
|---|---|---|
| BA-2010 | AS-CL | 8 LP |
| NBA | AS-CL | 8 LP |
| ABA | V01 | 6 LP |
| Master | SS-CL, SS-TAC | 8 LP |
| Magister | - | - |
| Dozenten/-innen | Simone Paolo Ponzetto |
| Veranstaltungsart | Hauptseminar |
| Veranstaltungsbeginn | 21.10.2010 |
| Zeit und Ort | Do, 13:30–15:45, INF 325 / SR 24 (SR) |
Teilnahmevoraussetzungen
Voraussetzungen sind die bestandene Zwischenprüfung (Magister) und Programmierprüfung. Vorkenntnisse in statistischer NLP oder Maschinellem Lernen sind von Vorteil.
Leistungsnachweis
Aktive Teilnahme und regelmäßige Abgabe von Projektenarbeit in kleinen Gruppen. Vortrag/Präsentation.
Zusammensetzung der Endnote:
- 1/3: Präsentation
- 1/3: Beteiligung an den Seminarprojekten
- 1/3: Beteiligung an den Diskussionen im Seminar
Inhalt
In the last years Natural Language Processing (NLP) researchers have shown a considerable amount of interest in developing methods based on graph theoretic models, with a large variety of NLP applications adopting efficient and elegant solutions from graph-theoretical frameworks.
This seminar will provide a gentle introduction to state-of-the-art graph-based methods for NLP applications. These include, but are not limited to:
- Word sense disambiguation
- Information extraction
- Automatic summarization
- Co-reference resolution
- Named entity recognition and disambiguation
The course will be offered as a project seminar. Students will present current work from the literature in short, seminar-format presentations (i.e., Referate). In addition, they will be expected to form small groups of 2-3 people and work on a project, e.g. implement and/or extend an existing state-of-the-art graph-based NLP method. Each one of the groups is expected to submit a short report (2-4 pages), as well as to regularly give an update on the status of their project -- i.e. as a very short, informal presentation on a regular basis. Students are expected to *actively* participate in the class discussions during their fellow students' presentations, as well as in the seminar's projects. This means that you'll have to read the papers *before* the class period in which they will be presented and discussed, as well as *clearly* present to the audience what your specific work was as part of the seminar's projects.
Kursübersicht
Seminarplan
| Datum | Sitzung | Materialien | 21.10 | Organisatorisches, Einfuehrung I | Folien | 28.10 | Themenvergabe, Einfuehrung II | Folien | 4.11 | Katharina Waeschle WikiWalk: Random walks on Wikipedia for Semantic Relatedness Eleftherios Matios Pageranking WordNet synsets: An application to opinion mining |
Folien Folien |
11.11 | Cai Jie Learning from Labeled and Unlabeled Data using Graph Mincuts Andreas Doerr Graph-Based Generation of Referring Expressions |
18.11 | Samuel Broscheit Experiments in Graph-based Semi-Supervised Learning Methods for Class-Instance Acquisition Eva Mujdricza-Maydt Chinese Whispers - an Efficient- Graph Clustering Algorithm and its Application to Natural Language Processing Problems |
Folien Folien |
25.11 | Xiaoxi Pang The PageRank Citation Ranking: Bringing Order to the Web Thierry Goeckel A sentimental education: Sentiment analysis using subjectivity summarization based on minimum cuts |
Folien Folien |
2.12 | Faellt aus | 9.12 | Amin Kiem PageRank on Semantic Networks, with application to Word Sense Disambiguation Huiqin Qu Coreference resolution (TBD) |
Folien Folien |
16.12 | Wrap-up |
Literatur
We will mostly read and meditate on papers from NLP conferences such as ACL, NAACL, EMNLP, COLING etc. and related workshops (i.e. the TextGraphs workshop series). Students are encouraged to select and read 1-2 papers from this list (DOC/PDF) before the beginning of the course, in order to have a taste of its content.
