From Information Extraction to 'Deep' Semantic Parsing
Kursbeschreibung
Studiengang | Modulkürzel | Leistungs- bewertung |
---|---|---|
BA-2010 | AS-CL | 8 LP |
NBA | AS-CL | 8 LP |
Master | SS-CL, SS-TAC | 8 LP |
Magister | - | - |
Dozenten/-innen | Anette Frank |
Veranstaltungsart | Hauptseminar |
Erster Termin | 24.10.2013 |
Zeit und Ort | Do, 11:15–12:45, INF 327 / SR 5 (SR) |
Commitment-Frist | 02.12.2013 – 17.01.2014 |
Leistungsnachweis
- regelmäßige und aktive Teilnahme
- Referat
- Hausarbeit oder Projekt
Inhalt
Open Information Extraction (OpenIE), Never Ending Language Learning (NELL), Machine Reading, ... - these terms describe recent developments in web-scale information extraction, knowledge acquisition and relation learning in the era of 'big data'.
This trend goes hand in hand with the evolving 'Web of Linked Data'. Starting from automatically harvested knowledge bases such as DBpedia, large-scale IE techniques are used to induce novel ontological resources that link knowledge about entities and relations to natural language expressions (see, e.g., YAGO, PATTY).
We will start with reviewing techniques and limitations of current research in large-scale IE approaches. We will study in particular techniques that aim at the automatic induction of (ontological) resources linking facts to language, and how to apply the acquired knowledge for enhanced 'deep' semantic parsing.
We will in particular focus on aspects such as:
- relation and sense clustering
- linking relations to syntactic configurations
- induction of semantic types
On the basis of this review, we will discuss how to further exploit this novel paradigm for the automated induction of linguistic knowledge for enhanced NLP using data sets such as Google syntactic n-grams. Here, the seminar participants will be encouraged to develop own ideas and scenarios on how to link knowledge to language for enhanced NLP.
Kursübersicht
Seminarplan
Datum | Sitzung | Papiere und Referent/in |
24.10. | Einführung | Anette Frank Slides |
31.10. moved to 7.11. |
State of the Art OpenIE | Fader et
al. 2011 Etzioni et al. 2011 (opt/bgr:) Banko et al. 2007 Elefterios Matios |
7.11. | Analysis of State of the Art: Shallow vs. Deep; Limitations of web-scale OpenIE |
Mesquita et
al. 2013 Weikum et al. 2012 Anette Frank |
14.11. | Distant Supervision | Mintz et
al. 2009 Zhang et al. 2013 Anette Frank |
14.11. 13:30 SR 2, INF 327 |
Semantic Methods in Relation Extraction | Yao et
al. 2012 Chan & Roth 2011 alt: Plank & Moschitti 2013 Ofer Bronstein |
28.11. | More Semantic Methods in Relation Extraction | Lao et
al. 2012 Nakashole et al. 2012 Mirjam Eppinger |
5.12. | From Relations to Events and Parsing | McClosky et
al. 2011 (opt.) Singh et al. Goldberg & Orwant 2012 Tri Duc Nghiem |
12.12. | Events in Context | Gerber & Chai 2010/2012 (opt.) Silberer & Frank 2012 Chambers & Jurafsky 2009 Thomas Haider |
19.12. | Unsupervised Event Schema/Frame Induction | Cheung et
al. 2013 Chambers 2013 Joachim Bingel |
9.01. | Event Coreference Resolution | Bejan &
Harabagiu 2010 Lee et al. 2012 group discussion |
16.01. | Cross-document Event Alignment and Search | Roth & Frank
2012 Glavaš & Šnajder 2013 Chen Li |
23.01. | Applications I: Semantic parsing for instruction-giving | Zettlemoyer
& Collins
2007 Artzi & Zettlemoyer 2013 Mareike Hartmann |
30.01. | Applications II: Multi-Document Summarization | Chi et
al. 2013 Zoe Bylinovich |
30.01. 13:30 SR7 INF 325 |
Wrap-up / Discussion |
Literatur
- Nakashole, Weikum and Suchanek (2012). PATTY: A Taxonomy of Relational Patterns with Semantic Types . EMNLP 2012.
- Lao, Subramanya, Pereira and Cohen (2012). Reading The Web with Learned Syntactic-Semantic Inference Rules. EMNLP 2012.
- Weikum et al. (2012). Big Data Methods for Computational Linguistics. IEEE.
- Goldberg and Orwant (2013). A Dataset of Syntactic-Ngrams over Time from a Very Large Corpus of English Books . *SEM 2013
More literature will be provided with the beginning of the seminar.