Presentation is loading. Please wait.

Presentation is loading. Please wait.

SALSA The Saarbrücken Lexical Semantics Annotation & Acquisition Project Aljoscha Burchardt, Katrin Erk, Anette Frank, Andrea Kowalski, Sebastian Pado,

Similar presentations


Presentation on theme: "SALSA The Saarbrücken Lexical Semantics Annotation & Acquisition Project Aljoscha Burchardt, Katrin Erk, Anette Frank, Andrea Kowalski, Sebastian Pado,"— Presentation transcript:

1 SALSA The Saarbrücken Lexical Semantics Annotation & Acquisition Project Aljoscha Burchardt, Katrin Erk, Anette Frank, Andrea Kowalski, Sebastian Pado, Manfred Pinkal

2 Semantic Annotation in SALSA Manual semantic annotation of 0.8 million words of syntactically annotated German newspaper text (TIGER Corpus, Releases 1, 2) with frames and frame elements (Berkeley FrameNet Database), staying as close as possible to the Berkeley FrameNet database

3 SALSA: What's special? SALSA is about German Cross-lingual divergencies?

4 Cross-lingual Divergencies Convincing cross-lingual portability results (E  D) in general Adaptation necessary because of Inappropriate granularity of distinctions between FEs Missing FEs (Rare cases of) inappropriate granularity of frames

5 SALSA: What's special? SALSA is about German Cross-lingual divergencies? Corpus-driven lexicon development through exhaustive full-text annotation Difficult cases Incompleteness of Berkeley FrameNet

6 Difficult cases Metaphors Support Verb Constructions Idioms

7 Difficult phenomena: Some Figures Sample of 246 LemmasSub-corpus nehmen Number% % Standard readings463885,7%4217,4% Metaphor3696,8%3815,8% Support3266,0%13254,8% Idiom791,5%2912,0% Non-literal use77414,3%19982,6% Total5412100,0%241100,0%

8 SALSA corpus: Release I Total size of 20.000 annotated instances Consistent annotation through different verification steps All occurrences/readings of > 400 German verbal predicates (different frequency bands) Scheduled for Summer 2006

9 The SALTO Annotation Tool

10 SALSA II: Automatic Annotation and Acquisition Fred, Rosy, and Shalmaneser: A tool- chain for shallow semantic analysis  Talk by Katrin and Sebastian

11

12 SALSA II: Automation Fred, Rosy, and Shalmaneser: A tool- chain for shallow semantic analysis  Talk by Katrin and Sebastian The Detour System (through WordNet to FrameNet)  Talk by Anette and Al

13 Fred & Rosy Fred, Detour & Rosy

14 SALSAII: Automation Fred, Rosy, and Shalmaneser: A tool- chain for shallow semantic analysis  Talk by Katrin and Sebastian The Detour System (through WordNet to FrameNet)  Talk by Anette and Al Cross-lingual projection of frame- semantic information  Katrin and Sebastian

15 Cross-lingual Projection

16 SALSAII: Automation & Application Fred, Rosy, and Shalmaneser: A tool-chain for shallow semantic analysis  Talk by Katrin and Sebastian The Detour System (through WordNet to FrameNet)  Talk by Anette and Al Cross-lingual projection of frame-semantic information  Katrin and Sebastian Textual Entailment (RTE)  Anette and Al

17 t: In 1983, Aki Kaurismäki directed his first full-time feature. h: Aki Kaurismäki directed a film.

18 t: In 1983, Aki Kaurismäki directed his first full-time feature. h: Aki Kaurismäki directed a film. WordNet related Grammatically related

19 SALSA: Future Work Bottstrapping frame information by data expansion techniques Linking lexical semantic resourcs with upper-model ontologies Analysis of non-compositional phenomena A worked-out semantic lexicon Application to textual entailment


Download ppt "SALSA The Saarbrücken Lexical Semantics Annotation & Acquisition Project Aljoscha Burchardt, Katrin Erk, Anette Frank, Andrea Kowalski, Sebastian Pado,"

Similar presentations


Ads by Google