Refine
Document Type
- Part of a Book (3)
- Conference Proceeding (3)
- Article (2)
Has Fulltext
- yes (8)
Keywords
- Einbettung <Linguistik> (8) (remove)
Publicationstate
- Zweitveröffentlichung (4)
- Postprint (2)
- Veröffentlichungsversion (2)
Reviewstate
- Peer-Review (5)
- (Verlags)-Lektorat (2)
We present recognizers for four very different types of speech, thought and writing representation (STWR) for German texts. The implementation is based on deep learning with two different customized contextual embeddings, namely FLAIR embeddings and BERT embeddings. This paper gives an evaluation of our recognizers with a particular focus on the differences in performance we observed between those two embeddings. FLAIR performed best for direct STWR (F1=0.85), BERT for indirect (F1=0.76) and free indirect (F1=0.59) STWR. For reported STWR, the comparison was inconclusive, but BERT gave the best average results and best individual model (F1=0.60). Our best recognizers, our customized language embeddings and most of our test and training data are freely available and can be found via www.redewiedergabe.de or at github.com/redewiedergabe.
Die Polysemie satzeinbettender Prädikate spielt eine wichtige Rolle fur deren Einbettungsverhalten. Konkret wird gezeigt, dass Polysemie mit struktureller Ambiguität als Kontroll- vs. Anhebungsverb assoziiert sein kann (Beispiel drohen/versprechen) und dass NEG-Raising auf bestimmte Lesarten eines polysemen Verbs beschränkt sein kann. Des Weiteren wird beleuchtet, welche Faktoren die syntaktische Flexibilität satzeinbettender Prädikate, d.h. das Einbettungspotenzial bzgl. der wichtigsten Satzkomplementtypen des Deutschen, begünstigen und welche Rolle dabei Umdeutungen (z.B. von bedauern zu 'mit Bedauern äußern'), die Polysemie induzieren, spielen. Alle betrachteten Phänomene deuten darauf hin, dass sie S-Selektion (semantische Selektion) eine zentrale Rolle in der Satzeinbettung spielt.
The paper discusses particular logical consistency conditions satisfied by German proposition-embedding predicates which determine the question type (external and internal whether-form as well as exhaustive and non-exhaustive wh-form), the correlate type (es- or da-correlate) as well as the impact of the correlate on the respective consistency condition. It will turn out that some consistency conditions also determine the embedding of verb second and subject-control.
Previous accounts addressing the question what semantic properties of a matrix predicate determine the possible clause type of the embedded clause have not provided a general answer (e.g. Grimshaw 1979, Zifonun et al. 1997, Ginzburg & Sag 2000). This paper proposes that clause-embedding predicates fulfill characteristic logical conditions, so-called consistency conditions, which rule the syntactic potential of the matrix clause: for instance, the clause type of the embedded clause (declarative, ob- and/or wh-interrogative) and the correlate type, the matrix predicate can co-occur with (es and/or ProPP). Furthermore, they predict the logical forms of legitimate constructions with embedded ob- or wh-interrogatives, respectively, and how a legitimate optional correlate modifies the meaning of the matrix predicate.
Previous accounts addressing the question what semantic properties of a matrix predicate determine the possible clause type of the embedded clause have not provided a general answer (e.g. Grimshaw 1979, Zifonun et al. 1997, Ginzburg & Sag 2000). This paper proposes that clause-embedding predicates fulfill characteristic logical conditions, so-called consistency conditions, which rule the syntactic potential of the matrix clause: for instance, the clause type of the embedded clause (declarative, ob- and/or wh-interrogative) and the correlate type, the matrix predicate can co-occur with (es and/or ProPP). Furthermore, they predict the logical forms of legitimate constructions with embedded ob- or wh-interrogatives, respectively, and how a legitimate optional correlate modifies the meaning of the matrix predicate.
German subjectively veridical sicher sein ‘be certain’ can embed ob-clauses in negative contexts, while subjectively veridical glauben ‘believe’ and nonveridical möglich sein ‘be possible’ cannot. The Logical Form of F isn’t certain if M is in Rome is regarded as the negated disjunction of two sentences ¬(cf σ ∨ cf ¬σ) or ¬cf σ ∧ ¬cf ¬σ. Be certain can have this LF because ¬cf σ and ¬cf ¬σ are compatible and nonveridical. Believe excludes this LF because ¬bf σ and ¬bf ¬σ are incompatible in a question-under-discussion context. It follows from this incompatibility and from the incompatibility of bf σ and bf ¬σ that bf ¬σ and ¬bf σ are equivalent. Therefore believe cannot be nonveridical. Be possible doesn’t allow the LF either. Similar to believe, ¬pf σ and ¬pf ¬σ are incompatible. But unlike believe, pf σ and pf ¬σ are compatible.