Identifying Semantic Argument Types in Predication and Copredication Contexts: A Zero-Shot Cross-Lingual Approach
20113Authors
nan
Abstract
AbstractIdentifying semantic argument types in predication contexts is not a straightforward task for several reasons, such as inherent polysemy, coercion, and copredication phenomena. In this paper, we train monolingual and multilingual classifiers with a zero-shot cross-lingual approach to identify semantic argument types in predications using pre-trained language models as feature extractors. We train classifiers for different semantic argument types and for both verbal and adjectival predications. Furthermore, we propose a method to detect copredication using these classifiers through identifying the argument semantic type targeted in different predications over the same noun in a sentence. We evaluate the performance of the method on copredication test data with Food•Event nouns for 5 languages.
Keywords
Full Content
Proceedings of Recent Advances in Natural Language Processing, pages 310–320 Varna, Sep 4–6, 2023 https://doi.org/10.26615/978-954-452-092-2_035 310 Identifying Semantic Argument Types in Predication and Copredication Contexts: A Zero-Shot Cross-Lingual Approach Deniz Ekin Yavas1, Laura Kallmeyer1, Rainer Osswald1, Elisabetta Jezek2, Marta Ricchiardi3, Long Chen1 Heinrich Heine University D¨usseldorf1, University of Pavia2,3 {deniz.yavas, laura.kallmeyer, rainer.osswald, chen.long}@hhu.de1, jezek@unipv.it2, marta.ricchiardi01@universitadipavia.it3 Abstract Identifying semantic argument types in pred- ication contexts is not a straightforward task for several reasons, such as inherent polysemy, coercion, and copredication phenomena. In this paper, we train monolingual and multilin- gual classifiers with a zero-shot cross-lingual approach to identify semantic argument types in predications using pre-trained language mod- els as feature extractors. We train classifiers for different semantic argument types and for both verbal and adjectival predications. Further- more, we propose a method to detect copredica- tion using these classifiers through identifying the argument semantic type targeted in different predications over the same noun in a sentence. We evaluate the performance of the method on copredication test data with Food•Event nouns for 5 languages. 1 Introduction This paper is concerned with the question of how to automatically decide which semantic type is tar- geted in predications over nouns. In our case, the predicate can be a verb or an adjective. This ques- tion is particularly interesting in cases where com- plex type nouns1 are arguments of predications. But even with nouns that, lexically, have only a single type, the predication can target a different type and thereby trigger a coercion in the noun (Pustejovsky, 1991). Examples are given in (1). In both (1-a) as well as (1-b), the respective predi- cates target one of the two types of a complex type noun (a dinner is inherently both an Event and a Food item). In (1-c), the noun is a simple type noun (soup is only of type Food), and its type is targeted in the predication. The predication in (1-d) involves a coercion since it targets a type that is 1Also “dot object” nouns (Pustejovsky, 1995), “nouns with facets” (Cruse, 1995), “dual aspect nouns” (Asher, 2011) in the literature. different from the lexical type of the noun. Finally, for complex type nouns, we can have cases where different component types of the same noun are targeted, either by different predicates as in (1-e) or by a single predicate as in (1-f) where book is a physical object and an informational content at the same time and the predicate targets both. The first case is an instance of copredication (see below). (1) a. They chose the vegetarian dinner. → (target: Food) b. I organized a dinner for them. → (target: Event) c. I ate my soup. → (target: Food) d. I finished my soup. → (target: Event, coercion) e. They organized a vegetarian dinner. → (target: Event and Food) f. He wrote a lot of books. → (target: Phys Obj•Information) Our main goal is to develop classifiers that, given a predicate and an argument noun in their sentential context, decide whether a specific …
Content truncated to 500 words
Labeled Text
AbstractIdentifying semantic argument types in predication contexts is not a straightforward task for several reasons, such as inherent polysemy, coercion, and copredication phenomena. In this paper, we train monolingual and multilingual classifiers with a zero-shot cross-lingual approach to identify semantic argument types in predications using pre-trained language models as feature extractors. We train classifiers for different semantic argument types and for both verbal and adjectival predications. Furthermore, we propose a method to detect copredication using these classifiers through identifying the argument semantic type targeted in different predications over the same noun in a sentence. We evaluate the performance of the method on copredication test data with Food•Event nouns for 5 languages.
Publication Details
- Paper ID:
- 20113
- Year:
- 2023
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan