Corpus-based Identification of Verbs Participating in Verb Alternations Using Classification and Manual Annotation

4007

Authors

nan

Abstract

AbstractEnglish verb alternations allow participating verbs to appear in a set of syntactically different constructions whose associated semantic frames are systematically related. We use ENCOW and VerbNet data to train classifiers to predict the instrument subject alternation and the causative-inchoative alternation, relying on count-based and vector-based features as well as perplexity-based language model features, which are intended to reflect each alternation’s felicity by simulating it. Beyond the prediction task, we use the classifier results as a source for a manual annotation step in order to identify new, unseen instances of each alternation. This is possible because existing alternation datasets contain positive, but no negative instances and are not comprehensive. Over several sequences of classification-annotation steps, we iteratively extend our sets of alternating verbs. Our hybrid approach to the identification of new alternating verbs reduces the required annotation effort by only presenting annotators with the highest-scoring candidates from the previous classification. Due to the success of semi-supervised and unsupervised features, our approach can easily be transferred to further alternations.

Keywords

nan

Full Content

Proceedings of the 28th International Conference on Computational Linguistics, pages 4044–4055 Barcelona, Spain (Online), December 8-13, 2020 4044 Corpus-based Identification of Verbs Participating in Verb Alternations Using Classification and Manual Annotation Esther Seyffarth Heinrich Heine University D¨usseldorf D¨usseldorf, Germany seyffarth@phil.hhu.de Laura Kallmeyer Heinrich Heine University D¨usseldorf D¨usseldorf, Germany kallmeyer@phil.hhu.de Abstract English verb alternations allow participating verbs to appear in a set of syntactically different constructions whose associated semantic frames are systematically related. We use ENCOW and VerbNet data to train classifiers to predict the instrument subject alternation and the causative- inchoative alternation, relying on count-based and vector-based features as well as perplexity- based language model features, which are intended to reflect each alternation’s felicity by sim- ulating it. Beyond the prediction task, we use the classifier results as a source for a manual annotation step in order to identify new, unseen instances of each alternation. This is possi- ble because existing alternation datasets contain positive, but no negative instances and are not comprehensive. Over several sequences of classification-annotation steps, we iteratively extend our sets of alternating verbs. Our hybrid approach to the identification of new alternating verbs reduces the required annotation effort by only presenting annotators with the highest-scoring can- didates from the previous classification. Due to the success of semi-supervised and unsupervised features, our approach can easily be transferred to further alternations. 1 Introduction Verb alternations are a phenomenon at the syntax-semantics interface in which several syntactic patterns are available for certain verbs, assigning different semantic roles to the elements in individual slots of the constructions. English verbs that participate in the instrument subject alternation (ISA, Levin (1993)) can take an optional instrument argument, encoded as a with-PP or using-phrase, which can also be realized in subject position (see (1)). Replacing with with using is also a standard test for instrumenthood, i.e., a means to exclude other cases of with-PPs. Not all verbs with an (optional) instrument participate in ISA (see (2)). This is usually explained by the specific nature of the instrument in the context of the verb (Van Hooste, 2018). The English causative-inchoative alternation (CIA) involves two main constructions: Verbs can appear intransitively, as in (3-a), or transitively, as in (3-b). The syntactically intransitive construction is associated with an inchoative sense, while the transitive construction has a causative sense. (1) a. She broke the glass with/using a hammer. b. The hammer broke the glass. (2) a. She ate the soup with/using a spoon. b. *The spoon ate the soup. (3) a. The door opens. b. I open the door. Verb alternations pose a challenge to tasks like semantic role labeling or frame induction, and knowl- edge about which verbs participate in which alternations is highly useful for such tasks: Certain syntactic patterns need to be interpreted differently, depending on whether or not the verb participates in the al- ternation. For instance, the syntactic subject in a sentence with the syntactic structure of (3-a) would This work is licensed under a Creative Commons Attribution 4.0 International License. License …

Content truncated to 500 words

Labeled Text

AbstractEnglish verb alternations allow participating verbs to appear in a set of syntactically different constructions whose associated semantic frames are systematically related. We use ENCOW and VerbNet data to train classifiers to predict the instrument subject alternation and the causative-inchoative alternation, relying on count-based and vector-based features as well as perplexity-based language model features, which are intended to reflect each alternation’s felicity by simulating it. Beyond the prediction task, we use the classifier results as a source for a manual annotation step in order to identify new, unseen instances of each alternation. This is possible because existing alternation datasets contain positive, but no negative instances and are not comprehensive. Over several sequences of classification-annotation steps, we iteratively extend our sets of alternating verbs. Our hybrid approach to the identification of new alternating verbs reduces the required annotation effort by only presenting annotators with the highest-scoring candidates from the previous classification. Due to the success of semi-supervised and unsupervised features, our approach can easily be transferred to further alternations.

Publication Details

Paper ID:
4007
Year:
2020
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan