“The Rodney Dangerfield of Stylistic Devices”: End-to-End Detection and Extraction of Vossian Antonomasia Using Neural Networks

Abstract

Vossian Antonomasia (VA) is a well-known stylistic device based on attributing a certain property to a person by relating them to another person who is famous for this property. Although the morphological and semantic characteristics of this phenomenon have long been the subject of linguistic research, little is known about its distribution. In this paper, we describe end-to-end approaches for detecting and extracting VA expressions from large news corpora in order to study VA more broadly. We present two types of approaches: binary sentence classifiers that detect whether or not a sentence contains a VA expression, and sequence tagging of all parts of a VA on the word level, enabling their extraction. All models are based on neural networks and outperform previous approaches, best results are obtained with a fine-tuned BERT model. Furthermore, we study the impact of training data size and class imbalance by adding negative (and possibly noisy) instances to the training data. We also evaluate the models' performance on out-of-corpus and real-world data and analyze the ability of the sequence tagging model to generalize in terms of new entity types and syntactic patterns.

“Who is the Madonna of Italian-American Literature?”: Target Entity Extraction and Analysis of Vossian Antonomasia

“…While Jäschke et al (2017); Fischer and Jäschke (2019) used semi-automated approaches to detect VA expressions, Schwab et al (2019) developed the first automated approach for the detection of VA expressions on the sentencelevel. They developed a finer extraction approach on the word-level (Schwab et al, 2022). In particular, they employed pre-trained contextual language models, for instance, BERT (Devlin et al, 2019), and fine-tuned them on an annotated dataset modeling the problem as a sequence tagging task.…”

Proceedings of the 7th Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature

“…While Jäschke et al (2017); Fischer and Jäschke (2019) used semi-automated approaches to detect VA expressions, Schwab et al (2019) developed the first automated approach for the detection of VA expressions on the sentencelevel. They developed a finer extraction approach on the word-level (Schwab et al, 2022). In particular, they employed pre-trained contextual language models, for instance, BERT , and fine-tuned them on an annotated dataset modeling the problem as a sequence tagging task.…”

“…We use the dataset from Schwab et al (2022), which is an annotated VA dataset on the word-level. The dataset consists of 5,995 sentences, of which 3,066 contain VA expressions and 2,929 do not.…”

“…In order to separate them, we created copies of such sentences for each VA expression with just one annotation resulting in 3,115 sentences. Two trained students annotated all target names inside each NYT article containing a VA expression from the dataset by Schwab et al (2022). Specifically, the annotators took a closer look at the article in which the sentence occurred and extracted the name of the entity to which the VA expression referred.…”

“…The task of extracting VA expressions focusing on source and modifier has already been covered Schwab et al, 2019Schwab et al, , 2022. The models are also able to identify a reference of the target inside a sentence where source and modifier appear, but in most cases, these references are in the form of pronouns or other mentions of the entity that cannot be linked to a knowledge base for disambiguation.…”