Word Sense Disambiguation Dataset in Historical Danish and Norwegian Texts
Overview
This README provides information about the dataset for word sense disambiguation in historical texts, focusing on the transformation of the concept of fate ('skæbne') from its pre-modern to modern sense in the latter part of the 19th century. The dataset is established and annotated by a Danish-speaking literary scholar.
Task Description
The primary task addressed by this dataset is word sense disambiguation (WSD) in the context of historical texts. Specifically, the dataset explores the evolution of the concept of fate from its pre-modern, religiously and metaphysically inflected sense to a modern meaning that incorporates a secular and material understanding of the world.
Dataset Creation
The dataset is introduced through a novel annotation process carried out by one of the authors, a Danish-speaking literary scholar. The focus is on the semantic and ideological division of the concept of fate in the latter part of the 19th century.
Dataset Details
Size: The dataset consists of 650 segments in total, with 48 segments representing the pre-modern sense and 56 segments representing the modern notion of fate. Labeling: Segments are labeled as either pre-modern or modern based on predefined criteria. The labels generally correspond to the dichotomy between religious/metaphysical notions and secular/material notions. Content Focus: The dataset was produced with a specific focus on conceptual content. Note that the word 'skæbne' may not appear in every segment. Word Occurrence: The word 'skæbne' is present in 27 out of the 48 segments representing the pre-modern sense and in 24 out of the 56 segments representing the modern notion. Splitting: The dataset is split into training, development, and testing sets with proportions 80%, 10%, and 10%, respectively. Usage Researchers and practitioners interested in WSD in historical texts, particularly in the evolution of the concept of fate, can utilize this dataset for training, development, and evaluation purposes.
Citation
- Downloads last month
- 35