A Deep Learning System for Automatic Extraction of Typological Linguistic Information from Descriptive Grammars

Proceedings of Recent Advances in Natural Language Processing, pages 1480–1489 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_166 1480 A Deep Learning System for Automatic Extraction of Typological Linguistic Information from Descriptive Grammars Shafqat Mumtaz Virk Språkbanken Text, Dept. of Swedish University of Gothenburg 405 30 Gothenburg, Sweden virk.shafqat@gmail.com Daniel Foster University of Gothenburg 405 30 Gothenburg, Sweden daniel.a.foster@gmail.com Azam Sheikh Muhammad Chalmers University of Technology 412 96 Gothenburg, Sweden sheikhazam@gmail.com Raheela Saleem Gift University Gujranwala, Pakistan raheela.saleem50@yahoo.com Abstract Linguistic typology is an area of linguistics concerned with analysis of and comparison be- tween natural languages of the world based on their certain linguistic features. For that purpose, historically, the area has relied on manual extraction of linguistic feature values from textural descriptions of languages. This makes it a laborious and time expensive task and is also bound by human brain capacity. In this study, we present a deep learning sys- tem for the task of automatic extraction of lin- guistic features from textual descriptions of natural languages. First, textual descriptions are manually annotated with special structures called semantic frames. Those annotations are learned by a recurrent neural network, which is then used to annotate un-annotated text. Fi- nally, the annotations are converted to linguis- tic feature values using a separate rule based module. Word embeddings, learned from gen- eral purpose text, are used as a major source of knowledge by the recurrent neural network. We compare the proposed deep learning sys- tem to a previously reported machine learning based system for the same task, and the deep learning system wins in terms of F1 scores with a fair margin. Such a system is expected to be a useful contribution for the automatic curation of typological databases, which other- wise are manually developed. 1 Introduction and Background Linguistic typology is an area of linguistics con- cerned with analysis of and comparison between natural languages of the world in terms of their structural and functional attributes. Among oth- ers, major objectives of the area are exploring the range of possibilities for expressing different lin- guistic categories, trying to understand the extent to which the presence of different features depend on one another in larger patterns, and to study how the global distribution of language traits has come about through an interplay of tendencies inherent to languages and historical contingencies (Bickel, 2015). For achieving such aims, historically, the area has relied heavily on scholars having to read textual documents (commonly known as descrip- tive grammars) describing languages, manually ex- tracting values of a pre-defined set of features, and then comparing languages based on the extracted feature values. To make the whole exercise system- atic and structured, traditionally, the features are expressed in the form of questions e.g. ‘What is the order of adjectives and noun in language X?’, and scholar’s job is then to find answers to such questions by reading descriptive documents about language X. The answers are often formulated to be simple strings e.g. ’NA’, ’AN’, or ’Both’ represent- ing the fact that nouns …

Content truncated to 500 words