Automatic Extraction of the Romanian Academic Word List: Data and Methods

Proceedings of Recent Advances in Natural Language Processing, pages 234–241 Varna, Sep 4–6, 2023 https://doi.org/10.26615/978-954-452-092-2_026 234 Automatic Extraction of the Romanian Academic Word List: Data and Methods Ana-Maria Bucur1,2, Andreea Dinc˘a2, M˘ad˘alina Chitez2, Roxana Rogobete2 1 Interdisciplinary School of Doctoral Studies, University of Bucharest 2 West University of Timis,oara, Romania ana-maria.bucur@drd.unibuc.ro {madalina.chitez, andreea.dinca, roxana.rogobete}@e-uvt.ro Abstract This paper presents the methodology and data used for the automatic extraction of the Ro- manian Academic Word List (Ro-AWL). Aca- demic Word Lists are useful in both L2 and L1 teaching contexts. For the Romanian language, no such resource exists so far. Ro-AWL has been generated by combining methods from corpus and computational linguistics with L2 academic writing approaches. We use two types of data: (a) existing data, such as the Ro- manian Frequency List based on the ROMBAC corpus, and (b) self-compiled data, such as the expert academic writing corpus EXPRES. For constructing the academic word list, we follow the methodology for building the Academic Vocabulary List for the English language. The distribution of Ro-AWL features (general distri- bution, POS distribution) into four disciplinary datasets is in line with previous research. Ro- AWL is freely available and can be used for teaching, research and NLP applications. 1 Introduction Since academic language differs from everyday social language and is an essential acquisition tar- get in current education, extracting salient features contributes to linguistic, register, genre and disci- plinary feature identification that can benefit stu- dents, teachers and educational app developers alike. Compiling an Academic Word List (AWL) is an effective solution to support both language teaching and NLP tasks. From the didactic perspec- tive, AWLs reflecting either the L1 (i.e. mother tongue) or the L2 (i.e. foreign language) academic vocabulary can be used to offer linguistic support to novice academic writers in the form of discipline- specific and general lexical prompts. Teachers of all disciplines can integrate AWLs into teaching materials to help students write better (see, for ex- ample, Wangdi and Shimray (2022)). NLP studies can exploit AWL datasets on top- ics such as text classification (Zampieri, 2012) and topic modelling (Murakami et al., 2017). For ex- ample, field-specific academic lists can be used to automatically classify texts into disciplinary areas. The same can be applied for the automatic distri- bution of texts in academic versus non-academic batches. In machine learning methods for language modelling tasks, AWLs are essential in training models to generate accurate academic writing sam- ples. By combining NLP tasks with linguistic ap- proaches in relation to AWLs, important advances can be achieved in the frame of lexical and syntac- tic analyses that evaluate the use of collocations and phraseology specific to the academic varieties. For the Romanian language, there have been few attempts to extract a valid Romanian Word List (Szabo, 2015) and only one study has extracted and analysed multiword units using academic writing corpora (Muresan et al., 2022). In recent years, researchers have worked to cre- ate several academic writing corpora. EXPRES – Corpus of Expert Writing …

Content truncated to 500 words