Summarising Historical Text in Modern Languages

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 3123–3142 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 3123 Summarising Historical Text in Modern Languages Xutan Peng Yi Zheng Chenghua Lin ∗ Advaith Siddharthan Department of Computer Science, The University of Sheffield, UK School of Computer Science and Engineering, Beihang University, China Knowledge Media Institute, The Open University, UK {x.peng, c.lin}@shef.ac.uk zhengyi53@buaa.edu.cn advaith.siddharthan@open.ac.uk Abstract We introduce the task of historical text sum- marisation, where documents in historical forms of a language are summarised in the corresponding modern language. This is a fundamentally important routine to histo- rians and digital humanities researchers but has never been automated. We compile a high-quality gold-standard text summarisation dataset, which consists of historical German and Chinese news from hundreds of years ago summarised in modern German or Chinese. Based on cross-lingual transfer learning tech- niques, we propose a summarisation model that can be trained even with no cross-lingual (historical to modern) parallel data, and fur- ther benchmark it against state-of-the-art algo- rithms. We report automatic and human eval- uations that distinguish the historic to mod- ern language summarisation task from stan- dard cross-lingual summarisation (i.e., mod- ern to modern language), highlight the dis- tinctness and value of our dataset, and demon- strate that our transfer learning approach out- performs standard cross-lingual benchmarks on this task. 1 Introduction The process of text summarisation is fundamental to research into history, archaeology, and digital humanities (South, 1977). Researchers can better gather and organise information and share knowl- edge by first identifying the key points in historical documents. However, this can cost a lot of time and effort. On one hand, due to cultural and linguis- tic variations over time, interpreting historical text can be a challenging and energy-consuming pro- cess, even for those with specialist training (Gray et al., 2011). To compound this, historical archives can contain narrative documents on a large scale, ∗Chenghua Lin is the corresponding author. adding to the workload of manually locating im- portant elements (Gunn, 2011). To reduce these burdens, specialised software has been developed recently, such as MARKUS (Ho and Weerdt, 2014) and DocuSky (Tu et al., 2020). These toolkits aid users in managing and annotating documents but still lack functionalities to automatically process texts at a semantic level. Historical text summarisation can be regarded as a special case of cross-lingual summarisa- tion (Leuski et al., 2003; Or˘asan and Chiorean, 2008; Cao et al., 2020), a long-standing research topic whereby summaries are generated in a tar- get language from documents in different source languages. However, historical text summarisa- tion posits some unique challenges. Cross-lingual (i.e., across historical and modern forms of a lan- guage) corpora are rather limited (Gray et al., 2011) and therefore historical texts cannot be handled by traditional cross-lingual summarisers, which re- quire cross-lingual supervision or at least large sum- marisation datasets in both languages (Cao et al., 2020). Further, language use evolves over time, including vocabulary and word spellings …

Content truncated to 500 words