Cross-lingual Science Journalism: Select, Simplify and Rewrite Summaries for Non-expert Readers
38780Authors
nan
Abstract
AbstractAutomating Cross-lingual Science Journalism (CSJ) aims to generate popular science summaries from English scientific texts for non-expert readers in their local language. We introduce CSJ as a downstream task of text simplification and cross-lingual scientific summarization to facilitate science journalists’ work. We analyze the performance of possible existing solutions as baselines for the CSJ task. Based on these findings, we propose to combine the three components - SELECT, SIMPLIFY and REWRITE (SSR) to produce cross-lingual simplified science summaries for non-expert readers. Our empirical evaluation on the Wikipedia dataset shows that SSR significantly outperforms the baselines for the CSJ task and can serve as a strong baseline for future work. We also perform an ablation study investigating the impact of individual components of SSR. Further, we analyze the performance of SSR on a high-quality, real-world CSJ dataset with human evaluation and in-depth analysis, demonstrating the superior performance of SSR for CSJ.
Keywords
Full Content
Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 1843–1861 July 9-14, 2023 ©2023 Association for Computational Linguistics Cross-lingual Science Journalism: SELECT, SIMPLIFY and REWRITE Summaries for Non-expert Readers Mehwish Fatima and Michael Strube Heidelberg Institute for Theoretical Studies (mehwish.fatima|michael.strube)@h-its.org Abstract Automating Cross-lingual Science Journalism (CSJ) aims to generate popular science sum- maries from English scientific texts for non- expert readers in their local language. We in- troduce CSJ as a downstream task of text sim- plification and cross-lingual scientific summa- rization to facilitate science journalists’ work. We analyze the performance of possible ex- isting solutions as baselines for the CSJ task. Based on these findings, we propose to com- bine the three components - SELECT, SIMPLIFY and REWRITE (SSR) to produce cross-lingual simplified science summaries for non-expert readers. Our empirical evaluation on the WIKIPEDIA dataset shows that SSR significantly outperforms the baselines for the CSJ task and can serve as a strong baseline for future work. We also perform an ablation study investigat- ing the impact of individual components of SSR. Further, we analyze the performance of SSR on a high-quality, real-world CSJ dataset with hu- man evaluation and in-depth analysis, demon- strating the superior performance of SSR for CSJ. 1 Introduction Cross-lingual Science Journalism (CSJ) produces popular science stories1 from English scientific texts for non-expert readers in their local language. CSJ focuses on simultaneously reducing linguistic complexity and length of the original text. Au- tomating CSJ can facilitate science journalists in their work for writing popular science stories. A real-world example of CSJ is Spektrum der Wis- senschaft (SPEKTRUM)2. It is a popular monthly science magazine, the German version of Scientific American. The magazine is considered a bridge between scientific literature and the non-expert pub- lic. Our work is initiated by a collaboration with SPEK- 1We quote stories and summaries interchangeably in this work. 2https://www.spektrum.de/ Target: ein mädchen aus südafrika lebt seit neun jahren mit einer hiv-infektion, ohne den erreger mit medikamenten kon- trollieren zu müssen. das berichteten fachleute um avy violari von der university of the witwatersrand in johannesburg auf einer konferenz in paris. [...] Translation: a girl from south africa has been living with an hiv infection for nine years without having to control the pathogen with medication. this is what experts around avy vi- olari from the university of the witwatersrand in johannesburg reported at a conference in paris. [...] mBART: das kind, das fiziert wird mit dem hi-virus bei der geburt, ist die geburtsache für das infizieren des virus bei der geburt. die infektion des ungeborenen kindes mit dem hiv wird als latente hiv-infektion bezeichnet. die ursache für die erkrankung ist noch ungeklärt. Translation: the child who is <UNK> with the hi virus at birth is the <UNK> for infecting the virus at birth. the infec- tion of the unborn child with hiv is referred to as latent hiv infection. the cause of the disease is still unclear. SSR: ein neunjähriges kind, das ohne behandlung mit …
Content truncated to 500 words
Labeled Text
AbstractAutomating Cross-lingual Science Journalism (CSJ) aims to generate popular science summaries from English scientific texts for non-expert readers in their local language. We introduce CSJ as a downstream task of text simplification and cross-lingual scientific summarization to facilitate science journalists’ work. We analyze the performance of possible existing solutions as baselines for the CSJ task. Based on these findings, we propose to combine the three components - SELECT, SIMPLIFY and REWRITE (SSR) to produce cross-lingual simplified science summaries for non-expert readers. Our empirical evaluation on the Wikipedia dataset shows that SSR significantly outperforms the baselines for the CSJ task and can serve as a strong baseline for future work. We also perform an ablation study investigating the impact of individual components of SSR. Further, we analyze the performance of SSR on a high-quality, real-world CSJ dataset with human evaluation and in-depth analysis, demonstrating the superior performance of SSR for CSJ.
Publication Details
- Paper ID:
- 38780
- Year:
- 2023
- Publication:
- TIPSTER
- Publication Type:
- conf
- DOI:
- nan