Giacomo Gonella, Stefano Menini, Marco Guerini · 2026 · arXiv (Cornell University)
DOI: 10.48550/arxiv.2610.31356 · cited by 0 works (OpenAlex) · open-access PDF
Berkomunikasi dengan publik selama krisis melibatkan pengiriman pesan peringatan dengan cepat. Di Eropa, seringkali diperlukan untuk memproduksi mereka dalam lebih dari satu bahasa. Pekerjaan baru-baru ini telah menunjukkan bahwa model bahasa dapat dilatih untuk menghasilkan pesan peringatan saat krisis terjadi, tetapi data yang tersedia hanya mencakup bahasa Inggris. Untuk mengisi kesenjangan ini, kami menyajikan CrisiLangs, ekstensi multibahasa dari CrisiText (set data krisis bahasa Inggris dikombinasikan dengan pesan peringatan yang ditulis sesuai dengan pedoman komunikasi ahli) ke bahasa Italia, Bulgaria, Jerman, dan Yunani. Pembicara asli ahli menerjemahkan pesan peringatan dari 1.000 peristiwa yang diambil dari 411 skenario, untuk total 5.000 pasangan pesan peristiwa di lima bahasa. Kami menguji sumber daya untuk menyesuaikan model multibahasa di semua lima bahasa secara bersama-sama, dengan dan tanpa
Machine translation (Meta M2M100); the English original follows.
Communicating with the public during an emergency entails issuing warning messages promptly. In Europe, it often requires producing them in more than one language. Recent work has shown that language models can be trained to generate warning messages as a crisis unfolds, but the available data covers only English. To fill this gap, we present CrisiLangs, a multilingual extension of CrisiText (an English dataset of crisis events paired with warning messages written according to expert communication guidelines) to Italian, Bulgarian, German, and Greek. Expert native speakers translated the warning messages of 1,000 events drawn from 411 scenarios, for a total of 5,000 event-message pairs across five languages. We test the resource to fine-tune a multilingual model on all five languages jointly, with and without the emergency guidelines in the prompt, and compare it with a zero-shot baseline. The fine-tuned models outperform the baseline on ROUGE-L, BLEU and BERTScore in every language and configuration. In a human preference study, low decoding temperatures receive the most preferences in all five languages and in both configurations, while the effect of the guidelines varies by language.
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
Topics: Computer science, Resource (disambiguation), Warning system, Preference, Natural language processing, Language model
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.31356, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange