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.
Figures & tables
Figure 1: Workflow for the creation of CrisiLangs . Event descriptions and FEMA guidelines are machine-translated into each target language, while the warning messages are translated and adapted by a native speaker.
Figure 2: Example of event and warning message from CrisiLangs in the five languages. The English event and message are the originals from CrisiText . In the other four languages the messages are human translations produced by native speakers, while the events are machine-translated.
Category
Scenarios
Events
Human-caused
253
500
explosion
105
201
fire
76
153
attack
72
146
Natural
158
500
tornado
33
111
Table 1: Distribution of the event and message main categories over the 1,000 selected elements, grouped into human-caused and natural crises.
All
EN
IT
DE
EL
BG
306.9
268.6
316.4
324.1
331.1
294.4
Table 2: Mean warning message length in characters per language in CrisiLangs .
User
† {Based on the provided guidelines,} Create a warning message informing on the current happening, providing a suggestion, for the last line in the following chain of events (be short, max 300 characters). No other output other than the message.
† Guidelines: [GUIDELINES]
Chain of events: [EVENTS]
Figure 3: Prompt example. The instruction is always in English, while [EVENTS] is replaced by the chain of events of the instance, in its language, with the target event as its last line. The segments marked with † appear only in the +guidelines configuration, where [GUIDELINES] is replaced by the translated FEMA guidelines of the crisis categories of the event.
All
English
Italian
German
Greek
Bulgarian
T
R-L
B
BS
R-L
B
BS
R-L
B
BS
R-L
B
BS
R-L
B
BS
R-L
B
BS
Zero-shot
plain
1.0
.168
.037
.878
.209
.072
.890
.175
.041
.882
.185
.038
.884
.130
.019
.865
.143
.016
.871
0.8
.184
.043
.884
.232
.083
.893
.196
.045
.886
.202
.047
.888
.145
.021
.874
.148
.021
.878
0.6
.201
.048
.888
.234
.084
.895
.215
.049
.891
.218
.055
.892
.176
.032
.883
.166
.021
.883
0.4
.205
.051
.890
.232
.083
.894
.214
.053
.890
.226
.050
.894
.174
.036
.887
.181
.035
.887
0.2
.211
.056
.892
.240
.090
.894
.221
.056
.892
.233
.059
.896
.182
.039
.888
.182
.035
.889
Table 3: ROUGE-L F1 (R-L), BLEU (B) and BERTScore F1 (BS) on the test split, per configuration, temperature and language. Bold marks the best temperature within each block. Zero-shot uses the un-finetuned Llama-3.1-8B-Instruct with identical configuration.
Length
BLEU w/o BP
Gold
304
–
ZS, plain
173
.114
ZS, +guid.
205
.104
FT, plain
295
.177
FT, +guid.
299
.171
Table 4: Mean generation length in characters and BLEU without the brevity penalty (BP), at T=0.2 , over all five languages, for the zero-shot (ZS) and fine-tuned (FT) models. Per-language values are in Appendix B .
plain
+guidelines
0.6
0.4
0.2
N
0.6
0.4
0.2
N
English
8
15
27
50
12
8
30
50
Italian
6
15
29
50
7
15
27
49
German
12
13
24
49
7
17
26
50
Greek
11
2
37
50
4
22
24
50
Bulgarian
4
17
29
50
3
9
25
37
Table 5: Preference counts per configuration and language. N is the number of annotated events out of 50.
Appendix figures & tables2 assets
Supplementary material from the paper’s appendix.
Appendix
All
EN
IT
BG
DE
EL
mean length (characters)
Gold
304
267
315
289
317
330
ZS, plain
173
188
163
163
182
168
ZS, +guid.
205
204
214
192
221
195
FT, plain
295
269
297
272
326
309
FT, +guid.
299
270
312
284
317
314
Appendix
Table 6: Per-language mean generation length in characters and BLEU without the brevity penalty, at T=0.2 , for the zero-shot (ZS) and fine-tuned (FT) models with the plain and +guidelines (+guid.) prompts. As in Table 3 , the comparison is intra-language. Languages are English (EN), Italian (IT), Bulgarian (BG), German (DE) and Greek (EL).
Translation task
The task consists in creating alert messages in your language based on the English alert messages we provide. We ask you to translate the messages and, when necessary, adapt them to sound more natural and appropriate as alert messages in your language.
Preference task
Your task is to:
– For each row, select your preferred message using the dropdown menu, based on grammaticality and consistency with the task.
– At the top of each sheet, indicate which configuration (column) you preferred overall, and briefly explain why.
– Let us know which of the two models (sheets) you preferred overall, and why.
Appendix
Figure 4: Instructions given to the annotators, reproduced verbatim. The translation instructions were provided together with the English source; the preference instructions accompanied one spreadsheet per training configuration.