@inproceedings{lorandi-belz-2023-data,
title = "Data-to-text Generation for Severely Under-Resourced Languages with {GPT}-3.5: A Bit of Help Needed from {G}oogle {T}ranslate ({W}eb{NLG} 2023)",
author = "Lorandi, Michela and
Belz, Anya",
editor = "Gatt, Albert and
Gardent, Claire and
Cripwell, Liam and
Belz, Anya and
Borg, Claudia and
Erdem, Aykut and
Erdem, Erkut",
booktitle = "Proceedings of the Workshop on Multimodal, Multilingual Natural Language Generation and Multilingual WebNLG Challenge (MM-NLG 2023)",
month = sep,
year = "2023",
address = "Prague, Czech Republic",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.mmnlg-1.9",
pages = "80--86",
abstract = "LLMs are great at tasks involving English which dominates in their training data. We explore their ability to address tasks involving languages that are severely under-represented in their training data. More specifically, we do this in the context of data-to-text generation for Irish, Maltese, Welsh and Breton. During the prompt-engineering phase we tested GPT-3.5 and{\textasciitilde}4 with a range of prompt types and formats on a small sample of example input/output pairs. We then fully evaluated the two most promising prompts in two scenarios: (i) direct generation into the under-resourced languages, and (ii) generation into English followed by translation into the under-resourced languages. We find that few-shot prompting works better for direct generation into under-resourced languages, but that the difference disappears when pivoting via English. The few-shot + translation system variants were submitted to the WebNLG 2023 shared task where they outperformed all other systems by substantial margins in all languages on all automatic metrics. We conclude that good performance can be achieved with state-of-the-art LLMs out-of-the box for under-resourced languages. However, best results (for Welsh) of BLEU 25.12, ChrF++ 0.55, and TER 0.64 are well below the lowest ranked English system at WebNLG{'}20 with BLEU 0.391, ChrF++ 0.579, and TER 0.564.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lorandi-belz-2023-data">
<titleInfo>
<title>Data-to-text Generation for Severely Under-Resourced Languages with GPT-3.5: A Bit of Help Needed from Google Translate (WebNLG 2023)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Michela</namePart>
<namePart type="family">Lorandi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anya</namePart>
<namePart type="family">Belz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-09</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Workshop on Multimodal, Multilingual Natural Language Generation and Multilingual WebNLG Challenge (MM-NLG 2023)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Albert</namePart>
<namePart type="family">Gatt</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Claire</namePart>
<namePart type="family">Gardent</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liam</namePart>
<namePart type="family">Cripwell</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anya</namePart>
<namePart type="family">Belz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Claudia</namePart>
<namePart type="family">Borg</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aykut</namePart>
<namePart type="family">Erdem</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Erkut</namePart>
<namePart type="family">Erdem</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Prague, Czech Republic</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>LLMs are great at tasks involving English which dominates in their training data. We explore their ability to address tasks involving languages that are severely under-represented in their training data. More specifically, we do this in the context of data-to-text generation for Irish, Maltese, Welsh and Breton. During the prompt-engineering phase we tested GPT-3.5 and~4 with a range of prompt types and formats on a small sample of example input/output pairs. We then fully evaluated the two most promising prompts in two scenarios: (i) direct generation into the under-resourced languages, and (ii) generation into English followed by translation into the under-resourced languages. We find that few-shot prompting works better for direct generation into under-resourced languages, but that the difference disappears when pivoting via English. The few-shot + translation system variants were submitted to the WebNLG 2023 shared task where they outperformed all other systems by substantial margins in all languages on all automatic metrics. We conclude that good performance can be achieved with state-of-the-art LLMs out-of-the box for under-resourced languages. However, best results (for Welsh) of BLEU 25.12, ChrF++ 0.55, and TER 0.64 are well below the lowest ranked English system at WebNLG’20 with BLEU 0.391, ChrF++ 0.579, and TER 0.564.</abstract>
<identifier type="citekey">lorandi-belz-2023-data</identifier>
<location>
<url>https://aclanthology.org/2023.mmnlg-1.9</url>
</location>
<part>
<date>2023-09</date>
<extent unit="page">
<start>80</start>
<end>86</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Data-to-text Generation for Severely Under-Resourced Languages with GPT-3.5: A Bit of Help Needed from Google Translate (WebNLG 2023)
%A Lorandi, Michela
%A Belz, Anya
%Y Gatt, Albert
%Y Gardent, Claire
%Y Cripwell, Liam
%Y Belz, Anya
%Y Borg, Claudia
%Y Erdem, Aykut
%Y Erdem, Erkut
%S Proceedings of the Workshop on Multimodal, Multilingual Natural Language Generation and Multilingual WebNLG Challenge (MM-NLG 2023)
%D 2023
%8 September
%I Association for Computational Linguistics
%C Prague, Czech Republic
%F lorandi-belz-2023-data
%X LLMs are great at tasks involving English which dominates in their training data. We explore their ability to address tasks involving languages that are severely under-represented in their training data. More specifically, we do this in the context of data-to-text generation for Irish, Maltese, Welsh and Breton. During the prompt-engineering phase we tested GPT-3.5 and~4 with a range of prompt types and formats on a small sample of example input/output pairs. We then fully evaluated the two most promising prompts in two scenarios: (i) direct generation into the under-resourced languages, and (ii) generation into English followed by translation into the under-resourced languages. We find that few-shot prompting works better for direct generation into under-resourced languages, but that the difference disappears when pivoting via English. The few-shot + translation system variants were submitted to the WebNLG 2023 shared task where they outperformed all other systems by substantial margins in all languages on all automatic metrics. We conclude that good performance can be achieved with state-of-the-art LLMs out-of-the box for under-resourced languages. However, best results (for Welsh) of BLEU 25.12, ChrF++ 0.55, and TER 0.64 are well below the lowest ranked English system at WebNLG’20 with BLEU 0.391, ChrF++ 0.579, and TER 0.564.
%U https://aclanthology.org/2023.mmnlg-1.9
%P 80-86
Markdown (Informal)
[Data-to-text Generation for Severely Under-Resourced Languages with GPT-3.5: A Bit of Help Needed from Google Translate (WebNLG 2023)](https://aclanthology.org/2023.mmnlg-1.9) (Lorandi & Belz, MMNLG-WS 2023)
ACL