@inproceedings{mishra-haghighi-2021-improved,
title = "Improved Multilingual Language Model Pretraining for Social Media Text via Translation Pair Prediction",
author = "Mishra, Shubhanshu and
Haghighi, Aria",
editor = "Xu, Wei and
Ritter, Alan and
Baldwin, Tim and
Rahimi, Afshin",
booktitle = "Proceedings of the Seventh Workshop on Noisy User-generated Text (W-NUT 2021)",
month = nov,
year = "2021",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.wnut-1.42/",
doi = "10.18653/v1/2021.wnut-1.42",
pages = "381--388",
abstract = "We evaluate a simple approach to improving zero-shot multilingual transfer of mBERT on social media corpus by adding a pretraining task called translation pair prediction (TPP), which predicts whether a pair of cross-lingual texts are a valid translation. Our approach assumes access to translations (exact or approximate) between source-target language pairs, where we fine-tune a model on source language task data and evaluate the model in the target language. In particular, we focus on language pairs where transfer learning is difficult for mBERT: those where source and target languages are different in script, vocabulary, and linguistic typology. We show improvements from TPP pretraining over mBERT alone in zero-shot transfer from English to Hindi, Arabic, and Japanese on two social media tasks: NER (a 37{\%} average relative improvement in F1 across target languages) and sentiment classification (12{\%} relative improvement in F1) on social media text, while also benchmarking on a non-social media task of Universal Dependency POS tagging (6.7{\%} relative improvement in accuracy). Our results are promising given the lack of social media bitext corpus. Our code can be found at: \url{https://github.com/twitter-research/multilingual-alignment-tpp}."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mishra-haghighi-2021-improved">
<titleInfo>
<title>Improved Multilingual Language Model Pretraining for Social Media Text via Translation Pair Prediction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shubhanshu</namePart>
<namePart type="family">Mishra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aria</namePart>
<namePart type="family">Haghighi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2021-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Seventh Workshop on Noisy User-generated Text (W-NUT 2021)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alan</namePart>
<namePart type="family">Ritter</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tim</namePart>
<namePart type="family">Baldwin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Afshin</namePart>
<namePart type="family">Rahimi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Online</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>We evaluate a simple approach to improving zero-shot multilingual transfer of mBERT on social media corpus by adding a pretraining task called translation pair prediction (TPP), which predicts whether a pair of cross-lingual texts are a valid translation. Our approach assumes access to translations (exact or approximate) between source-target language pairs, where we fine-tune a model on source language task data and evaluate the model in the target language. In particular, we focus on language pairs where transfer learning is difficult for mBERT: those where source and target languages are different in script, vocabulary, and linguistic typology. We show improvements from TPP pretraining over mBERT alone in zero-shot transfer from English to Hindi, Arabic, and Japanese on two social media tasks: NER (a 37% average relative improvement in F1 across target languages) and sentiment classification (12% relative improvement in F1) on social media text, while also benchmarking on a non-social media task of Universal Dependency POS tagging (6.7% relative improvement in accuracy). Our results are promising given the lack of social media bitext corpus. Our code can be found at: https://github.com/twitter-research/multilingual-alignment-tpp.</abstract>
<identifier type="citekey">mishra-haghighi-2021-improved</identifier>
<identifier type="doi">10.18653/v1/2021.wnut-1.42</identifier>
<location>
<url>https://aclanthology.org/2021.wnut-1.42/</url>
</location>
<part>
<date>2021-11</date>
<extent unit="page">
<start>381</start>
<end>388</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Improved Multilingual Language Model Pretraining for Social Media Text via Translation Pair Prediction
%A Mishra, Shubhanshu
%A Haghighi, Aria
%Y Xu, Wei
%Y Ritter, Alan
%Y Baldwin, Tim
%Y Rahimi, Afshin
%S Proceedings of the Seventh Workshop on Noisy User-generated Text (W-NUT 2021)
%D 2021
%8 November
%I Association for Computational Linguistics
%C Online
%F mishra-haghighi-2021-improved
%X We evaluate a simple approach to improving zero-shot multilingual transfer of mBERT on social media corpus by adding a pretraining task called translation pair prediction (TPP), which predicts whether a pair of cross-lingual texts are a valid translation. Our approach assumes access to translations (exact or approximate) between source-target language pairs, where we fine-tune a model on source language task data and evaluate the model in the target language. In particular, we focus on language pairs where transfer learning is difficult for mBERT: those where source and target languages are different in script, vocabulary, and linguistic typology. We show improvements from TPP pretraining over mBERT alone in zero-shot transfer from English to Hindi, Arabic, and Japanese on two social media tasks: NER (a 37% average relative improvement in F1 across target languages) and sentiment classification (12% relative improvement in F1) on social media text, while also benchmarking on a non-social media task of Universal Dependency POS tagging (6.7% relative improvement in accuracy). Our results are promising given the lack of social media bitext corpus. Our code can be found at: https://github.com/twitter-research/multilingual-alignment-tpp.
%R 10.18653/v1/2021.wnut-1.42
%U https://aclanthology.org/2021.wnut-1.42/
%U https://doi.org/10.18653/v1/2021.wnut-1.42
%P 381-388
Markdown (Informal)
[Improved Multilingual Language Model Pretraining for Social Media Text via Translation Pair Prediction](https://aclanthology.org/2021.wnut-1.42/) (Mishra & Haghighi, WNUT 2021)
ACL