@inproceedings{lu-etal-2026-seem,
title = "{S}e{EM}: Selective Dictionary-based Generation",
author = "Lu, Hongyuan and
Li, Zixuan and
Lam, Wai",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {IJCNLP}-{AACL} 2026",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-aacl.245/",
pages = "4809--4830",
ISBN = "979-8-89176-444-6",
abstract = "While dictionary-based approaches have shown promise in applying Large Language Models (LLMs) to generation tasks such as Neural Machine Translation (NMT), existing studies overlook several key challenges. Most methods rely on full or rule-selected dictionaries for prompting, assume that all dictionary entries are effectively utilised by LLMs, and construct dictionaries from bilingual parallel corpora, each of which limits performance. We propose a new dictionary-based translation framework that addresses these issues by fine-tuning LLMs with LoRA to autonomously select salient dictionary entries, choosing translations with the highest dictionary utilisation, and introducing a Multilingual Lexicon Induced Dictionary that leverages multilingual parallel corpora to improve dictionary quality. Experiments on the FLORES benchmark across multiple languages demonstrate the superior performance of our framework."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lu-etal-2026-seem">
<titleInfo>
<title>SeEM: Selective Dictionary-based Generation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hongyuan</namePart>
<namePart type="family">Lu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zixuan</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wai</namePart>
<namePart type="family">Lam</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-444-6</identifier>
</relatedItem>
<abstract>While dictionary-based approaches have shown promise in applying Large Language Models (LLMs) to generation tasks such as Neural Machine Translation (NMT), existing studies overlook several key challenges. Most methods rely on full or rule-selected dictionaries for prompting, assume that all dictionary entries are effectively utilised by LLMs, and construct dictionaries from bilingual parallel corpora, each of which limits performance. We propose a new dictionary-based translation framework that addresses these issues by fine-tuning LLMs with LoRA to autonomously select salient dictionary entries, choosing translations with the highest dictionary utilisation, and introducing a Multilingual Lexicon Induced Dictionary that leverages multilingual parallel corpora to improve dictionary quality. Experiments on the FLORES benchmark across multiple languages demonstrate the superior performance of our framework.</abstract>
<identifier type="citekey">lu-etal-2026-seem</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.245/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>4809</start>
<end>4830</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SeEM: Selective Dictionary-based Generation
%A Lu, Hongyuan
%A Li, Zixuan
%A Lam, Wai
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-444-6
%F lu-etal-2026-seem
%X While dictionary-based approaches have shown promise in applying Large Language Models (LLMs) to generation tasks such as Neural Machine Translation (NMT), existing studies overlook several key challenges. Most methods rely on full or rule-selected dictionaries for prompting, assume that all dictionary entries are effectively utilised by LLMs, and construct dictionaries from bilingual parallel corpora, each of which limits performance. We propose a new dictionary-based translation framework that addresses these issues by fine-tuning LLMs with LoRA to autonomously select salient dictionary entries, choosing translations with the highest dictionary utilisation, and introducing a Multilingual Lexicon Induced Dictionary that leverages multilingual parallel corpora to improve dictionary quality. Experiments on the FLORES benchmark across multiple languages demonstrate the superior performance of our framework.
%U https://aclanthology.org/2026.findings-aacl.245/
%P 4809-4830
Markdown (Informal)
[SeEM: Selective Dictionary-based Generation](https://aclanthology.org/2026.findings-aacl.245/) (Lu et al., Findings 2026)
ACL
- Hongyuan Lu, Zixuan Li, and Wai Lam. 2026. SeEM: Selective Dictionary-based Generation. In Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026, pages 4809–4830, Hengqin, China. Association for Computational Linguistics.