We study the task of automatically expanding WordNet-style lexical resources to new languages through sense generation. We generate senses by associating target-language lemmas with existing lexical concepts via semantic projection. Given a sense-tagged English corpus and its translation, our method projects the annotated synsets onto aligned target-language tokens and assigns the corresponding lemmas to those synsets. To generate alignments and ensure their quality, we augment a pretrained base aligner with a bilingual dictionary, which is also used to filter incorrect sense projections. We evaluate the method on multiple languages, comparing it to prior methods, as well as dictionary-based and large language model baselines. Results show that the proposed project-and-filter strategy improves precision while remaining interpretable and resource-efficient. We release our code, documentation, and generated sense inventories at https://github.com/UAlberta-NLP/ExpandNet.
Figures & tables
French (dev)
Spanish
Chinese
Urdu
Method
Prec.
Cov.
HM
Prec.
Cov.
HM
Prec.
Cov.
HM
Prec.
Cov.
HM
UWN
46.1
56.9
50.9
71.9
51.4
60.0
45.0
5.8
10.3
59.9
13.1
21.5
Dict
5.2
93.7
9.9
8.5
88.2
15.5
7.8
75.1
14.1
9.4
44.0
21.5
OC14
71.1
8.2
14.6
72.0
9.9
17.4
34.3
2.8
5.2
44.0
0.5
1.0
M23
52.3
67.2
59.0
55.3
67.8
60.9
34.7
48.4
40.6
6.3
24.4
10.0
ChatGPT
44.3
44.3
44.3
55.3
55.3
55.3
41.8
41.8
41.8
50.5
50.5
50.5
Table 1. Results (in % ) on the SC20 corpus, showing the Sense Precision, Synset Coverage, and Harmonic Mean (HM) of the two.
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
Role
Content
System
You are a bilingual lexicon expert.
Given a dictionary definition, produce the single word in {TARGET_LANGUAGE} that best matches this definition.
Provide only the {TARGET_LANGUAGE} word without explanations.
User
{DICTIONARY_DEFINITION}
Appendix
Figure B1 . Prompt template for our LLM baselines.
Role
Content
System
You are an expert translator.
Translate from {SOURCE_LANGUAGE} to {TARGET_LANGUAGE}.
Provide only the translation without explanations.
User
{SOURCE_TEXT}
Appendix
Figure B2 . Prompt template for our translations.
n
NOUN PROPN PRON NUM NN PN PP GR PD D CA RP KP AP REP
v
VERB AUX VB AA TA KER MUL
a
ADJ AKP Q OR AD KD
r
ADV ADP PART SCONJ NEG P SE QW WALA SC I FR A
x
CCONJ INTJ SYM PUNCT DET X SPACE CC SM G U X PM EXP INT
Appendix
Table C1 . Our mapping of POS tags from multiple tagsets to the basic POS categories. All tags mapped to x are considered to denote function words and are excluded from projection.
Aligner
Dict.
POS
Prec.
Cov.
HM
DBAlign
Used
Used
82.8
66.6
73.8
DBAlign
Used
Off
80.8
68.3
74.0
SimAlign
Used
Used
84.1
62.2
71.5
SimAlign
Used
Off
83.9
59.5
69.6
SimAlign
Off
Off
49.1
68.1
57.1
Appendix
Table D1 . Results for variants of ExpandNet on SC20 , with Spanish as the target language. The Sense Precision, Synset Coverage, and their Harmonic Mean (HM) are reported for each method (in % ). “Dict.” indicates whether a dictionary-based filter was used. Similarly, “POS” indicates whether the POS filter was used.
Word Sense Disambiguation has advanced rapidly for English and a handful of well-resourced modern languages, but it continues to assume the existence of a sense inventory and a word-to-sense mapping in the source language (Navigli, 2026). These assumptions break down for most historical and low-resource languages, whose dedicated WordNets are either incomplete or still under construction. We present Inspicio, an open-vocabulary retrieval pipeline that links tokens in context to synsets of the Open English WordNet (McCrae et al., 2020) without requiring any source-language inventory or mapping. For each occurrence, an instruction-tuned LLM produces two English translations of the surrounding sentence, a small set of candidate dictionary-style definitions, and a few candidate English lemmas. These outputs drive a hybrid retrieval step that combines dense definition-synset similarity, sparse lemma matching, and Maximal Marginal Relevance re-ranking. We evaluate the pipeline across a 6x6 grid of LLMs and sentence-embedding models on a new bilingual set of manually annotated Latin and Ancient Greek perception verbs, on a subset of PREMOVE dataset (Farina, 2025), and on a diachronic sample of Italian. The best configuration reaches 96% Recall@50 on the perception-verb test set, with each component contributing measurable gains, and remains competitive in the out-of-domain and cross-lingual settings.
Michele Ciletti
Department of Humanities, University of Foggia / Via Arpi 176, 71121 Foggia, Italy
This paper proposed an algorithm for part-of-speech (POS) tagging senses of a bilingual dictionary. The algorithm is applied on the Al-Mawrid Arabic-English dictionary. The tagging task is accomplished by transferring the POS tags of the English translation equivalences (TEs) to the dictionary senses after dis-ambiguities process. The English POS tags of senses are acquired from the Princeton WordNet. POS tagging of bilingual dictionary senses is prerequisite to link a bilingual dictionary to WordNet and/or standardizing that dictionary into WordNet-LMF format where the synset (set of synonyms), not word, is the basic brick. The registered accuracy is high though the cost is little. Building NLP/HLT tools needs linguistic experts, large investments, and long time. For statistical approach, we need large annotated corpora and for rule-based approach, we need large lexicon that contains rich linguistic and world knowledge. That motivates the appearance of what are called resource-light approaches to develop natural language processing (NLP) tools for poor-resource languages.
Diaa M. Fayed, Aly A. Fahmy, Mohsen A. Rashwan +1
Computer Science, Faculty of Computers and Information, Giza, Egypt · EECE, Faculty of Engineering, Giza, Egypt · Arabic Language and Literatures, Faculty of Arts, Giza, Egypt
Sense representations (explicit, per-token meaning decompositions) are useful for disambiguation, steering, and cross-lingual alignment, but existing approaches require models to be pretrained with sense structure baked in. We introduce ACROS, which induces an explicit sense pathway into a frozen pretrained decoder LM through a gated residual addition. On SmolLM2-360M, ACROS preserves base LM quality while supporting three uses of the same induced variables: zero-shot word-sense disambiguation (64.95 F1 on Raganato ALL, competitive with the WordNet first-sense heuristic), low-KL lexical steering across 5,161 CoInCo cases where a simple non-oracle proxy recovers about 90% of positive shifts, and SENSIA cross-lingual adaptation to four languages (mean R@1 0.988, target FLORES PPL 7.94). ACROS makes sense representations an inducible interface for ordinary pretrained LMs.