Please see our paper for more details! Overall, I really enjoyed working on this project, and I’m excited to see what the future holds in terms of more research that bridges psycholinguistic and artificial intelligence theories

Code: github.com/EshaanT/Bili...

Arxiv: arxiv.org/abs/2501.09127

For English homographs, LLMs identify the language and use the word's correct semantic meaning to resolve ambiguous sentences. But for most non-English homographs, they rely on the meaning of the English counterpart instead.

7/

Multilingual LLMs also seem to adopt different strategies to handle the complexity of interlingual homographs, particularly in cross-lingual sentence contexts involving English and non-English words.

6/

Furthermore, we found no correlation between the semantic judgment capacity of LLMs and their word pair disambiguation ability. Indicating the lack of utility of semantic cues in the disambiguation task and the lack of utility of orthographic signals in semantic retrieval.

5/

However, this does not extend to the semantic retrieval ability of models, where no single category is easier for the models to retrieve the meaning of, which is contrary to the expectation of cognate facilitation.

4/

LLMs demonstrate a facilitation effect when disambiguating cognates, mirroring human bilingual processing patterns, where shared orthographic and lexical representations simplify word recognition and understanding.

3/

🔹Cognates (e.g. exit in English and éxito in Spanish),
🔹 Non-cognates (e.g. dog in English and perro in
Spanish), and
🔹Interlingual homographs (e.g. gift meaning “present” in English but “poison” in German).

We focus on understanding how LLMs process these word types.

2/