andreaschari/mt5-ZH_MMARCO_50_MIXED_ANSERINI

mt5-base Reranker ZH mMARCO/v2 50/50 Native Transliterated Queries tokenised with Anserini

This is a variation of Unicamp's mt5-base Reranker initially finetuned on mMARCOv/2.

The queries are a 50/50 split between native Chinese and transliterated Chinese to English text using uroman. The queries were tokenised with pyterrier_anserini.

The model was used for the SIGIR 2025 Short paper: Lost in Transliteration: Bridging the Script Gap in Neural IR.