@inproceedings{wang-etal-2025-bimax,
    title = "{B}i{M}ax: Bidirectional {M}ax{S}im Score for Document-Level Alignment",
    author = "Wang, Xiaotian  and
      Utsuro, Takehito  and
      Nagata, Masaaki",
    editor = "Christodoulopoulos, Christos  and
      Chakraborty, Tanmoy  and
      Rose, Carolyn  and
      Peng, Violet",
    booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2025",
    month = nov,
    year = "2025",
    address = "Suzhou, China",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/author-page-yu-wang-polytechnic/2025.findings-emnlp.704/",
    doi = "10.18653/v1/2025.findings-emnlp.704",
    pages = "13095--13116",
    ISBN = "979-8-89176-335-7",
    abstract = "Document alignment is necessary for the hierarchical mining, which aligns documents across source and target languages within the same web domain. Several high-precision sentence embedding-based methods have been developed, such as TK-PERT and Optimal Transport (OT). However, given the massive scale of web mining data, both accuracy and speed must be considered.In this paper, we propose a cross-lingual Bidirectional Maxsim score (BiMax) for computing doc-to-doc similarity,to improve efficiency compared to the OT method.Consequently, on the WMT16 bilingual document alignment task,BiMax attains accuracy comparable to OT with an approximate 100-fold speed increase.Meanwhile, we also conduct a comprehensive analysis to investigate the performance of current state-of-the-art multilingual sentence embedding models."
}