摘要：This article presents a robust system of phrase punctuation recognition. It is an adaptive multilingual one. Its accuracy exceeds 99% in different linguistic corpora (French, greek and Arabic). The system is a rule-based segmentation algorithm. The rules are statistics issued themselves based on distributional analysis [Harris, 1954]. No previous knowledge or dictionary are previously. System’s portability as well as adaptability are precious for any further application or analysis like machine translation, information extraction, text alignment and document summarization.
