Chinese automatic segmentation is the basis of web text mining and other Chinese information processing applications. Booming Chinese information processing applications put forward a higher requirement for Chinese automatic segmentation. This paper presents a new segmentation algorithm FPLS, which uses a dictionary with a first letter of the Pinyin as a first level index and words count as the secondary index structure. A bidirectional matching method and rules are employed to resolve ambiguity segmentation problem in the algorithm. Comparing with the existing algorithm, algorithm FPLS gets higher accuracy and efficiency.