The present disclosure discloses a method and apparatus of selecting a word sequence for a text written in a language without word boundary in order to solve the problem of having excessively large computation load when selecting an optimal word sequence in existing technologies. The disclosed method includes: segmenting a segment of the text to obtain different word sequences; determining a common word boundary for the word sequences; and performing optimal word sequence selection for portions of the word sequences prior to the common word boundary. Because optimal word sequence selection is performed for portions of word sequences prior to a common word boundary, shorter independent units can be obtained, thus reducing computation load of word segmentation.