It has been a widespread tendency that large amount of new words are emerging in web text corpus. Among these are many new words created by netizens or arising from social focuses

and are also many colloquial expressions

abbreviations in the social intercourse corpus created by SNS. All the above cases together make it diffi cult for words segmentation. In this essay a new extraction method for new words is proposed

aiming to extract new words in a certain corpus

to generate a dictionary and to segment the Chinese expressions more accurately. The new method fi rstly extracts candidate words from the corpus

and then calculates its support and confi dence

sifts the new words out

and fi nally extracts new words accurately and rapidly from huge text data.