METHOD AND APPARATUS FOR TERMINOLOGY RECOGNITION BASED ON WEB SEARCH

摘要

PURPOSE: A web search based word recognition method and a device thereof are provided to use a web search result for a word candidate extracted from a document group as statistical information for assigning a weighted value for the word candidate, thereby reflecting the weighted value and recognizing a new word. CONSTITUTION: A word candidate extraction unit obtains part of speech information and original word information by analyzing sentences of an input document. The word candidate extraction unit extracts word candidates by using the part of speech information, the original word information, and a stored word candidate pattern. A quality extraction unit obtains basic quality for the word candidates and web quality of a web search result(S306). A word recognition unit assigns a weighted value by applying machine learning to the basic quality and the web quality and recognizes a word candidate which the weighted value is the highest as a word(S308). [Reference numerals] (AA) Start; (BB) End; (S302) Extracting word candidates by analyzing an input document; (S304) Obtaining basic quality for the extracted word candidates and web quality of a web search resu (S306) Obtaining basic quality of each word candidate and web quality of the web search resu (S308) Recognizing a word candidate which the weighted value is the highest as a word