engleski

Detecting Measurement Expressions using NooJ

We present a NooJ module implementing a general method for detection and classification of measurement expressions in English and Croatian newspaper texts using local regular grammars. Expressions involving the most frequently used units of measurement are covered for both languages. Insight on module design is provided along with its evaluation. Overall accuracy of the module reaches above 96 and 98 percent for Croatian and English, respectively. Issues regarding normalization of detected measurement expressions are also discussed.

measurement expressions; units of measurement; NooJ; detection; normalization

nije evidentirano