Syntactic Chronology: Dating Text in the History of English
Satoru Tsukamoto · 2004
This study illustrates an application of a statistical technique to text analysis. Using the syntactic annotation in the Penn-Helsinki Parsed corpus of Middle English Phase1 and imitating carbon dating technique, this research employs multiple regression analysis to estimate the date of a given text. A regression equation with seven variables was developed for estimating text date. About 75% of variation in the period is accounted for. The predictor variables are non-argument NPs, WH-words, free relative subordinate clauses, untensed auxiliary verbs, floated quantifiers, negation, and degree complement subordinate clauses. The regression equation was able to date about two thirds of texts within a range of 50 years deviation from the date of manuscript. The equation was verified by using independent texts in ME. The result indicates that syntactic development can be quantified effectively by internal information.