Effect of Statistical POS Tagger on Syntactic Analysis of Punjabi Sentences

Objectives: In this research article, author has explored the effect of statistics based part of speech tagger on the syntactic analysis of Punjabi sentences. Methods/Statistical Analysis: To study the effect of statistical POS tagger on the syntactic analysis of Punjabi sentence, author performed two experiments; first a rule based POS tagger is used for syntactic analysis and second this rule based POS tagger is replaced with HMM based statistical POS tagger. An annotated corpus of 20,000 words has been used to train the HMM based POS tagger. Findings: The system is tested on three types of errors; first subject/object and verb agreement error second noun and modifier agreement (in attributed form) error and third modifier and noun agreement error. On using HMM based POS tagger, the system shows a precision of 80.67 for subject/object and verb agreement error whereas on using rule based POS tagger the system shows a precision of 72.81. Similarly for noun and modifier agreement (in attributed form) error, author claims a precision of 82.45 on using HMM based tagger whereas on using rule based tagger, the precision is 76.00. And in case of modifier and noun agreement error, a precision of 97.56 is claimed by the author by using HMM based tagger which was 95.45 when rule based POS tagger is used. Application/Improvements: The result indicates that the grammar checker performs better when rule based POS tagger is replaced with statistics based POS tagger.