Abstract:
A grammar checker checks each sentence in a text whether it conforms to
the grammar, and in case it finds conflicts in the conformity, it would
suggest alternatives to the word(s) that causes the conflict. An approach to
detecting and correcting grammatical mistakes due to subject and finiteverb
disagreement with regard to person, number and/or gender and due to
disagreement in tense aspects in Tamil sentences is proposed in this paper.
Hierarchical POS tags of words are used to detect the grammatical mistakes
and two sets of Tamil grammar rules are used to generate suggestions. First
set of Tamil grammar rules is used to generate suggestions to correct
mistakes due to subject and finite-verb disagreement, and the other one is
used to generate suggestions for tense aspect disagreement. These grammar
rules have been defined in the form of Python conditional statements based
on standard Tamil grammar. Alternative suggestions for both subject and
finite-verb are given for personal pronouns of third person and alternative
suggestion is given for finite-verb for all types of subjects, not just for
pronouns. Moreover, a list of tense-aspects words mapped to tenses is
prepared and used for giving suggestions in case of tense aspect
disagreement. Test results show that the proposed grammatical mistake
detection and correction system performs well. For 100 testing sentences
with all kinds of grammatical mistakes mentioned above, it gives 136
suggestions, a Scholar in Tamil language approved 132 suggestions as most
suited and the remaining 4 are as acceptable.