Part Of Speech Tagger For Afaan Oromo Language Using Transformational Error Driven Learning (tel) Approach

purpose of this research is to develop part-of-speech tagger for Afaan Oromo usingrnTransformational Error driven Learning (TEL) approach and compare it with other approach.rnMost natural language processing systems use part-of-speech (POS) tagger as a one of theirrncomponent in their system. Specially, it is very significant for developing parser, machinerntranslator, speech recognizer and search engines.rnAfaan Oromo literatures on grammar and morphology are reviewed to understand nature of thernlanguage and also to identify possible tagsets. Based on this, 18 tagsets are identified and used onrn223 sentences (1708 words) for the experiment.rnThe study customized Brill transformational error driven learning tagger for Afaan Oromo. Somerntemplate in the original Brill tagger was modified to fit Afaan Oromo morphological nature. Afterrntraining data is analyzed for its appropriateness using learning curve analysis, the study used 10-rnfold validation method for the experiment. Moreover experiment was conducted to determine thernpercentage of training data for contextual and lexical rule learner. Best accuracy of the tagger wasrnachieved when contextual rule learner training data is 35% and lexical rule learning data isrn65%.This shows the morphological rule dominance over contextual rule for the language.rnAfter modification on the templates of the Brill’s tagger about 2.44% improvements over thernoriginal Brill tagger was achieved. This means 80.08% accuracy of the tagger was achieved inrnmodifying the templates where the accuracy of the original tagger is 77.64%. Error of the modifiedrntagger was also analyzed for further improvements using confusion matrix for the tagger.rnThe result obtained in both original Brill tagger and modified Brill tagger is compared withrnHidden Markov Model approach (bigram and unigram approach).The comparison shows that Brillrntagger is by far better than Hidden Markov Model in all the cases for Afaan Oromo i.e HiddenrnMarkov Model accuracy for bigram approach is 70.63% and for unigram 68.08% whereas that ofrnoriginal Brill tagger without modification is 77.64 and 80.08% for modified Brill tagger.rnKeywords: Natural Language processing, parts of speech tagging, Brill Tagger,rnTransformational Error driven Learning, Hidden Markov Model, Bigram, N-Gram.

Get Full Work

Report copyright infringement or plagiarism