ALGORITMA PENENTUAN KELAS KATA PADA KATA-KATA YANG MEMILIKI SIFAT AMBIGU DALAM PART-OF-SPEECH TAGGING BAHASA INDONESIA DENGAN METODE RULE BASED

Indonesian has a fundamental role in the communication of Indonesian society. However, there are problems that arise In the use of Indonesian, that is about ambiguity. Ambiguity is a character or thing that has understanding or meaningful more than one. In the Natural Language Processing study, Part of Speech (POS) tagging plays a role in the determination of word class labels. This study aims to know the word class determinitation algorithm of ambiguous words in Indonesian POS tagging with rule based method. Rule based is one of the methods in the expert system that uses the if-then rule to present its knowledge. This research belongs to the basic research with experimental research design, with the following stages: 1) knowledge inventory, 2) making algorithms, 3) implementation, 4) Testing, 5) Analysis, 6) Conclusions, and 7) Preparation of reports. This research uses explanatory framework with analytical method. Explanatory is a frame of mind that explains the properties of the study object. The analysis method used in this study is descriptive by describing the reality of situations or events coherently and systematically. The result of this research is a word class determination algorithm resulted from several stages as follows: Understand Indonesian sentence patterns, define reasoning strategies, compile ambiguous words, create patterns of ambiguity, make rules, consult with experts, and ensure algorithm. The paths in the algorithm include: 1) input sentence, 2) segmentation of the sentence being words, 3) check ambiguity, 4) if there’s not ambiguous, then do the word labeling, 6) if the words are ambiguous, then test with syntactic rules, 7) show the results of word class labeling. The algorithm that has been prepared is implemented in Indonesian POS tagging application developed by learning group of NLP UIN Sunan kalijaga. This application is developed with PHP programming language and MySQL DBMS. The test results that the implementation of this algorithm successfully provides 92 labeling of 100 tested words (92%). The results of the implementation are influenced by the availability of rules in handling the ambiguity of word classes in various forms, such as the algorithm still can not handle two adjoining ambiguous words and ambigous words that are at the beginning of the sentence