Myanmar Compound Word Errors Detection and Suggestion Generation

Abstract

In Myanmar Language, Pronunciation and orthography has differences because spelling is often not an accurate reflection of pronunciation. Pronunciation of a word may lead to misspelling. In this paper, we present Myanmar Compound misused words errors detection and suggestion generation system. In this paper, we propose Myanmar compound misused word error detection algorithm to apply in Myanmar spell checker. After detecting Myanmar compound misused words errors, we provide suggestion list, which implemented with two methods: Cosine Similarity and Levenshtein Distance Algorithm. To evaluate the efficiency of the system, we tested with various types of Myanmar sentences which contain various types of spell errors. According to the evaluation results, our proposed system achieves promising accuracy (over 90 %) for Myanmar compound misused words errors detection. By analyzing these two results, we found that Levenshtein Distance Algorithm can provide better relevant suggestion list.