Analyzing Word Error Rate (WER) over Part of Speech (POS) Tagging
The task of the part of speech (POS) tagging has been applied by various ways. This paper represents POS tagging by measuring the minimum edit distance between English and Myanmar phonetic based on English language. The input of this task is voice that is automatically generated the text output by android speech recognizer. The proposed system is originally defined as the minimum number of deletions, insertions, and substitutions required for transforming one word into another using Levenshtein Distance (LD) algorithm. This paper presents a framework for extraction of linguistics details from standard words error rates WER over ten basics POS classes. It also carried out a detailed analysis of inflectional errors and missing words error between English and Myanmar phonetic. Tested on the 1000 corpus on English and this system is developing on android platform using eclipse workbench.
Keywords- Word Error Rate (WER), POS Tagging, Speech Recognition.