an efficient method to add chunker rules in persian to english rule-based apertium machine translation system

Rule-based machine translation (rbmt) captures linguistic information about the source and target languages. this information is retrieved from (bilingual) dictionaries and grammar rules. this paper proposes an active learning (al) method to grow structural transfer rules at the chunker level. to this end, two sets of experiments are performed based on two types of sentences extracted from mizan english-persian parallel corpus which are selected manually and randomly. the results show adding newly written chunker rules to the transformation file using pool-based al technique improves translation system more compared to a random chunker rule selection baseline.