Knowledge based Approach for English-Malayalam Parallel Corpus Generation

摘要

Objective: This paper aims in providing an overview about a part of Natural Language Generation – Parallel sentence generation which involves the generation of the English sentence as well as its Malayalam translated version. Methods/Analysis: A template based sentence generator approach is followed here. A system is proposed which takes input from a manually created bilingual dictionary and fills the slots in the template for parallel sentence generation. Finding: Using the proposed method, we have generated a total of 25,208 parallel sentences. This can be used in bilingual Machine Translation dictionary. Application/Improvement: In the proposed case use only four templates but by increasing the number of templates and by updating the dictionary, we can increase the size of the parallel corpus that can be generated.