Data Driven Approach to Brahmi OCR Error Correction and Sinhala Meaning Generation from Brahmi Character Array

Brahmi is one of ancient scripting languages used in India, South and Central Asia. Brahmi inscriptions are the main source of getting information about ancient Sri Lanka. But manual translation process is very time consuming because Brahmi alphabet, word pattern, meaning and grammar are completely differing than the current Sinhala context. As these inscriptions have been discovered under the drip ledges of caves, rocks and slabs, it is a big challenge for Archaeologists to getting the data from these inscriptions and translating into Sinhala sentences because of various noises. This research paper is containing the after process of OCR character recognition phase, which proposed and implemented an efficient system which can take recognized Brahmi character array without word boundaries as input and convert it into meaningful Sinhala sentences by overcoming the OCR errors and word identification scoring method. Full system which consist of four modules is taking inputs of Brahmi inscription images, and this research module is the third module which takes the recognised Brahmi character array from the output of second module and mainly focused on word detection, overcome OCR errors using 3 error correction methods, Brahmi language model using 2 Bigram models and 1 Trigram model and translation.