Researcher Utilizes Machine Learning to Digitize Ancient Texts from Indus Civilization
The civilization of Indus River Valley is renowned as one of the oldest in world history, alongside Mesopotamia and Egypt. Spanning what is now Pakistan and India from 3300 BCE, this civilization possessed remarkable uniformity in weights and measures, skilled artisans, an intricate trade system, and over 500 symbols for communication.
A longstanding debate among scholars revolves around whether these symbols represent a language or pictograms. Debasis Mitra, a computer science professor, has delved into this mystery with a unique project funded by the National Endowment for the Humanities. His initiative, titled Ancient Script Digitization and Archival (ASDA) of Indus Valley Artifacts using Deep Learning, aims to decode the ancient script.
Mitra, aided by graduate student Deva Atturu, leverages machine learning to analyze over 1,000 images of Indus seals. Their automated script recognition system achieves an impressive 88% success rate in detecting graphemes, despite challenges posed by limited and occasionally distorted data.
Collaborating with experts and students from the Indian Statistical Institute, Mitra’s team is making strides in digitizing motifs and graphemes. The ultimate goal is to develop a system enabling archaeologists to photograph texts or symbols in the field for instant digitization.
By shedding light on one of India’s ancient civilizations, Mitra feels a strong sense of purpose in his work. His efforts not only benefit academia but also foster cultural exchange, as demonstrated by the interest sparked among American students.
This groundbreaking research signals a significant step towards unraveling the mysteries of the Indus civilization, bridging the gap between the past and the present.
