Ancient Indus seals have hidden secrets and over centuries researchers have tried to understand the texts. Lately, advanced technology is being employed to decipher the same. Utilizing automated script recognition (ASR) system, more than 1,000 photographs of the Indus seals have been analyzed. The ASR system extracts coded sequences of symbols from the images of seals and the success rate is 88% in detecting the symbols.

Amid excitements, the process is facing some challenges as well. Unlike typical machine learning projects that benefit from vast amounts of data, it is facing scarcity and sometimes even distorted data. Researcher Mitra, who is involved in the project, said the challenges are usual for those who encounter in medical imaging.

Mitra has employed various machine learning techniques to overcome the challenges. He has collaborated with archaeologists at conferences such as the Annual Conference of South Asia for the purpose. These interactions provide valuable data inputs for the project.

Moreover, Mitra is taking the help of students at the Indian Statistical Institute to digitize motifs and symbols from the seals. The primary focus is to create a database that can be accessed by archaeologists in the field. The database allows researchers in taking photo of a seal or symbols using a smartphone. This is thereafter routed into the database for analysis.

Mitra’s motivation for the project is deeply personal and it is basically focused in shedding light on one of the great civilizations in the history of India. He emphasized the importance of the project for himself and also for Indian students who share same enthusiasm. However, the project has also attracted interest from American students with a strong fascination for India. This means the project has global significance.