Automated Analysis through Natural Language Processing of DGMS Fatality Reports on Indian Non-Coal Mines

Mining activities pose a major threat to the safety of mine workers due to the inherent risky nature of the job. To realize the objective of preventing fatal and non-fatal injuries, it is imperative for a mine management to analyze site specific mine safety data. This paper is the first attempt to automate the analysis of Directorate General Mines Safety (DGMS) fatality reports on Indian non-coal mines. It is done by text mining and natural language processing (NLP) techniques. The proposed method sidesteps multiple man hours of scrutinizing each report and technical expertise required in analyzing mining accidents. The authors have taken 6 years' worth of data ranging from 2010–2015 published by DGMS. The attempt revealed valuable insights into the causes of injuries such as accidents occurrence is highest in states of Rajasthan followed by Andhra Pradesh. The leading indicators of fatality included falling from heights, age between 28 to 32 years, worker class of ‘mazdoors’, and shift timings between 10 AM and 2 PM. Furthermore, results show accidents happened mostly in small scale stone mines. This will be a step forward for the Indian mining industry to extract meaningful results from more such text data quickly and accurately to make better decisions. The methodology can be extended to the Indian coal mining industry easily.