Annual Reports Assessment Dataset

Annual reports Assessment Dataset 

This dataset will help investors, merchant bankers, credit rating agencies, and the community of equity research analysts explore annual reports in a more automated way, saving them time.

Following Sub Dataset(s) are there :

a) pdf and corresponding OCR text of 100 Indian annual reports
These 100 annual reports are for the 100 largest companies listed on the Bombay Stock Exchange.
The total number of words in OCRed text is 12.25 million.

b) A Few Examples of Sentences with Corresponding Classes
The author defined 16 widely used topics used in the investment community as classes like:

Accounting Standards

Accounting for Revenue Recognition

Corporate Social Responsbility

Credit Ratings

Diversity Equity and Inclusion

Electronic Voting   

Environment and Sustainability

Hedging Strategy

Intellectual Property Infringement Risk

Litigation Risk

Order Book   

Related Party Transaction

Remuneration

Research and Development

Talent Management

Whistle Blower Policy

These classes should help generate ideas and investment decisions, as well as identify red flags and early warning signs of trouble when everything appears to be proceeding smoothly. 

ABOUT DATA ::

"scrips.json" is a json with name of companies
    "SC_CODE" is BSE Scrip Id 
    "SC_NAME" is Listed Companies Name
    "NET_TURNOV" is Turnover on the day of consideration

"source_pdf" is folder containing both PDF and OCR Output from Tesseract
    "raw_pdf.zip" contains raw PDF and it can be used to try another OCR.
    "ocr.zip" contains json file (annual_report_content.json) containing OCR text for each pdf.
    "annual_report_content.json" is an array of 100 elements and each element is having two keys "file_name" and "content"

"classif_data_rank_freezed.json" is used for evaluation of results
    contains "sentence" and corresponding "class"