Digital-Alpha-ML

The ML repository for the Problem Statement - Digital Alpha SaaS Analyzer at Inter IIT Tech Meet 10.0. The deployed version of the website can be found here. The report can be found here

Files Description

1. dict-sentiment.ipynb -> For sentiment analysis(6 classes - lexicon based)

   • Input: The input of the file is specified in the temp_text variable.
   • Output: The input is passed to the get_class_counter function, which returns the sentiment dictionary containing the results.

2. finbert_inference.ipynb -> For sentiment analysis(3 classes - transformer)

   • Input: The input of the file is specified in the temp_text variable.
   • Output; The input is passed to the get_output function, which returns the sentiment dictionary.

3. mdna_extractor.ipynb -> For extracting contents(section wise)

   • Input: The input to the function is the filing_url and section_name, where the names have their usual meanings
   • Output: The output is obtained from the get_section function, which returns the desired section text

4. find_company_trends_using_lda.py -> For extracting the latest trending topics relevant to the company

   • Input: The input to the file is the company title and the number of tweets we want to extract
   • Output: The output of the file is a list of top keywords relevant to the company

5. extract_metrics_from_fillings.ipynb -> For extracting metrics from the fillings

   • Input: The inputs are:
     • api_key - for accessing the fillings using sec-api
     • url - url to the filing
     • metric - name of the metric in lowercase
     • val_type - metric data type - one of ['PERCENT', 'MONEY', 'NUMBER', 'RATIO']
     • k - window size for metric search, default = 6
     • relevant_sections - list of sections to search for the metric
   • Output: The output of the file is value of the metric extracted from the filing stored in correct_value variable

6. extract_tables.ipynb -> For extracting tables from the fillings

   • Input: The inputs are api_key for accessing the fillings using sec-api, url to the filing and the section
   • Output: The output of the file is the tables extracted from the filing stored in tables variable

7. qna_on_tables.ipynb -> For question answering on the tables

   • Input: The inputs are table and ques (a list of questions)
   • Output: The output of the file is the answers to the question based on the table

8. theme-vocab-builder.ipynb -> To build vocabulary for various sectors

   • Input: any important data file related to various sectors
   • Output: The output of the file is the vocabulary file for various sectors

9. exposure-calc.ipynb -> to calculate the exposure of a company to various sectors

   • Input: The inputs are -
     • filing.txt - sec filing of a company
     • theme.txt - vocabulary file for a specific sector
   • Output: The output of the file is the similarity score with respect to the vocabulary of a specific sector

10. generate_questions_answers.ipynb -> to generate questions and answers from the text given

    • Input: The only input is the text
    • Output: The output of the file is the generated questions and answers in the dictionary qna_dict

11. summarize_text.ipynb -> to summarize the text given

    • Input: The only input is the text
    • Output: The output of the file is the summary of the text in the variable summary

12. 10Q_parser.ipynb -> For extracting contents(section wise)

    • Input: The input to the function is the link of the filing and section number
    • Output: The output is obtained from the parse_10q_filing function, which returns the desired section text

13. find_metric.ipynb -> complete pipeline for extracting metrics from filings of a company

    • Input: The inputs are -
      • api_key - for accessing the fillings using sec-api
      • company_cik - cik of the cmopany
      • metric - name of the metric in lowercase
      • val_type - metric data type - one of ['PERCENT', 'MONEY', 'NUMBER', 'RATIO']
      • k - window size for metric search, default = 6
    • Output: The output of the file is value of the metric extracted from the filings