PBL_tool: keyword search

Idea:

We are using the API PubMed to get access to published papers and the textmining is done with regex. The user is will be able to use the tool on a webpage, we created with html. For the Layout we use CSS. All html files are resided in the templates folder.

Workflow:

• For now the tool can be executed by running the start.py script and opening the link that is given in the output terminal.
• By opening the link, the input.html file is opened. Here the user input can be parsed, we allow a keyword, a number for the amount of papers that is wished to be returned, as well as filter options (filter options are not included into relevance score yet).
• When the submit button is pressed, the input_post function in the script start.py is executed. The user input is requested from the input.html file and the runner.py file is being called and gets parsed the user input.
• The runner.py file opens PubMed and returns the wished amount of papers, that contain the keyworld, for each paper an instance of the class Paper in the script paper.py is created.
• TODO: Make the scanning Progress faster! It's taking AAAAGES!! ANY IDEAS ARE WELCOME
• You can run the runner.py with an argument parser for testing purposes. For more info on the parser execute runner.py --help
• In the Paper object we are saving the title, a list of the authors, the publishing date, and the PubMed Id for each paper, as well as a count of the occuring keyword.
• When the paper is created, the abstracts and titles are scanned for the keyword using regex and the relevance score is computed.
• TODO: IMPROVE the relevance score
• The list of papers is then sorted by score
• The script runner.py creates a list of all the Paper-objects and parses it to the script dataframe.py, which creates a dataframe for a pandas table.
• This table is returned to the function input_post in the script start.py, which convertes it into an html table and returns the output.html tamplate, which gets parsed the html table.
• On the webpage, the output.py is shown, displaying a the table with the search results. It is possible to retun to the input.html at any time, clicking on input again.
• Here it is possible to resort the returned papers by date, author, title and score. When submitted, the function sorting in start.py requests the radio input and resorts the globally saved dataframe of papers (called list) by column.
• TODO: links to PubMed pages for each resulting Paper.