An experiment in computationally writing and compiling academic papers based around machine learning.
Currently working on:
- Automated scraping of academic articles
Completed:
Flow:
- User enters a topic string. Ex) Canadian macro economic policy (Possibly add subtopics to refine search)
- InstantPaper references online journal and academic sources and delivers a rating (0-100) on the availability of information on the chosen topic
- Using a PageRank-esq algorithm, create a hierachy of academic papers based on imporance. Scrape content from top 100 papers on topics and using machine learning find commonly cited argument
- Extrapelate 3-4 top 'themes' for each topic. These will act as individual body paragraphs
- Rank indexed papers context and relevence to themes and add 5-6 'points' to each theme
- Intelligently gather author information so we can get sentences like: "Scottish political economist Adam Smith wrote in the 'Wealth of Nations' that "arument/quote..." (Could also cross reference this information w/ Wikipedia)
- Allow the user to select reference and citation format
- Have a few other user settings: max word count, limit citations to 'n' length, user name, class, professor...etc.
- Output as PDF/DOC/LaTeX