README.md

megathon2019

Fill in the blank question generation from textbooks. Made for Megathon 2019 @ IIIT Hyderabad

Pre-requisites

1. Removal of paragraphs from the text - Bhavyajeet / Jai
   • (i) and other bullet points
   • Split by \n\n
   • Remove Fig lines
   • 1.2.x are headings
   • Split by headings
   • Remove numerical lines (Later)
2. Removal of sentences from the paragraphs - Pranav / Jai
   • Using Simple split by "."
3. Grouping of the sentence and paras and relevant things such as Subsection Title, etc. - Pranav / Jai
   • Find last heading and subheading of the paragraph

Sentence Selection

1. Feature Extraction
   • Sentence Similarity - Ahish --> DONE
   • Common words to the title - Bhavyajeet --> DONE
   • Similar to summary - Bhavyajeet
   • Superlative - Pranav --> DONE
   • Abbreviations - Pranav --> DONE
   • Number (Y/N) - Pranav --> DONE
   • Certain Keyword (later)
   • Sentence Position in Para - Already provided by preprocessing
   • Length of sentence - Already provided by preprocessing
   • Number of Nouns / Length - Pranav --> DONE
   • Number of Pronouns / Length - Pranav --> DONE
   • Discourse Connective - Jai
2. Neural Network Design - Ahish / Jai

Keywords Selection

1. Feature Extraction
   • Parse Tree Height
   • Similarity to Title
   • Other Sequence Labelling things
2. Neural Network Design