Skip to content
Permalink
Branch: master
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
70 lines (46 sloc) 3.71 KB

TutorialBank: A Manually-Collected Corpus for Prerequisite Chains, Survey Extraction and Resource Recommendation

A corpus designed to facilitate NLP education and promote research in the areas of generating surveys for scientific topics, prerequisite chains of learning, resource recommendations, among others. Also check out our search engine!

Folder Structure:

  • /data/

      - resources.sql -- resource dump with meta-data for our corpus
      - resources.csv -- csv version of the above file
      - taxonomy.csv -- contains the topic taxonomy for our corpus
      - topics_to_resources.csv -- contains the 200 topics separate from the taxonomy and their reading lists
      - images_urls.csv -- contains filtered images from resources
      - prerequisite_topics.csv -- contains the 200 topics above along with their wikipedia page
      - prerequisite_annotations.csv -- prerequisite annotations
               -- If prereq_relation ==1, then source_topic_id is a prerequisite of target_topic_id 
               -- both source_topic_id and target_topic_id are foreign keys for prereq_id in prerequisite_topics.csv
      - survey_annotations.csv -- annotations for survey extraction with content cards where a score of -1 means DROP, 
               and 0-2 is the score as in the paper. (only includes PDF cards)
      - survey_annotations_pdf_and_html.csv -- the same as above, except includes both PDFs and HTML pages.
      - projects.csv -- abstract+title pairs from the  resource recommendation experiments
      - doc2vec/ -- a folder which contains the output recommendations from our Doc2Vec model for 10 title+abstract pairs
      - lda/ -- a folder which contains the output recommendations from our Doc2Vec model for 10 title+abstract pairs
    
  • /src/

      - download_all.py -- downloads the corpus, creating folders /data/files and /data/url/
      - get_cards.py -- splits resources into cards for survey extraction, creating the folder /data/cards/
      - kappa.py -- computes inter-annotator agreement scores
      - split-md/ -- folder containing utils for splitting markdown files
    
  • /annotation_process/

      - annotation_guidelines.pdf -- contains guidelines from our annotation process
      - prerequisite_interface.png -- picture of the prerequisite annotation interface
      - survey_extraction_interface.png  -- picture of the survey extraction annotation interface
      - resource_recommendation.png -- picture of our current resource recommendation system
    

Usage:

Download all the resources from the corpus:

Install lynx (follow these instructions or these for OSX).

python src/download_all.py

Create the cards for survey generation (must be run after downloading resources). For this part you need to go to https://pdfbox.apache.org/download.cgi#18x and download pdfbox-app-1.8.13.jar and place it in the folder /src/

python src/get_cards.py

Access command line tool that returns prerequisites/valuable cards/images for a particular topic.

Get a description of the options:

./interface -h

Corpus Statistics (at the time of paper submission):

1480 files (non-html pages)
4184 urls
5664 all

There are now over 6,300 resources in our corpus and search engine.

pedagogical annotations kappa: 0.69
prerequisite annotations kappa: 0.30
survey annotations kappa: 0.33

You can’t perform that action at this time.