Generating Factoid Questions from a given paragraph or article.
Done as part of NLP course - Monsoon 2018, IIIT Hyderabad.
Link to the primary details about this project.
Update: Completed Question Generation Part of this project.
Instructions:
- Download and install StanfordCoreNLP.
- Install wordnet with
nltk.download('wordnet')in your python shell. - Also download and install spacy along with its English language model.
(from corenlp folder)
java -Xmx2g -cp "*" edu.stanford.nlp.pipeline.StanfordCoreNLPServer -port 9000 -timeout 15000
(from this repo)
bash main.sh "---sentence---"
Examples:
bash main.sh "I saw an eagle"
OUTPUT:
Did I saw an eagle?
Did I see an eagle?
Who saw an eagle?
What did I saw?
What did I see?
bash main.sh "Peter is studying in NY"
OUTPUT:
Is Peter studying in NY?
Who is studying in NY?
Where is Peter studying?
bash main.sh "I have made a huge mistake"
OUTPUT:
Have I made a huge mistake?
Who has made a huge mistake?
What have I made?
bash main.sh "John has seen Mary"
OUTPUT:
Has John seen Mary?
Who has seen Mary?
Who has John seen?
bash main.sh "John's car is fast"
OUTPUT:
Is John 's car fast?
Whose car is fast?
bash main.sh "John read over 200 comic books"
OUTPUT:
Did John read over 200 comic books?
Who read over 200 comic books?
How many comic books did John read?
bash main.sh "John was a nice person"
OUTPUT:
Was John a nice person?
Who was a nice person?
What did John be?
Note that in the first example, questions 1 and 4 have 'saw' as verb because they consider 'saw' in the sentence to be the act of 'sawing' and not past form of 'see'.