2

I am working on Natural Language Generation project. Now I have bags of keywords and I am trying to generate sentence of pattern Subject+verb+object. Is there any tool or package or library to get sentences when i give bag of subject,bag of verb and bag of object? Example: subject=[teacher,student] verb=[teach] object=[book] answer: teacher teaching student from books.

Shraddha Deshmukh
  • 41
  • 1
  • 6
  • 1
    Train a language model on large collection of english text. Then create a list of candidates by going through permutations of words you are given. Use your language model to assign each candidate a probability. Most probably the likeliest candidate is also a correct english sentence... – user3639557 Apr 20 '16 at 10:55

1 Answers1

2

Assuming the words in your bag of words are tagged with word categories such as verb or noun, you could use a realiser such as SimpleNLG.

Effectively you would write a bunch of sentence specification templates and a script to look into your bag of words and use your templates.

For example, for your sentence "teacher teaching students from books" you could have the following sentence specification template:

  • Subject = a noun, e.g. teacher
  • Verb = a verb, e.g. teach (form = present participle)
  • Object = a noun, e.g. student (number = plural)
  • Postmodifier = { preposition = from , noun = books (number = plural) }

Note this approach will give you morphologically and syntacticly valid sentences, even though some may sound funny such as "books teaching teachers from students".

Rodrigo
  • 321
  • 2
  • 11