The code for KRISPER (extension of KRISP to handle ambiguous data, see the paper)
can be found here.
If the pairings of natural language sentences and meaning representations is not
ambiguous in the training data then it reduces to KRISP (see the paper).
For questions contact: Rohit Kate