Skip to content
Go to file

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

Relation Schema Induction

Given a set of documents from a specific domain (e.g., medical research journals), how do we automatically build a Knowledge Graph (KG) for that domain? Automatic identification of relations and their schemas, i.e., type signature of arguments of relations (e.g., undergo(Patient, Surgery)), is an important first step towards this goal. We refer to this problem as Relation Schema Induction (RSI).


SICTF solves the problem of Relation Schema Induction, defined below. Given the surface triples as a Tensor, Noun Phrase Side Information as a Matrix and Relation similarity as a Matrix, SICTF performs a joint factorization.


To get the project's source code, clone the github repository:

$ git clone

Install VirtualEnv in python3 using the following (optional):

$ [sudo] apt-get install python3-pip
$ [sudo] pip3 install virtualenv 

Create and activate your virtual environment (optional):

$ virtualenv venv
$ source venv/bin/activate

Install all the required packages (dependencies):

$ pip3 install -r requirements.txt


Please install the following dependencies for SCITF to run: numpy, scipy, scikit-learn, joblib, python3.3 or above.


To use sictf, you can start with these following samples commands:

$ python3 ../Data/Medline/ Triples.txt NpSideInfo.txt RelSim.txt
$ python3 outfolder 10 10 1 3 3 20 False 0.1 0.1 0.1 0.1 0.1


Help for running the setup.

Place the triples and side information in a folder at the same level as *.py scripts

Run the Following commands sequentially afterwards ( with -h flag for help)


usage: [-h] inputFolder fname sideInfoMatrix relSim

positional arguments:

  • inputFolder -- Enter the name of folder which containts the data file
  • fname -- Enter name of the file that is the source of triplets
  • sideInfoMatrix -- Enter the name the file that contains side information
  • relSim -- Enter the name of the file that contains Verb phrase Similarity information

optional arguments:

-h, --help show this help message and exit


python3 -h

usage: [-h] outputFolderName minRank maxRank step Top TopRC maxIters fitFlag lA lR lV Ws Wrs

positional arguments:

  • outputFolderName -- Enter the name of the Output Folder :
  • minRank -- Enter Rank for min Rescal Decomposition :
  • maxRank -- Enter Rank for max Rescal Decomposition :
  • step -- Enter step for Rank (Please ensure minRank +n*step = maxRank):
  • Top -- Enter cut-off for top Entities:
  • TopRC -- Enter cut-off for top RelMatrix Entries:
  • maxIters -- Enter Max no. of iterations:
  • fitFlag -- Enter True is fit computation desired, False if not, None for uncertainity (Advised to keep as False)
  • lA -- Enter Lambda A
  • lR -- Enter Lambda R
  • lV -- Enter Lambda V
  • Ws -- Enter Side Info term weight
  • Wrs -- Enter Relation Similarity term weight

optional arguments:

  • -h, --help show this help message and exit


[1] Madhav Nimishakavi, Uday Singh Saini and Partha Talukdar. Relation Schema Induction using Tensor Factorization with Side Information. November 2016. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing (EMNLP 2016).


Relation Schema Induction using SICTF




No releases published


No packages published


You can’t perform that action at this time.