TIFMO (Textual Inference Forward-chaining MOdule) is an unsupervised Recognizing Textual Entailment (RTE) system based on Dependency-based Compositional Semantics (DCS) and logical inference.
This repository is a fork of https://github.com/tianran/tifmo by @tianran.
During my internship at National Institute of Informatics, Japan in 2014, I extended his work to introduce support for generalized quantifiers. The result will be published in the 28th Pacific Asia Conference on Language, Information and Computing:
Yubing Dong, Ran Tian and Yusuke Miyao. "Encoding Generalized Quantifiers in Dependency-based Compositional Semantics." Proceedings of the 28th Pacific Asia Conference on Language, Information and Computing. Phuket, Thailand, 2014. [paper][slides]
- Download the code:
git clone https://github.com/tomtung/tifmo.git - Build with
./sbt compile - Run FraCaS demo:
./sbt "run-main tifmo.demo.FraCaS input/fracas.xml" > fracas-out.txt - Check accuracy:
./AccuracyFraCaS.sh fracas-out.txt single 1 80 - Run RTE demo:
./sbt "run-main tifmo.demo.RTE input/RTE2_dev.xml" > rte2-dev-out.txt - Check accuracy:
./AccuracyRTE.sh rte2-dev-out.txt 0.4