Hy-NLI: a Hybrid System for Natural Language Inference
Hy-NLI is a hybrid system for NLI. It computes inference with a deep learning (DL) and a symbolic approach and then its hybrid component determines which of the two labels should be trusted for a given pair. For the DL approach, it currently uses the BERT (Devlin et al, 2018) model, which is further fine-tuned on the SICK (Marelli et al, 2014) corpus. The symbolic approach computes inference based on a version of Natural Logic (Valencia, 1991; MacCartney, 2009) and on the Graphical Knowledge Representation (Kalouli et Crouch, 2018). The hybrid classifier is an MLP trained model. Please find more details in our paper:
Kalouli, A.-L., R. Crouch and V. de Paiva. 2020. Hy-NLI: a Hybrid system for Natural Language Inference. In Proceedings of COLING 2020.
Note that Hy-NLI targets performance, rather than explainability. If you are interested in the explainability of our system, check out our demo on XplaiNLI
The source code of Hy-NLI is publicly available on github .
Enter a sentence pair below:
You can also choose from the given examples:
- Hy-NLI : a Hybrid system for state-of-the-art Natural Language Inference Kalouli, A.-L. 2021. Dissertation. Konstanz: University of Konstanz
- Explainable Natural Language Inference through Visual Analytics . Kalouli, A.-L., R. Sevastjanova, R. Crouch, V. de Paiva and M. El-Assady. 2020. In Proceedings of the COLING 2020 System Demonstrations.
- Hy-NLI: a Hybrid system for Natural Language Inference . Kalouli, A.-L., R. Crouch and V. de Paiva. 2020. In Proceedings of COLING 2020.
- GKR: Bridging the gap between symbolic/structural and distributional meaning representations Kalouli, A.-L., R. Crouch and V. de Paiva. 2019, 1st International Workshop on Designing Meaning Representations (DMR) @ACL 2019.
- GKR: the Graphical Knowledge Representation for semantic parsing Kalouli, A.-L. and R. Crouch. 2018. SEMBEaR @NAACL 2018.
- Named Graphs for Semantic Representations Crouch, R. and A.-L. Kalouli. 2018. *SEM 2018.
- Graph Knowledge Representations for SICK Kalouli, A.-L., R. Crouch, V. de Paiva and L. Real. 2018. 5th Workshop on Natural Language and Computer Science @FLoC 2018
- Katerina: katerina (dot) kalouli (at) hotmail (dot) com (check out my homepage for more similar projects)
- Dick: dick (dot) crouch (at) gmail (dot)com
- Valeria: valeria (dot) depaiva (at) gmail (dot) com