Decomposable neural attention for natural language inference in spaCy and Kerasgithub.com/explosion3 pointssyllogism10 years ago