SentenceAx

The Openie6 (O6) software splits complex or compound sentences into simple ones. Simple sentences are essentially the same as the triples (subject, relationship, object) which, when visualized as a directed or undirected graph, is called a “knowledge graph”. Sentence splitting is also a necessary step in doing causal DAG extraction from text (causal DEFT), as is done by my software Mappa Mundi.

SentenceAx (Sax) is a complete rewrite, from stem to stern, of O6.

SentenceAx is a fine-tuning of BERT written with PyTorch and Lightning.

SentenceAx is a stand-alone app, but, just like the SCuMpy app, it's also a key component of the Mappa Mundi Project which started with the MappaMundi app. The SentenceAx app, SCuMpy app and Mappa Mundi app were written by www.ar-tiste.xyz

• Mappa Mundi Project
  • MappaMundi app
  • SCuMpy app
  • SentenceAx (this repo)
  • Causal Fitbit (coming soon)

Input Dataset and Weights

Due to a < 50MB per file limitation at Github, the input datasets and weights will be located at HuggingFace, at this location.

Download and unzip the zipped files at that loccation. Use them to replace the directories called input_files and weights in your local copy of SentenceAx.

Documentation

• My blog posts about SentenceAx

• Chapter entitled ``Sentence Splitting with SentenceAx" in my free open source book Bayesuvius

• Appendix to the Chapter in Bayesuvius