|
--- |
|
license: lgpl-3.0 |
|
task_categories: |
|
- summarization |
|
- text-classification |
|
tags: |
|
- code |
|
configs: |
|
- config_name: default |
|
data_files: |
|
- "input_data.zip" |
|
- "training_data_ex_epoch%3D20_tune_epoch_acc%3D0.5189.zip" |
|
--- |
|
|
|
# Welcome to the HuggingFace repo for SentenceAx |
|
|
|
![SentenceAx](pics/sentence_ax_logo.jpg) |
|
![SentenceAx Bayesian Network](pics/sentence-ax-bnet.jpg) |
|
|
|
|
|
The [Openie6 (O6) software](https://github.com/dair-iitd/openie6) |
|
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](https://github.com/rrtucci/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 it is also a vital part of the Mappa Mundi |
|
project for doing causal AI/ML and causal inference. |
|
|
|
This repo contains various large data files (input data, weights) |
|
necessary for training or generated by training Sax. |
|
The Python source code for Sax can be found at |
|
[its GitHub repo](https://github.com/rrtucci/SentenceAx). |
|
|
|
|