- No more Topic Modeling
- No need Training
- No more Machine Learning but Human-like Reading
- Get the Insights of Text Big and Small
KeypartX: a graph-based approach to represent perception (text in general) by key parts of speech. KeypartX solved the coherence crux that current topic modeling algorithms are trying to deal with but failed. KeypartX extracts the topics from text corpus syntactically, semantically and pragmatically instead of a meaningless combination of words from topic modeling.
KeypartX Vs Topic Modeling results from the following text:
“Thai food was great we loved it. Thiland also has beautiful beach resorts, we will come to Thailand again👍”
- KeypartX Result
- Topic Modeling Result
['food','thailand','resort','great','love', 'beautiful']
if need coreferee:
pip install keypartx[coreferee_spacy]
#!pip install keypartx[crosslingual-coreference_spacy] # a alternative coreference package
python3 -m coreferee install en
python -m spacy download en_core_web_lg
else:
pip install spacy
pip install keypartx
python -m spacy download en_core_web_lgFor an in-depth overview of the features of KeypartX you can check the Documents or you can follow along with one of the examples as follows:
| Name | Link |
|---|---|
| KeypartX Quick Start |  |
| KeypartX with Real Example | |
| KeypartX VS Topic Modelling | |
| KeypartX Network Comparison | |
- 1 NLP Target
Original sentence: """Thai food was great,delicousr and not expensive, we loved it. We visited 3 beach resorts, they are higly recommened... We had "Fire-Vodka" !!!"""
- 2 Keyparts Wordclouds
The following wordclouds are generated from a real example of corpus comprised of reviews by those who visit Thailand.
- 3 Community and Gray Perceptual Unit Networks
To cite the KeypartX paper, please use the following bibtex reference:
@article{pengyang2022keypartx,
title={KeypartX: Graph-based Perception (Text) Representation},
author={Peng, Yang},
journal={arXiv preprint arXiv:2209.11844},
year={2022}
}