Awesome
comment analysis experiment
Enter comments -> Fix typos -> Get sentiments, semantic similarity graph, and comment groups/classes.
Live demo: https://comment-analysis.surge.sh (yarn deploy
after running yarn dev
)
Local demo: yarn dev
(if you run into Error: Expected content key ### to exist
, try running rm -rf .parcel-cache
first)
references and things this projects builds on
USE (Universal Sentence Encoder) (see my text-similarity-test)
umap-js (UMAP vs t-SNE vs SNE)
nlp.js (note: SpellCheck
currently seems incompatible with one of parcel's transformers)
(otherwise just look at package.json)
typo fix suggestions under input box to fix input box
- https://github.com/axa-group/nlp.js/blob/master/docs/v4/similarity.md#spellcheck (seems to have compatibility issue with parcel)
- https://github.com/axa-group/nlp.js/blob/master/docs/v4/similarity.md#spellcheck-trained-with-words-trained-from-a-text
- https://naturalnode.github.io/natural/spellcheck.html (doesn't come with a dictionary)
- https://github.com/wooorm/dictionaries#example-use-with-nspell (provides a dictionary like
dictionary-en
ordictionary-en-ca
or more languages, but you'll likely have to manually copy and include some specific code)
sentiments above chart
- https://github.com/axa-group/nlp.js/blob/master/docs/v3/sentiment-analysis.md
- https://naturalnode.github.io/natural/sentiment_analysis.html
knn under chart
- KNN for automatic grouping, maybe with tweakable k input box: https://github.com/tensorflow/tfjs-models/blob/master/knn-classifier/README.md