Language models

WinkNLP comes with pre-trained language models with gzipped sizes starting from ~890KB (expanded sizes less than 3MB) onwards:

wink-eng-lite-model

Size	~890KB gzipped.
Usage	Node.js server side. Recommended for Node.js version 12 and 14.
Pipeline	Sentence Boundary Detection (`sbd`), Negation Handling (`negation`), Sentiment Analysis (`sentiment`), Named Entity Recognition (`ner`), Part-of-speech tagging (`pos`), Custom Entity Recognition (`cer`).
Model specific helpers	`its.pos` — part-of-speech tag. `its.lemma` — pos specific lemma. `its.stem` — based on Porter Stemmer version 2. `its.negationFlag` — true if a token is negated. `its.stopWordFlag` — true if a token is a stop word. `its.type, its.detail, its.span` — for `.entities.out()` and `.customEntities.out() methods.` `its.readabilityStats` — includes Flesch Reading Ease Score (fres), list complex words and their count, reading time in mins & seconds, sentiment score and more.
Word Vectors	NA
License	MIT

Size	~1MB gzipped.
Usage	Broswer and Node.js both. Recommended for Node.js version 16 and 18.
Pipeline	Sentence Boundary Detection (`sbd`), Negation Handling (`negation`), Sentiment Analysis (`sentiment`), Named Entity Recognition (`ner`), Part-of-speech tagging (`pos`), Custom Entity Recognition (`cer`).
Model specific helpers	`its.pos` — part-of-speech tag. `its.lemma` — pos specific lemma. `its.stem` — based on Porter Stemmer version 2. `its.negationFlag` — true if a token is negated. `its.stopWordFlag` — true if a token is a stop word. `its.type, its.detail, its.span` — for `.entities.out()` and `.customEntities.out() methods.` `its.readabilityStats` — includes Flesch Reading Ease Score (fres), list complex words and their count, reading time in mins & seconds, sentiment score and more.
Word Vectors	NA
License	MIT