NLU: The Power of Spark NLP, the Simplicity of Python

John Snow Labs' NLU is a Python library for applying state-of-the-art text mining, directly on any dataframe, with a single line of code.

As a facade of the award-winning Spark NLP library, it comes with 1000+ of pretrained models in 100+, all production-grade, scalable, and trainable, with everything in 1 line of code.

NLU in Action

See how easy it is to use any of the thousands of models in 1 line of code, there are hundreds of tutorials and simple examples you can copy and paste into your projects to achieve State Of The Art easily.

NLU & Streamlit in Action

This 1 line let's you visualize and play with 1000+ SOTA NLU & NLP models in 200 languages

streamlit run https://raw.githubusercontent.com/JohnSnowLabs/nlu/master/examples/streamlit/01_dashboard.py

NLU provides tight and simple integration into Streamlit, which enables building powerful webapps in just 1 line of code which showcase the.

View the NLU&Streamlit documentation or NLU & Streamlit examples section.

The entire GIF demo and

All NLU resources overview

Take a look at our official NLU page: https://nlu.johnsnowlabs.com/ for user documentation and examples

| Ressource | Description|

|-----------------------------------------------------------------------|-------------------------------------------|

| Install NLU | Just run pip install nlu pyspark==3.0.2

| The NLU Namespace | Find all the names of models you can load with nlu.load()

| The nlu.load() function | Load any of the 1000+ models in 1 line

| The nlu.load().predict(data) function | Predict on Strings, List of Strings, Numpy Arrays, Pandas, Modin and Spark Dataframes

| The nlu.load().fit(data) function | Train a text classifier for 2-Class, N-Classes Multi-N-Classes, Named-Entitiy-Recognition or Parts of Speech Tagging

| The nlu.load().viz(data) function | Visualize the results of Word Embedding Similarity Matrix, Named Entity Recognizers, Dependency Trees & Parts of Speech, Entity Resolution,Entity Linking or Entity Status Assertion

| The nlu.load().viz_streamlit(data) function | Display an interactive GUI which lets you explore and test every model and feature in NLU in 1 click.

| General Concepts | General concepts in NLU

| The latest release notes | Newest features added to NLU

| Overview NLU 1-liners examples | Most common used models and their results

| Overview NLU 1-liners examples for healthcare models | Most common used healthcare models and their results

| Overview of all NLU tutorials and Examples | 100+ tutorials on how to use NLU on text datasets for various problems and from various sources like Twitter, Chinese News, Crypto News Headlines, Airline Traffic communication, Product review classifier training,

| Connect with us on Slack | Problems, questions or suggestions? We have a very active and helpful community of over 2000+ AI enthusiasts putting NLU, Spark NLP & Spark OCR to good use

| Discussion Forum | More indepth discussion with the community? Post a thread in our discussion Forum

| John Snow Labs Medium | Articles and Tutorials on the NLU, Spark NLP and Spark OCR

| John Snow Labs Youtube | Videos and Tutorials on the NLU, Spark NLP and Spark OCR

| NLU Website | The official NLU website

|Github Issues | Report a bug

Getting Started with NLU

To get your hands on the power of NLU, you just need to install it via pip and ensure Java 8 is installed and properly configured. Checkout Quickstart for more infos

pip install nlu pyspark==3.0.2

Loading and predicting with any model in 1 line python

import nlu 

nlu.load('sentiment').predict('I love NLU! <3') 

Loading and predicting with multiple models in 1 line

Get 6 different embeddings in 1 line and use them for downstream data science tasks!

nlu.load('bert elmo albert xlnet glove use').predict('I love NLU! <3') 

What kind of models does NLU provide?

NLU provides everything a data scientist might want to wish for in one line of code!

• NLU provides everything a data scientist might want to wish for in one line of code!

• 1000 + pre-trained models

• 100+ of the latest NLP word embeddings ( BERT, ELMO, ALBERT, XLNET, GLOVE, BIOBERT, ELECTRA, COVIDBERT) and different variations of them

• 50+ of the latest NLP sentence embeddings ( BERT, ELECTRA, USE) and different variations of them

• 100+ Classifiers (NER, POS, Emotion, Sarcasm, Questions, Spam)

• 300+ Supported Languages

• Summarize Text and Answer Questions with T5

• Labeled and Unlabeled Dependency parsing

• Various Text Cleaning and Pre-Processing methods like Stemming, Lemmatizing, Normalizing, Filtering, Cleaning pipelines and more

Classifiers trained on many different datasets

Choose the right tool for the right task! Whether you analyze movies or twitter, NLU has the right model for you!

• trec6 classifier

• trec10 classifier

• spam classifier

• fake news classifier

• emotion classifier

• cyberbullying classifier

• sarcasm classifier

• sentiment classifier for movies

• IMDB Movie Sentiment classifier

• Twitter sentiment classifier

• NER pretrained on ONTO notes

• NER trainer on CONLL

• Language classifier for 20 languages on the wiki 20 lang dataset.

Utilities for the Data Science NLU applications

Working with text data can sometimes be quite a dirty job. NLU helps you keep your hands clean by providing components that take away from data engineering intensive tasks.

• Datetime Matcher

• Pattern Matcher

• Chunk Matcher

• Phrases Matcher

• Stopword Cleaners

• Pattern Cleaners

• Slang Cleaner

Where can I see all models available in NLU?

For NLU models to load, see the NLU Namespace or the John Snow Labs Modelshub or go straight to the source.

Supported Data Types

• Pandas DataFrame and Series

• Spark DataFrames

• Modin with Ray backend

• Modin with Dask backend

• Numpy arrays

• Strings and lists of strings

Overview of all tutorials using the NLU-Library

In the following tabular, all available tutorials using NLU are listed. These tutorials will help you learn the

usage of the NLU library and on how to use it for your own tasks. Some of the tasks NLU does are

translating from any language to the english language, lemmatizing, tokenizing, cleaning text from

Symbol or unwanted syntax, spellchecking, detecting entities, analyzing sentiments and many more!

{:.table2}

| Tutorial Description | NLU Spells Used |Open In Colab | Dataset and Paper References |

|-----------------------------------------------------------------------------------------------------|---------------------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|

| Albert Word Embeddings with NLU | albert, sentiment pos albert emotion | | Albert-Paper, Albert on Github, Albert on TensorFlow, T-SNE, T-SNE-Albert, Albert_Embedding |

| Bert Word Embeddings with NLU | bert, pos sentiment emotion bert | | Bert-Paper, Bert Github, T-SNE, T-SNE-Bert, Bert_Embedding |

| BIOBERT Word Embeddings with NLU | biobert , sentiment pos biobert emotion | | BioBert-Paper, Bert Github , BERT: Deep Bidirectional Transformers, Bert Github, T-SNE, T-SNE-Biobert, Biobert_Embedding |

| COVIDBERT Word Embeddings with NLU | covidbert, sentiment covidbert pos | | CovidBert-Paper, Bert Github, T-SNE, T-SNE-CovidBert, Covidbert_Embedding |

| ELECTRA Word Embeddings with NLU | electra, sentiment pos en.embed.electra emotion | | Electra-Paper, T-SNE, T-SNE-Electra, Electra_Embedding |

| ELMO Word Embeddings with NLU | elmo, sentiment pos elmo emotion | | ELMO-Paper, Elmo-TensorFlow, T-SNE, T-SNE-Elmo, Elmo-Embedding |

| GLOVE Word Embeddings with NLU | glove, sentiment pos glove emotion | | Glove-Paper, T-SNE, T-SNE-Glove , Glove_Embedding |

| XLNET Word Embeddings with NLU | xlnet, sentiment pos xlnet emotion | | XLNet-Paper, Bert Github, T-SNE, T-SNE-XLNet, Xlnet_Embedding |

| Multiple Word-Embeddings and Part of Speech in 1 Line of code | bert electra elmo glove xlnet albert pos | | Bert-Paper, Albert-Paper, ELMO-Paper, Electra-Paper, XLNet-Paper, Glove-Paper |

| Normalzing with NLU | norm | | - |

| Detect sentences with NLU | sentence_detector.deep, sentence_detector.pragmatic, xx.sentence_detector | | Sentence Detector |

| Spellchecking with NLU | n.a. | n.a. | - |

| Stemming with NLU | en.stem, de.stem | | - |

| Stopwords removal with NLU | stopwords | | Stopwords |

| Tokenization with NLU | tokenize | | - |

| Normalization of Documents | norm_document | | - |

| Open and Closed book question answering with Google's T5 | en.t5 , answer_question | | T5-Paper, T5-Model |

| Overview of every task available with T5 | en.t5.base | | T5-Paper, T5-Model |

| Translate between more than 200 Languages in 1 line of code with Marian Models | tr.translate_to.fr, en.translate_to.fr ,fr.translate_to.he , en.translate_to.de | | Marian-Papers, Translation-Pipeline (En to Fr), Translation-Pipeline (En to Ger) |

| BERT Sentence Embeddings with NLU | embed_sentence.bert, pos sentiment embed_sentence.bert | | Bert-Paper, Bert Github, Bert-Sentence_Embedding |

| ELECTRA Sentence Embeddings with NLU | embed_sentence.electra, pos sentiment embed_sentence.electra | | Electra Paper, Sentence-Electra-Embedding |

| USE Sentence Embeddings with NLU | use, pos sentiment use emotion | | Universal Sentence Encoder, USE-TensorFlow, Sentence-USE-Embedding |

| Sentence similarity with NLU using BERT embeddings | embed_sentence.bert, use en.embed_sentence.electra embed_sentence.bert | | Bert-Paper, Bert Github, Bert-Sentence_Embedding |

| Part of Speech tagging with NLU | pos | | Part of Speech |

| NER Aspect Airline ATIS | en.ner.aspect.airline | | NER Airline Model, Atis intent Dataset |

| NLU-NER_CONLL_2003_5class_example | ner | | NER-Piple |

| Named-entity recognition with Deep Learning ONTO NOTES | ner.onto | | NER_Onto |

| Aspect based NER-Sentiment-Restaurants | en.ner.aspect_sentiment | | - |

| Detect Named Entities (NER), Part of Speech Tags (POS) and Tokenize in Chinese | zh.segment_words, zh.pos, zh.ner, zh.translate_to.en | | Translation-Pipeline (Zh to En) |

| Detect Named Entities (NER), Part of Speech Tags (POS) and Tokenize in Japanese | ja.segment_words, ja.pos, ja.ner, ja.translate_to.en | | Translation-Pipeline (Ja to En) |

| Detect Named Entities (NER), Part of Speech Tags (POS) and Tokenize in Korean | ko.segment_words, ko.pos, ko.ner.kmou.glove_840B_300d, ko.translate_to.en | | - |

| Date Matching | match.datetime | | - |

| Typed Dependency Parsing with NLU | dep | | Dependency Parsing |

| Untyped Dependency Parsing with NLU | dep.untyped | | - |

| E2E Classification with NLU | e2e | | e2e-Model |

| Language Classification with NLU | lang | | - |

| Cyberbullying Classification with NLU | classify.cyberbullying | | Cyberbullying-Classifier |

| Sentiment Classification with NLU for Twitter | emotion | | Emotion detection |

| Fake News Classification with NLU | en.classify.fakenews | | Fakenews-Classifier |

| Intent Classification with NLU | en.classify.intent.airline | | Airline-Intention classifier, Atis-Dataset |

| Question classification based on the TREC dataset | en.classify.questions | | Question-Classifier |

| Sarcasm Classification with NLU | en.classify.sarcasm | | Sarcasm-Classifier |

| Sentiment Classification with NLU for Twitter | en.sentiment.twitter | | Sentiment_Twitter-Classifier |

| Sentiment Classification with NLU for Movies | en.sentiment.imdb | | Sentiment_imdb-Classifier |

| Spam Classification with NLU | en.classify.spam | | Spam-Classifier |

| Toxic text classification with NLU | en.classify.toxic | | Toxic-Classifier |

| Unsupervised keyword extraction with NLU using the YAKE algorithm | yake | | - |

| Grammatical Chunk Matching with NLU | match.chunks | | - |

| Getting n-Grams with NLU | ngram | | - |

| Assertion | en.med_ner.clinical en.assert, en.med_ner.clinical.biobert en.assert.biobert, ... | | Healthcare-NER, NER_Clinical-Classifier, Toxic-Classifier |

| De-Identification Model overview | med_ner.jsl.wip.clinical en.de_identify, med_ner.jsl.wip.clinical en.de_identify.clinical, ... | | NER-Clinical |

| Drug Normalization | norm_drugs | | - |

| Entity Resolution | med_ner.jsl.wip.clinical en.resolve_chunk.cpt_clinical, med_ner.jsl.wip.clinical en.resolve.icd10cm, ... | | NER-Clinical, Entity-Resolver clinical |

| Medical Named Entity Recognition | en.med_ner.ade.clinical, en.med_ner.ade.clinical_bert, en.med_ner.anatomy,en.med_ner.anatomy.biobert, ... | | - |

| Relation Extraction | en.med_ner.jsl.wip.clinical.greedy en.relation, en.med_ner.jsl.wip.clinical.greedy en.relation.bodypart.problem, ... | | - |

| Visualization of NLP-Models with Spark-NLP and NLU | ner, dep.typed, med_ner.jsl.wip.clinical resolve_chunk.rxnorm.in, med_ner.jsl.wip.clinical resolve.icd10cm | | NER-Piple, Dependency Parsing, NER-Clinical, Entity-Resolver (Chunks) clinical |

| NLU Covid-19 Emotion Showcase | emotion | | Emotion detection |

| NLU Covid-19 Sentiment Showcase | sentiment | | Sentiment classification |

| NLU Airline Emotion Demo | emotion | | Emotion detection |

| NLU Airline Sentiment Demo | sentiment | | Sentiment classification |

| Bengali NER Hindi Embeddings for 30 Models | bn.ner, bn.lemma, ja.lemma, am.lemma, bh.lemma, en.ner.onto.bert.small_l2_128,.. | | Bengali-NER, Bengali-Lemmatizer, Japanese-Lemmatizer, Amharic-Lemmatizer |

| Entity Resolution | med_ner.jsl.wip.clinical en.resolve.umls, med_ner.jsl.wip.clinical en.resolve.loinc, med_ner.jsl.wip.clinical en.resolve.loinc.biobert | | - |

| NLU 20 Minutes Crashcourse - the fast Data Science route | spell, sentiment, pos, ner, yake, en.t5, emotion, answer_question, en.t5.base ... | | T5-Model, Part of Speech, NER-Piple, Emotion detection , Spellchecker, Sentiment classification |

| Chapter 0: Intro: 1-liners | sentiment, pos, ner, bert, elmo, embed_sentence.bert | | Part of Speech, NER-Piple, Sentiment classification, Elmo-Embedding, Bert-Sentence_Embedding |

| Chapter 1: NLU base-features with some classifiers on testdata | emotion, yake, stem | | Emotion detection |

| Chapter 2: Translation between 300+ languages with Marian | tr.translate_to.en, en.translate_to.fr, en.translate_to.he | | Translation-Pipeline (En to Fr), Translation (En to He) |

| Chapter 3: Answer questions and summarize Texts with T5 | answer_question, en.t5, en.t5.base | | T5-Model |

| Chapter 4: Overview of T5-Tasks | en.t5.base | | T5-Model |

| Graph NLU 20 Minutes Crashcourse - State of the Art Text Mining for Graphs | spell, sentiment, pos, ner, yake, emotion, med_ner.jsl.wip.clinical, ... | | Part of Speech, NER-Piple, Emotion detection, Spellchecker, Sentiment classification |

| Healthcare with NLU | med_ner.human_phenotype.gene_biobert, med_ner.ade_biobert, med_ner.anatomy, med_ner.bacterial_species,... | | - |

| Part 0: Intro: 1-liners | spell, sentiment, pos, ner, bert, elmo, embed_sentence.bert | | Bert-Paper, Bert Github, T-SNE, T-SNE-Bert , Part of Speech, NER-Piple, Spellchecker, Sentiment classification, Elmo-Embedding , Bert-Sentence_Embedding |

| Part 1: NLU base-features with some classifiers on Testdata | yake, stem, ner, emotion | | NER-Piple, Emotion detection |

| Part 2: Translate between 200+ Languages in 1 line of code with Marian-Models | en.translate_to.de, en.translate_to.fr, en.translate_to.he | | Translation-Pipeline (En to Fr), Translation-Pipeline (En to Ger), Translation (En to He) |

| Part 3: More Multilingual NLP-translations for Asian Languages with Marian | en.translate_to.hi, en.translate_to.ru, en.translate_to.zh | | Translation (En to Hi), Translation (En to Ru), Translation (En to Zh) |

| Part 4: Unsupervise Chinese Keyword Extraction, NER and Translation from chinese news | zh.translate_to.en, zh.segment_words, yake, zh.lemma, zh.ner | | Translation-Pipeline (Zh to En), Zh-Lemmatizer |

| Part 5: Multilingual sentiment classifier training for 100+ languages | train.sentiment, xx.embed_sentence.labse train.sentiment | n.a. | Sentence_Embedding.Labse |

| Part 6: Question-answering and Text-summarization with T5-Modell | answer_question, en.t5, en.t5.base | | T5-Paper |

| Part 7: Overview of all tasks available with T5 | en.t5.base | | T5-Paper |

| Part 8: Overview of some of the Multilingual modes with State Of the Art accuracy (1-liner) | bn.lemma, ja.lemma, am.lemma, bh.lemma, zh.segment_words, ... | | Bengali-Lemmatizer, Japanese-Lemmatizer , Amharic-Lemmatizer |

| Overview of some Multilingual modes avaiable with State Of the Art accuracy (1-liner) | bn.ner.cc_300d, ja.ner, zh.ner, th.ner.lst20.glove_840B_300D, ar.ner | | Bengali-NER

| NLU 20 Minutes Crashcourse - the fast Data Science route | - | | - |

Need help?

• Ping us on Slack

• Post an issue on Github

Simple NLU Demos

• NLU different output levels Demo

Features in NLU Overview

• Tokenization

• Trainable Word Segmentation

• Stop Words Removal

• Token Normalizer

• Document Normalizer

• Stemmer

• Lemmatizer

• NGrams

• Regex Matching

• Text Matching,

• Chunking

• Date Matcher

• Sentence Detector

• Deep Sentence Detector (Deep learning)

• Dependency parsing (Labeled/unlabeled)

• Part-of-speech tagging

• Sentiment Detection (ML models)

• Spell Checker (ML and DL models)

• Word Embeddings (GloVe and Word2Vec)

• BERT Embeddings (TF Hub models)

• ELMO Embeddings (TF Hub models)

• ALBERT Embeddings (TF Hub models)

• XLNet Embeddings

• Universal Sentence Encoder (TF Hub models)

• BERT Sentence Embeddings (42 TF Hub models)

• Sentence Embeddings

• Chunk Embeddings

• Unsupervised keywords extraction

• Language Detection & Identification (up to 375 languages)

• Multi-class Sentiment analysis (Deep learning)

• Multi-label Sentiment analysis (Deep learning)

• Multi-class Text Classification (Deep learning)

• Neural Machine Translation

• Text-To-Text Transfer Transformer (Google T5)

• Named entity recognition (Deep learning)

• Easy TensorFlow integration

• GPU Support

• Full integration with Spark ML functions

• 1000 pre-trained models in +200 languages!

• Multi-lingual NER models: Arabic, Chinese, Danish, Dutch, English, Finnish, French, German, Hewbrew, Italian, Japanese, Korean, Norwegian, Persian, Polish, Portuguese, Russian, Spanish, Swedish, Urdu and more

• Natural Language inference

• Coreference resolution

• Sentence Completion

• Word sense disambiguation

• Clinical entity recognition

• Clinical Entity Linking

• Entity normalization

• Assertion Status Detection

• De-identification

• Relation Extraction

• Clinical Entity Resolution

Citation

We have published a paper that you can cite for the NLU library:

@article{KOCAMAN2021100058,

    title = {Spark NLP: Natural language understanding at scale},

    journal = {Software Impacts},

    pages = {100058},

    year = {2021},

    issn = {2665-9638},

    doi = {https://doi.org/10.1016/j.simpa.2021.100058},

    url = {https://www.sciencedirect.com/science/article/pii/S2665963821000063},

    author = {Veysel Kocaman and David Talby},

    keywords = {Spark, Natural language processing, Deep learning, Tensorflow, Cluster},

    abstract = {Spark NLP is a Natural Language Processing (NLP) library built on top of Apache Spark ML. It provides simple, performant & accurate NLP annotations for machine learning pipelines that can scale easily in a distributed environment. Spark NLP comes with 1100+ pretrained pipelines and models in more than 192+ languages. It supports nearly all the NLP tasks and modules that can be used seamlessly in a cluster. Downloaded more than 2.7 million times and experiencing 9x growth since January 2020, Spark NLP is used by 54% of healthcare organizations as the world’s most widely used NLP library in the enterprise.}

    }

}