Gensim

Gensim

Unsupervised embeddings

Vector + Similarity Is All You Need

Gensim

Unsupervised embeddings

Online

Fast

Robust

Production-ready

Gensim

Unsupervised embeddings

When to use?

Text classification

Sentiment analysis

Search engines

NER

Topic modeling

...

Open-source

Maintainers

∞ - XX 2016

XX 2016 - May 2017

May 2017 - Feb 2019

Feb 2019 - ∞

Radim Řehůřek

Lev Konstantinovskiy

Ivan Menshikh

Michael Penkov

Open-source

Why I'm here?

1. Used gensim on work
2. Feature requests / bug fixes to gensim
3. "Lev, I want to change a job"
4. ....
5. PROFIT, you are fired hired!

Open-source

Core developers

Open-source

Core developers

Students

NMF

FastText (cython)

gensim-data

FastText (python)

TM Viz

Corpusfile

ATM

Doc

Open-source

Core developers

Students

Radom contributors

@544895340
@AMR
@AadityaJ
@Alexjmsherman
@AustenLamacraft
@CLearERR
@Cheukting
@DennisChen0307
@ELind77
@ElSaico
@Fil

@HodorTheCoder
@IrinaGoloshchapova
@Jayantj
@JonathanHourany
@Karamax
@KenjiOhtsuka
@KiddoZhu
@KokuKUSIAKU
@Kreiswolke
@LShostenko
@Laubeee

@MridulS
@MritunjayMohitesh
@PeteBleackley
@PeterHamilton
@RishabGoel
@RunHorst
@SamriddhiJain
@Shiki
@Stamenov
@Stigjb
@TheFlash10

@Utkarsh
@VorontsovIE
@Witiko
@Xinyi2016
@Zohaggie
@abhinavchawla
@accraze
@ajkl
@akarazeev
@akutuzov
@alantian

@alexgarel
@allenyllee
@andrewjlm
@aneesh
@anmol01gulat
@anmolgulati
@anujkhare
@aquatiko
@arlenk
@arttii
@bahbbc

Project structure

• RaRe-Technologies/gensim
  • Code & documentation
  • CI (tests, docs, code-style)
• RaRe-Technologies/gensim-data
  • Pre-trained models
  • Datasets
• RaRe-Technologies/smart-open
  • Universal reader/writer
  • (De) Compression
  • S3 / HTTP / HDFS

• MacPython/gensim-wheels
  • CI (tests + wheels PyPI)

• conda-forge/gensim-feedstock​
  • CI (tests + Conda)​

• Documentation

Community

• Google Groups
  • Any topic
  • Support

• Github
  • Feature requests
  • Bug reports
  • Holywars

• Twitter
  • Announces
  • Short discussions

• Gitter
  • Chat
  • Awful, really
  • Infinite context-switch

Maintainer?

Maintainer?

Expectation

Maintainer?

Expectation

Reality

Maintainer?

Goal: improve project

• Support
• Code-review & Merge
• Releases
• Roadmap
• Anything that nobody want to do
  • Setup env, CI, checkers, etc
  • Guides, documentation
  • Coordinates an contributors
• Sometimes (never) add a new features

What's next with gensim?

• Project in "slow maintenance mode" until ¯\_(ツ)_/¯​
  • Bugfixes / documentation improvements
  • Code cleanup
  • No roadmap
  • No GSoC 2019
  • No student incubator
  • No awesome features planned

When do you implement a model X ?

ULMFit, BERT, LASER, etc ...

Most likely never

How can I help?

• Fix any bug from issues
• Improve documentation
• Ask @mpenkov for advice