Recognai

This account is no longer maintained. We are now Argilla, follow us @argilla_io

[Recognai @recogn_ai]

🎉 Excited to release Selectra (Spanish Electra), a new set of models on the @huggingface Hub

3-5x times smaller than current SOTA Spanish models while achieving competitive results

🧵Overview below (1/4)

Thanks @GoogleAI TPU RC for their support

#python #opensource #nlproc

[Recognai @recogn_ai]

🥳 We're extremely excited to announce we're now Argilla

Please don't forget to follow us @argilla_io.

There are many more exciting things coming up!

Read more at: https://t.co/JiCU7iSOJb

#python #opensource #nlproc

[Daniel Vila Suero @dvilasuero]

Find text classification label errors with @CleanlabAI and correct them with @rubrixml

https://t.co/2JVbbe8xrn

#python #opensource #NLProc

[Awesome Machine Learning Repositories @MLRepositories]

rubrix: ✨ Rubrix, open-source framework for data-centric NLP. Data annotation and monitoring for enterprise NLP
Lang: Python
⭐️ 1351
Author: @recogn_ai
#MachineLearning
https://t.co/Utorh0TXzq

[Daniel Vila Suero @dvilasuero]

Get started with NLP with custom datasets

Create and label datasets for text classification, token classification and text generation

https://t.co/VlMC8wkqHB

#python #opensource #datascience

[Daniel Vila Suero @dvilasuero]

Don't have a lot of time to annotate data?

SetFit + Rubrix, few-shot classification with custom data 🤓

https://t.co/U1FEM5zn9j

#nlproc #datascience #opensource

[Argilla @argilla_io]

⚡ New release 0.18.0

> Better token classification validation

> Delete records by id & query for better dataset management

> New tutorials!

Thanks to our community contributors @AnkushChander, Tom Aarsen, & others

https://t.co/FmBpyHdGKk

#python #nlproc #opensource

[Daniel Vila Suero @dvilasuero]

https://t.co/vdYxmwTEd5

#python #nlproc #opensource

[Daniel Vila Suero @dvilasuero]

SetFit: Efficient few-shot learning with Sentence Transformers

So exciting!

Train robust models with very few examples, fast training, fast inference, and comparable/better than other LLMs and prompt-based methods.

https://t.co/sbFh3EcSK3

#python #opensource #NLProc

[Daniel Vila Suero @dvilasuero]

Active learning for text classification with @rubrixml and the wonderful small-text library by @webis_de

Learn how to build a custom active learning loop and teach a 🤗 transformers model

https://t.co/iux9l0rp7M

#python #opensource #NLProc

[Daniel Vila Suero @dvilasuero]

Want to analyze prediction explanations from your Transformer models? At the dataset level?

A new tutorial using SHAP and Transformers interpret!

https://t.co/3wP3B0MaCK

#python #opensource #xai

[Daniel Vila Suero @dvilasuero]

humap: Hierarchical Uniform Manifold Approximation and Projection

A very cool method and library by @EstecioJunior

Reduces visual burden when exploring clusters in large datasets and enables drill-down with hierarchical levels

https://t.co/zx0LNUCOTh

#python #opensource #umap

[Daniel Vila Suero @dvilasuero]

🎉 Rubrix just crossed 1.3K stars this week 🎉

https://t.co/DOrtloj95f

#python #opensource #nlp

[Daniel Vila Suero @dvilasuero]

Rubrix: the open-source framework for data-centric NLP

Build human-in-the-loop workflows for data annotation, monitoring, and review.

https://t.co/DOrtloj95f

Follow @rubrixml for updates

#python #nlp #opensource

[Daniel Vila Suero @dvilasuero]

What can we learn from model predictions vs. training data labels?

* Ambiguous examples
* (Some) wrong labels
* Model improvement patterns

A reproducible example using the @stanfordnlp sentiment treebank dataset & @rubrixml

https://t.co/fMRiIw2MB9

#python #opensource #NLProc

[Daniel Vila Suero @dvilasuero]

Weak supervision for multilabel text classification.

Get instant statistics about heuristics' coverage and precision with @rubrixml UI

Define rules programmatically with Python

Tutorial:
https://t.co/ovPVLLcyMT

#opensource #datacentricai #python

[Daniel Vila Suero @dvilasuero]

Every good model starts with good quality datasets.

Iteration and collaboration are key ingredients to achieve this.

Here's how you can iterate on data and models using the Hugging Face Hub.

https://t.co/agDomR7bnM

#nlproc #datascience #opensource

[Daniel Vila Suero @dvilasuero]

Weak supervision for multilabel text classification?

A step by step tutorial using @rubrixml

https://t.co/ovPVLLcyMT

#python #opensource #NLProc

[Daniel Vila Suero @dvilasuero]

Fine-tuning a sentiment classifier starting with no labeled data with @rubrixml

https://t.co/mBFZrqA8jw

Follow @rubrixml for more resources like this one

If you love NLP & open-source join our friendly community:
https://t.co/O0jE08KGdX

#python #opensource #nlp #transformers

[Daniel Vila Suero @dvilasuero]

BOND: BERT-Assisted Open-Domain Named Entity Recognition with Distant Supervision

1️⃣ Noisy labels using Wikidata and gazetteers (distant labels)
2️⃣ Fine-tune Roberta for NER with distant labels
3️⃣ Self-training

https://t.co/q17JEAWcLl
https://t.co/Cp0Kfba9WN

#python #NLProc

[Daniel Vila Suero @dvilasuero]

Stanza by @stanfordnlp is powerful for NER

Want to see how well it performs with your data? 👇

https://t.co/fhSzDk90HP

New to @rubrixml?
https://t.co/DOrtloiBfH

Join the community:
https://t.co/O0jE08KGdX

#python #nlp #opensource #datascience