Blog · Explosion

Against LLM maximalism

May 18, 2023 · LLMs are not a direct solution to most of the NLP use-cases companies have been working on. They are extremely useful, but if you want to deliver reliable software you can improve over time, you can't just write a prompt and call it a day. Once you're past prototyping and want to deliver the best system you can, supervised learning will often give you better efficiency, accuracy and reliability.

Against LLM maximalism

💌 Stay in the loop!

How the Guardian approaches quote extraction with NLP

Introducing Holmes 4.0

Introducing spaCy v3.3

Healthsea: an end-to-end spaCy pipeline for exploring health supplement effects

Introducing spaCy v3.2

Introducing spaCy v3.0

sense2vec reloaded: contextually-keyed word vectors

Explosion in 2017: Our Year in Review

Supervised learning is great — it's data collection that's broken

Building your bot's brain with Node.js and spaCy

Introducing Explosion AI

Multi-threading spaCy's parser and named entity recognizer

Dead Code Should Be Buried

Parsing English in 500 Lines of Python

Implementing a custom trainable component for relation extraction

Introducing spaCy v3.5

End-to-end Neural Coreference Resolution in spaCy

Introducing spaCy v3.4

Compact word vectors with Bloom embeddings

Universal Dependencies v2.5 Benchmarks for spaCy

We’ve sold 5% of Explosion

Explosion in 2020: Our Year in Review

Introducing spaCy v2.2

Introducing custom pipelines and extensions for spaCy v2.0

Supervised similarity: Learning symmetric relations from duplicate question data

spaCy v1.0: Deep Learning with custom pipelines and Keras

How front-end development can improve Artificial Intelligence

spaCy now speaks German

How spaCy Works

A Good Part-of-Speech Tagger in about 200 Lines of Python

The Tale of Bloom Embeddings and Unseen Entities

Explosion in 2022: Our Year in Review

spaCy behind the scenes: library patterns & design concepts explained

Spancat: a new approach for span labeling

Introducing spaCy Tailored Pipelines

Neural edit-tree lemmatization for spaCy

Introducing spaCy v3.1

Introducing spaCy v2.3

spaCy meets Transformers: Fine-tune BERT, XLNet and GPT-2

Pseudo-rehearsal: A simple solution to catastrophic forgetting for NLP

Deep text-pair classification with Quora's 2017 question dataset

An open-source named entity visualiser for the modern web

A natural language user interface is just a user interface

Statistical NLP in the Ten Hundred Most Common English Words

Introducing spaCy

Deploying a Prodigy cloud service for Posh’s financial chatbots

Fast transformer inference with Metal Performance Shaders

floret: lightweight, robust word vectors

Diary of a spaCy project: Predicting GitHub Tags

Explosion in 2021: Our Year in Review

spaCy v3's project and config systems are pretty great

Applied NLP Thinking: How to Translate Problems into Solutions

Explosion in 2019: Our Year in Review

Introducing spaCy v2.1

Prodigy: A new tool for radically efficient machine teaching

Embed, encode, attend, predict: The new deep learning formula for state-of-the-art NLP models

displaCy.js: An open-source NLP visualiser for the modern web

SyntaxNet in context: Understanding Google's new TensorFlow NLP model

Sense2vec with spaCy and Gensim

Writing C in Cython

Release radar