Large Language Models: From Prototype to ProductionEuroPython KeynoteLarge Language Models (LLMs) have shown some impressive capabilities and their impact is the topic of the moment. In this talk, Ines presents visions for NLP in the age of LLMs and a pragmatic, practical approach for how to use Large Language Models to ship more successful NLP projects from prototype to production today.
How Good is the Model in Model-in-the-loop Event Coreference Resolution Annotation?Ahmed, Nath, Regan, Pollins, Krishnaswamy, Martin (2023)Figure 6 illustrates the interface design of the annotation methodology on the popular model-in-the-loop annotation tool - Prodigy. We use this tool for the simplicity it offers in plugging in the various ranking methods we explained.
Against LLM maximalismLLMs 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.
Implementing a custom trainable component for relation extractionRelation extraction refers to the process of predicting and labeling semantic relationships between named entities. In this blog post, we'll go over the process of building a custom relation extraction component using spaCy and Thinc. We'll also add a Hugging Face transformer to improve performance at the end of the post. You'll see how you can utilize Thinc's flexible and customizable system to build an NLP pipeline for biomedical relation extraction.
Rulers, NER, and data iterationAbout the power of Rules + ML and the importance of iteration on your pipeline and your data.
Fiscal data in text: Information extraction from audit reports using Natural Language ProcessingBeltran (2023), Data & Policy, Cambridge University PressI relied on the text annotation software Prodigy in Python that offers a friendly user interface where the reviewer can read the text and assign a label to each paragraph.
Is it possible to have entities within entities within entities?PyData Global 2022Named entity recognition models might not be able to handle a wide variety of spans, but Spancat certainly can! Dive into named entity recognition, its limitations, and how we’ve solved them with a solution-focused talk and practical applications.
Finetuning and Bulk Labelling Images with Prodigy In this video, we’ll show how you might be able to improve the annotation experience by using bulk labelling for image classification.
Task Routers in ProdigyHow to use the new task routers to customize how examples are assigned in multi-annotator workflows.
✨ prodigy v1.12.0Jul 5, 2023LLM-assisted workflows for annotation and prompt engineering, task routing for multi-annotator setups
Concepts and measures of bureaucratic constraints in European Union laws from hand-coding to machine-learningFranchino, Migliorati, Pagano, Vignoli (2023)The models “learn” the relations between the text tokens and the entity categories from two randomly selected samples of sentences that are extracted from a pre-processed corpus and have been manually annotated using the Python-implemented platform “Prodigy”.
Newsletter May 2023We got so much amazing feedback from the spaCy user survey, thank you all for your contributions! The most requested feature was spaCy integration with LLMs, which is why we’re so excited to announce spacy-llm!
You are what you read: Building a personal internet front-page with spaCy and ProdigyPyCon DE & PyData Berlin
The Tale of Bloom Embeddings and Unseen EntitiesThe default Bloom embedding layer in spaCy is unconventional, but very powerful and efficient. We wrote about it before and showed the advantages it provides in terms of memory efficiency for our floret embeddings. Now we have released the first technical report by Explosion, where we explain Bloom embeddings in more detail and rigorously compare them to traditional embeddings. In this post we'll highlight some of our results with a special focus on unseen entities.
Slovak Dataset for Multilingual Question AnsweringHládek, Staš, Juhár, Koctúr (2023)We used the Prodigy annotation tool to annotate the questions and answers. One annotation task corresponds to one web application deployment and different configurations.
Introducing spaCy v3.5spaCy v3.5 introduces new CLI commands, fuzzy matching, improvements for entity linking and more.
Training spaCy NER Models with ProdigyThis handy flowchart contains our most common tips, tricks, and best practices for training and updating spaCy named entity recognition models with Prodigy.
Setting your ML project up for success“What can you do to maximize probability of success for your Machine Learning solution? Throughout my 15 years as data scientist in academia, big pharma and through consulting, one common theme has emerged: the most reliable predictor of success for any NLP or ML-based solution is whether or not you involve the data science team early on.”
Fast transformer inference with Metal Performance ShadersWe are happy to introduce support for Metal Performance Shaders in Thinc PyTorch layers. This makes it possible to run spaCy transformer-based pipelines on GPU on Apple Silicon Macs and improves inference speed up to 4.7 times.
SpanCat with spaCy and Prodigy on real dataYouTube series by WJB Mattingly showing an end-to-end project, from cultivating and annotating data to training, testing and visualizing a model.
Efficient Information Extraction From Text With spaCyJetBrains PyCharmThis webinar takes you through building a spaCy project that uses a named entity recognition (NER) model to extract entities of interest from restaurant reviews, like prices, opening hours and ratings.
Creating Custom Event Data Without Dictionaries: A Bag-of-TricksHalterman, Schrodt, Beger, Bagozzi, Scarborough (2023)While in the past the process of generating training case has been quite time consuming and tedious, newer approaches such as those incorporated into the web-based Prodigy annotation system allow this to be done much more quickly.
Deploying a Prodigy cloud service for Posh’s financial chatbotsA Prodigy case study of Posh AI's production-ready annotation platform and custom chatbot annotation tasks for banking customers.
Reflections on a year of spaCy consulting at ExplosionIn this post, Peter shares some lessons learned from chatting with practitioners about their NLP challenges, developing production-ready NLP pipelines for clients, and working with an open-source development team.
Multi hash embeddings in spaCyMiranda, Kádár, Boyd, Van Landeghem, Søgaard, Honnibal (2022)In this technical report we lay out a bit of history and introduce the embedding methods in spaCy in detail. Second, we critically evaluate the hash embedding architecture with multi-embeddings on Named Entity Recognition datasets from a variety of domains and languages. The experiments validate most key design choices behind spaCy’s embedders, but we also uncover a few surprising results.
Custom Interfaces with blocksYou can create custom annotation layouts in Prodigy using the annotation widgets that Prodigy provides by using the blocks feature. This video explains how to use this feature by building a custom interface that can manually annotate and transcribe audio.
Introducing spaCy v3.6spaCy v3.6 introduces the span finder component and trained pipelines for Slovenian.
🦙 spacy-llm v0.3.0Jun 14, 2023Cohere, Anthropic, OpenLLaMa, StableLM, logging, streamlit demo, lemmatization task
Inter-rater agreement for the annotation of neurologic signs and symptoms in electronic health recordsOommen, Howlett-Prieto, Carrithers, Hier (2023)Prodigy was used to annotate neurologic concepts in the EHR physician notes.
Large Disagreement Modelling“In this blogpost I’d like to talk about large language models. There’s a bunch of hype, sure, but there’s also an opportunity to revisit one of my favourite machine learning techniques: disagreement.”
spaCy Plugin for VSCodeThe spaCy VSCode Extension provides additional tooling and features for working with spaCy’s config files. Version 1.0.0 includes hover descriptions for registry functions, variables, and section names within the config as an installable extension.
Predicting relations between SOAP note sections: The value of incorporating a clinical information modelSocrates, Gilson, Lopez, Chi, Taylor, Chartash (2023), Journal of Biomedical InformaticsTo support human annotation, we first annotate 100 Assessment and Plan subsections manually using Prodigy, and then use spacy-transformers to fine-tune a general domain RoBERTa-base model pretrained on OntoNotes 5 for both the Assessment and Plan section NER tagging.
textaCy v0.13.0Utility library for NLP tasks before and after spaCy, including preprocessing, normalization and additional information extraction features.
The Nesta Skills Extractor LibraryEconomic Statistics Centre of ExcellenceA new library for extracting skills from job adverts and mapping them to a taxonomy of your choice, built on top of spaCy.
Towards a Tagalog NLP pipelineIn this blog post, Lj talks about how he built an NER pipeline for Tagalog, the gold-standard dataset, benchmarking results, and his hopes for the future of Tagalog NLP.
Explosion in 2022: Our Year in ReviewIt's been another exciting year at Explosion! We've developed a new end-to-end neural coref component for spaCy, improved the speed of our CNN pipelines up to 60%, and published new pre-trained pipelines for Finnish, Korean, Swedish and Croatian. We've also released several updates to Prodigy and introduced new recipes to kickstart annotation with zero- or few-shot learning.
Extracting Structured Information from Greek Legislation DataAlexios (2023)Worth noting is the existence of an application, called Prodigy, which takes advantage of an active learning framework and provides users with an interactive interface for data annotation.
The triangulation of ethical leader signals using qualitative, experimental, and data science methodsBanks, Ross, Toth, Tonidandel, Goloujeh, Dou, Wesslen (2022)This additional text was labeled by the same coding team using Prodigy, [...] a flexible user interface tool built on top of spaCy, a leading open source library in python for natural language processing. We created a spaCy end‐to‐end project workflow including package versioning, data pre‐processing, data ingestion into a database, annotation sessions using Prodigy’s user interface, model training, model evaluation, python packaging, and visual app for testing the model.
Coreference Resolution in spaCyIn everyday conversation, we use pronouns or other expressions to refer to entities in many different ways, but we effortlessly understand these references. In NLP this is a challenging problem known as Coreference Resolution. In this video, we’ll show how to train spaCy’s new component for Coreference Resolution and how to apply the pipeline to resolve references in a text.
