Practical overview
When it fits
Use the package when a spaCy pipeline needs lookup data that is not bundled with its pretrained model or the core installation. The README gives Serbian and Turkish as examples for lookup-based lemmatization and recommends the data for training new models, particularly blank models.
Who it is for
- Developers training spaCy models, especially blank models that need language-specific lookup data.
- NLP practitioners adding lemmatization or normalization resources for languages without a suitable pretrained model or third-party lemmatizer.
Documented uses
- Add lookup-based lemmatization and normalization data to spaCy workflows.
- Provide language resources to blank spaCy models during training.
- Include optional probability or Brown cluster data in spaCy workflows.
Requirements & limits
- The package must be installed in the same environment as spaCy for its language resources to be available.
- A pretrained model may already provide serialized vocabulary and lookup tables, so separate lookup data is not always needed.
- The resources support spaCy v2.2+ according to the repository README.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install lookup data for a spaCy environment
Make the package’s language-specific lookup resources available to spaCy for setting up its vocabulary and lookups.
Before you start
- spaCy v2.2 or later
- Install both packages in the same environment
Steps
Install spaCy in the environment you intend to use.
Install the spaCy lookups data package in that same environment.
Use spaCy in that environment so it can check the package’s language resources when setting up the vocabulary and lookups.
Expected outcome
The package’s available language-specific resources can be discovered by spaCy during vocabulary and lookup setup.
Guide limitations
- The package provides lookup data; the documented steps do not configure a particular language or pipeline component.
- Lookup data may already be supplied by a pretrained model. The package is recommended for languages without a suitable pretrained model, and for training blank models that need lemmatization or normalization data.
Include lookup tables when saving a trained blank spaCy model
Add the package’s lemmatization and normalization data to a blank model and save the model with its lookup tables.
Before you start
- spaCy and the lookup data package installed in the same environment
- A blank spaCy model being prepared for training
Steps
Install the lookup data package in the same environment as spaCy.
Create and train the blank spaCy model using that environment.
Save the model with
nlp.to_disk.
Expected outcome
The saved model includes lookup tables as part of its `Vocab`.
Guide limitations
- The package contains lemmatization and normalization data for 25+ languages; the source does not specify availability for every language.
- These steps cover saving the model, not a complete training configuration or deployment process.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 15 October 2026
- Original website title
- explosion/spacy-lookups-data · Official homepage title · retained in its source language