Practical overview
When it fits
Use Anago when you want a Python library workflow for training or applying sequence-labeling models from labeled examples, or for loading pretrained weights to process text.
Who it is for
- Python developers building NLP workflows that need sequence labels.
- Researchers or practitioners preparing labeled text to train and evaluate tagging models.
Documented uses
- Train a named-entity recognition model from labeled text.
- Add part-of-speech tagging or semantic role labeling to a text-processing workflow.
- Evaluate sequence-labeling results with the library's scoring method.
- Apply loaded model weights to label new text.
Requirements & limits
The repository is public and identifies the project as MIT-licensed. The README documents installation through pip or from source and states official support for Python 3.4–3.6.
- The documented language and framework are Python and Keras.
- The README lists named-entity recognition, part-of-speech tagging and semantic role labeling.
- The basic model exposes fit, score, analyze, save and load methods.
- Pretrained-weight loading and optional ELMo processing are documented.
- The README says further documentation is forthcoming.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Train and evaluate a sequence-labeling model
Prepare labeled examples, fit Anago's basic Sequence model and score it on test data.
Before you start
- Python; the README states official support for Python 3.4–3.6.
- Anago installed with pip or from the repository.
- Training and test files containing token sequences with corresponding labels.
Steps
Load the training and test sequences and labels with
load_data_and_labels.Create a model with
anago.Sequence()and train it withfit(x_train, y_train, epochs=15).Evaluate it with
model.score(x_test, y_test); the README identifies the reported measure as F1-micro.
Expected outcome
A trained sequence-labeling model and an F1-micro evaluation score.
Guide limitations
- The README notes that pretrained word embeddings are needed for better performance.
- The documented Python support is 3.4–3.6.
Apply pretrained weights to label new text
Load a pretrained model and use it to score test examples or analyze new text.
Before you start
- Anago installed in a Python environment.
- A pretrained model archive at a usable URL; the README demonstrates a CoNLL-2003 English model.
- Test examples if evaluating the loaded model.
Steps
Use
anago.utils.downloadwith the pretrained archive URL to retrieve weights, parameters and a preprocessor.Load the model with
anago.Sequence.load(weights, params, preprocessor).Call
scorewith test data to evaluate, oranalyzewith new text to obtain word and entity labels.
Expected outcome
A loaded model that can score labeled test data or return sequence labels for supplied text.
Guide limitations
- The example uses a particular English CoNLL-2003 archive; the README does not establish coverage for other pretrained models.
- The README states Python 3.4–3.6 support.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 17 October 2026
- Original website title
- Hironsan/anago · Official homepage title · retained in its source language