seqeval

What it is and what it does

seqeval is a Python evaluation framework for sequence labeling tasks—the kind of work you do when training models for named-entity recognition, part-of-speech tagging, semantic role labeling, or other chunking problems. It computes standard metrics (accuracy, precision, recall, F1) and generates classification reports, and it supports multiple tagging schemes (IOB1, IOB2, IOE1, IOE2, IOBES, BILOU). The package has two evaluation modes: a default mode compatible with the conlleval Perl script, and a strict mode that enforces schema compliance more rigidly.

The package has no runtime dependencies, making installation straightforward, but it is dormant—the last release was in October 2020 and there have been no commits since August 2024. If you need sequence labeling evaluation and the existing metrics and schemes cover your use case, seqeval is a lightweight, stable choice; if you need active maintenance, bug fixes, or support for newer tagging schemes, you should verify whether the package still works with your Python version and consider alternatives.

Use it for:

• Evaluate NER model predictions against ground-truth labels using standard metrics like precision and recall.
• Generate a classification report showing per-entity performance for a part-of-speech tagger or chunker.
• Compare predictions across different tagging schemes (IOB1, IOB2, IOBES) to understand how schema choice affects scoring.
• Validate sequence labeling output in strict mode to catch malformed tag sequences that default mode would overlook.
• Benchmark sequence labeling systems using conlleval-compatible metrics for reproducibility with published results.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

seqeval evaluates the performance of sequence labeling tasks like named-entity recognition and part-of-speech tagging by computing accuracy, precision, recall, and F1 scores across multiple tagging schemes.

Yes, if your sequence labeling evaluation needs match the supported schemes and metrics. The package is lightweight, has no dependencies, carries a permissive MIT license, and is well-tested against conlleval. However, it is dormant—last release October 2020—so verify compatibility with your Python version before adopting. Do not use if you need active maintenance or support for newer tagging schemes.

Install

seqeval on PyPI

pip install seqeval

uv add seqeval

poetry add seqeval

Installing seqeval

Before you install

High install friction with no runtime dependencies. The package is dormant—last release was 2020-10-24 and no commits since 2024-08-28—so expect no active maintenance or bug fixes going forward.

License in practice

MIT license is permissive and imposes no restrictions on use or redistribution, making it safe to adopt in any project.

Quickstart

pip install seqeval
from seqeval.metrics import f1_score
y_true = [['O', 'O', 'B-MISC', 'I-MISC']]
y_pred = [['O', 'B-MISC', 'I-MISC', 'I-MISC']]
print(f1_score(y_true, y_pred))

Verify before relying

• Whether the package works reliably with modern Python versions (classifiers list Python 2.6–3.6, but latest release was 2020; current compatibility unknown).
• Whether strict mode with IOBES and BILOU schemes is production-ready or has known limitations.

Package facts

Evidence: seqeval-1.2.2.tar.gz

Tags

More Testing packages