mallet を Homebrew, MacPorts でインストール

プロジェクトリンク

• ホームページ
• リポジトリ

コマンドとエイリアス

• classifier2info
• csv2classify
• csv2vectors
• mallet
• mallethon
• prepend-license.sh
• svmlight2vectors
• text2classify
• text2vectors
• vectors2classify
• vectors2info
• vectors2topics
• vectors2vectors

プロジェクトの歴史

The project is associated with Andrew McCallum's MALLET citation from 2002 and originated in an academic NLP context at UMass. The modern public repository is maintained under the mimno GitHub organization, and the official site identifies it as open source under the Apache 2.0 license.

MALLET became especially visible to digital humanities, information science, and NLP users because it packaged algorithms such as Maximum Entropy classifiers, Conditional Random Fields, Latent Dirichlet Allocation, Pachinko Allocation, and Hierarchical LDA behind reusable Java APIs and shell commands. Its 2.1.0 release in 2026 was described by the maintainers as the first stable release since 2.0.8 in 2016, modernizing the build around Java 17 and Maven.

採用の歴史

MALLET's package-manager footprint is modest but important: Homebrew and MacPorts carry it for macOS users who want the historical CLI workflow without building the Java project by hand. The GitHub repository shows roughly a thousand stars and hundreds of forks, which is consistent with a long-lived research toolkit rather than a fast-moving application.

The toolkit's adoption has been strongest where reproducible text-analysis workflows matter: topic-modeling corpora, classroom NLP exercises, digital humanities projects, and research code that wants a known implementation rather than a cloud service.

使われ方

The command-line workflow centers on `bin/mallet [command] --option value`, with helper commands for importing text or CSV data, converting data to vectors, training classifiers, classifying new text, and running topic models. The README recommends building with Maven or using binary downloads, then running `bin/mallet --help` or a command-specific help option.

In packages such as Homebrew, MALLET matters because it turns a Java research toolkit into ordinary terminal commands like `mallet`, `text2vectors`, `vectors2topics`, and `csv2classify`, making it scriptable in shell pipelines and notebooks.

パッケージ好きにとっての重要性

MALLET is a classic example of a research toolkit that became infrastructure for package-manager users: old enough to have stable workflows, specialized enough that vendoring it is annoying, and still useful when a project needs local NLP processing without a server stack.

For package curators, the interesting maintenance points are Java runtime requirements, binary wrapper scripts, and the 2026 shift to Java 17 and Maven. Those details directly affect whether older MALLET recipes keep working on current systems.

タイムライン

• 2002: Andrew McCallum's official MALLET citation identifies the toolkit.
• 2016: MALLET 2.0.8 became the last stable release before a long quiet period.
• 2021: A serialization-breaking 202108 release changed internal collections and licensing details.
• 2026: MALLET 2.1.0 shipped as the first stable release since 2.0.8, requiring Java 17.

Related projects

• GRMM was documented as an add-on package for inference in graphical models and CRF training with arbitrary graphical structure.
• MALLET overlaps historically with NLP libraries and topic-modeling stacks, but remains distinctive because it exposes mature algorithms as local Java command-line tools.

ソース

• https://github.com/mimno/Mallet
• https://github.com/mimno/Mallet/releases
• https://mimno.github.io/Mallet/
• source_facts.package-manager