LoginSign Up
← Back to all sparks
A

audubon

ANALYTICS
Velocity0.0

Japanese Text Processing Tools

github.com/paithiov909/audubon ↗

audubon's release feed is almost entirely Renovate bumping the JavaScript toolchain behind its Japanese text splitter.

japanese-nlptext-processingr-packagebudouxdependency-maintenance
◆Current state
An R package for Japanese text processing — normalisation, tokenisation via MeCab and SudachiPy, and phrase splitting through budoux. Ten releases since 2022, but the changelogs are dominated by automated dependency updates to a webpack, babel and prettier toolchain, because the budoux component is JavaScript that has to be bundled. Actual R-facing changes appear in perhaps one release in three.
◆Where it's heading
The package appears feature-stable and in maintenance. The last substantive R-level addition visible here is bind_lr() for bigram LR values back in 0.5.0; everything since has been dependency hygiene, a tokeniser refactor, and platform-specific test fixes. That is a reasonable end state for a wrapper whose value is the binding rather than ongoing invention, but it does mean the release feed carries almost no signal about the package itself — a reader watching this feed would learn more about webpack's version history than about Japanese text processing.
◆Prediction
Expect the Renovate cadence to continue setting the release rhythm, with R-facing changes arriving only when budoux itself gains capability or a platform breaks.

◆Recent moves

  1. 3mo ago

    M1 Mac locale crash worked around in examples

    Works around a locale-related crash in examples on Apple silicon and takes two automated dependency bumps. Platform maintenance with no change to the package's behaviour.

    View source ↗
  2. 7mo ago

    audubon 0.6.2

    A release whose entire body is a link to the version comparison, with no described changes. Nothing here indicates what shipped.

    View source ↗
  3. 7mo ago

    Automated dependency bumps, including a webpack security update

    A full release of Renovate-authored dependency updates across babel, webpack, prettier and budoux, one of them flagged as a security fix. Representative of what the majority of this package's releases contain.

    View source ↗
  4. 2y ago

    budoux bumped to 0.6.2; Renovate configured

    Updates budoux and adopts Renovate for automated dependency management — the change that set the pattern for every release since. The underlying phrase-splitting library moving forward is the only user-relevant part.

    View source ↗
  5. 3y ago

    MeCab and SudachiPy tokenisers refactored

    Refactors tokenize_mecab and tokenize_sudachipy with no described behaviour change, alongside dependency bumps and a test fix. Internal work the user would not observe.

    View source ↗
  6. 3y ago

    bind_lr() computes LR values for bigrams

    Adds bind_lr() for calculating the LR statistic of bigrams and makes pack() return a tibble consistently. The last R-facing feature addition in this window, which is what makes the dependency-bump releases that followed so conspicuous.

    View source ↗