LoginSign Up
← Back to all sparks
U

udpipe

AI-ASSISTANTS
Velocity0.0

Tokenization, Parts of Speech Tagging, Lemmatization and Dependency Parsing with the UDPipe NLP Toolkit

github.com/bnosac/udpipe ↗

udpipe's last six releases are entirely compiler fixes, with no NLP change among them.

rnlpdependency-parsingtext-miningmaintenance
◆Current state
udpipe provides tokenization, part-of-speech tagging, lemmatization and dependency parsing in R via the UDPipe library, plus a sizable set of text-mining helpers around document-term matrices and collocations. Every release in the visible window is toolchain maintenance — a bitwise-comparison warning, misaligned address and UBSan reports, dropping C++11, a C++20 declaration fix. The last functional additions were txt_grepl and the dtm_svd_similarity fix back in 0.8.8.
◆Where it's heading
This is a stable binding in custodial maintenance, released in bursts when CRAN's checks flag the vendored C++ tree — three of these releases went out within four minutes of each other. It is maintained alongside sentencepiece and nametagger, which receive the same fixes in the same sweeps, so the cadence reflects one maintainer's CRAN queue rather than demand for the package.
◆Prediction
Further compiler conformance releases are the expectation; a bump of the underlying UDPipe library or new pretrained models would be the change worth noticing, and nothing in the entries points to one.

◆Recent moves

  1. 6mo ago

    persistent_unordered_map declaration fixed for C++20

    A single C++20 fix to the persistent_unordered_map declaration, identical to the one nametagger shipped days earlier. Family-wide compiler upkeep.

    View source ↗
  2. 8mo ago

    C++11 dropped from Makevars

    Drops C++11 from Makevars, published within a minute of the matching change in sentencepiece. Build configuration only.

    View source ↗
  3. 8mo ago

    Authorship metadata corrected in Authors@R

    Adjusts authorship metadata in Authors@R for the udpipe.cpp portion. Packaging metadata with no code change.

    View source ↗
  4. 8mo ago

    Misaligned address and UBSan fixes

    Fixes a misaligned address load and UBSan messages reported by CRAN, one of three releases published within four minutes on the same day.

    View source ↗
  5. 8mo ago

    Always-true digit comparison in parse_int corrected

    Corrects an always-true overlapping comparison in parse_int, where the digit test used || instead of &&, plus documentation NOTEs. The logic error was in a code path CRAN's compiler flagged rather than one users reported.

    View source ↗
  6. 3y ago

    std::move fix for clang check warning

    Replaces move with std::move to clear an R CMD check warning on recent clang. No user-visible effect.

    View source ↗