AGENT SKILL

tokenize-tag-stem

Use NLTK tokenization, detokenization, POS tagging, tagger training, stemming, lemmatization, and sentiment preprocessing APIs without reopening the source repo.

License 许可
Apache-2.0
最近更新
2026-09-03
Stars
★ 351
Forks
27

01 前言 Front Matter

name
tokenize-tag-stem
license
Apache 2.0
metadata
{ "disco-role": "operating" }
description
Use NLTK tokenization, detokenization, POS tagging, tagger training, stemming, lemmatization, and sentiment preprocessing APIs without reopening the source repo.
disable-model-invocation
true

02 资源文件 5

03 SKILL.md 正文

Tokenize Tag Stem

Use this sub-skill when the task is about NLTK text preprocessing: tokenizing or detokenizing strings, running the nltk tokenize CLI, POS-tagging tokens, training simple taggers, stemming or lemmatizing words, or using VADER/basic sentiment preprocessing.

Route away from this sub-skill when the task is mainly about:

Fast Paths

  • For no-download word splitting, prefer WordPunctTokenizer, TreebankWordTokenizer, TweetTokenizer, or word_tokenize(text, preserve_line=True). sent_tokenize() and word_tokenize(..., preserve_line=False) need Punkt table data for the selected language.
  • For command-line tokenization, use nltk tokenize; it reads stdin and exposes --language, --preserve-line, --processes, --encoding, and --delimiter. The bundled CLI smoke checks the installed console script.
  • For tweets/social text, use TweetTokenizer(strip_handles=True, reduce_len=True, preserve_case=...); decide whether phone numbers should be matched with match_phone_numbers.
  • For POS tags, pass a list of tokens to pos_tag(tokens, tagset=None, lang="eng") or a list of token lists to pos_tag_sents. English uses averaged_perceptron_tagger_eng; Russian uses lang="rus" and averaged_perceptron_tagger_rus.
  • For a quick custom baseline tagger, combine RegexpTagger rules with UnigramTagger(train_sents, backoff=...); evaluate with accuracy, evaluate_per_tag, or a ConfusionMatrix when gold tagged sentences are available.
  • For stemming, use PorterStemmer(mode=...), SnowballStemmer(language, ignore_stopwords=False), LancasterStemmer, RegexpStemmer, or language-specific stemmers. Porter and most Snowball workflows do not need NLTK data.
  • For lemmatization, WordNetLemmatizer.lemmatize(word, pos="n") depends on WordNet for real morphology; download/check wordnet and omw-2.0 when multilingual or synonym-aware behavior is needed.
  • For sentiment, SentimentIntensityAnalyzer() requires vader_lexicon; route missing lexicon errors to the data/downloader sub-skill. SentimentAnalyzer can build features for a classifier workflow, which then routes to the ML sub-skill.

Reference Map

Minimum Validation Pattern

  1. Print nltk.__version__, the exact API or CLI route, and whether the workflow is allowed to download NLTK data.
  2. For tokenization, test one representative string and state whether sentence splitting is enabled (preserve_line=False) or skipped (preserve_line=True).
  3. For POS tagging, assert the input is list[str], not a raw string; for batches, assert list[list[str]].
  4. Before running data-backed APIs, probe exact resources such as tokenizers/punkt_tab/english/, taggers/averaged_perceptron_tagger_eng/, taggers/universal_tagset/, corpora/wordnet/, or sentiment/vader_lexicon.zip/vader_lexicon/vader_lexicon.txt.
  5. For custom taggers, keep training/testing data separated and report unknown-token behavior instead of hiding None tags.
  6. Run the bundled no-download smoke from any current working directory in an environment with NLTK installed:
python /path/to/skills/disco/nltk/sub-skills/tokenize-tag-stem/scripts/text_preprocess_smoke.py --json
python /path/to/skills/disco/nltk/sub-skills/tokenize-tag-stem/scripts/tokenize_cli_smoke.py --json

04 同仓库更多 Skills

VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-triton-opt

Make a single PassNet Triton kernel fast AND numerically faithful: performance model (when a replacement can win at all), block/grid/warp tuning, autotune policy, launch overhead, and per-op numeric recipes to pass the dtype baseline tolerances. Use when a pass matches and is correct but speedup below expected, or when correctness fails by small numeric margins.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-solve

END-TO-END playbook for solving one PassNet sample: analyze the computation graphs, decide the optimization strategy (what to fuse, what to replace, what to leave alone), drive the iteration loop, and maximize the sample score. This is the ENTRY skill — invoke it first for any PassNet optimization task; it tells you when to use passnet-pattern-fusion, passnet-triton-opt and passnet-feedback.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-skill

PassNet GPU kernel optimization via compiler passes. Design and implement Triton-based optimization passes, create pass files under ./pass_dir/, self-evaluate with pass_evaluator, and iterate to maximize GPU speedup.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-pattern-fusion

Author PassNet pass files: write patterns that actually MATCH the FX graph, pick fusion regions (which ops to absorb into one Triton kernel), and structure multi-pass submissions with the mandatory shared-dispatch architecture. Use when creating or fixing pattern()/replacement_args()/replacement_func() files, when a pass "failed to match", or when deciding how to fuse multiple kernels/ops.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-orchestrate

Round/batch planning and decision support for PassNet graph optimization. Use for multi-sample triage, worker allocation, eval-budget policy, and post-eval keep/revert decisions. For solving one concrete sample end to end, use passnet-solve as the entry skill; consult this skill only when you need broader planning or a specific decision gate.

★ 351
VectorSpaceLab
VectorSpaceLab/AREX-Skill

passnet-feedback

Fast, deterministic feedback for PassNet work: pre-flight pattern/match verification WITHOUT burning a GPU evaluation, per-node bottleneck analysis of a sample's graphs, and eval-log parsing into per-variant status + estimated score + failure classification. Use BEFORE every GPU evaluation (check_pattern), at the START of a sample (analyze_graph), and AFTER every evaluation (parse_eval_log).

★ 351