/compare

name: compare
description: "Compare several Japanese NLP libraries, models, or datasets for a keyword (a specific tool name, or a function/task like '形態素解析') across a handful of criteria chosen for that comparison, rendered as a ○/△/✕ table. Use when the user wants a side-by-side comparison of multiple Japanese NLP tools/libraries/datasets, not just the single best one. Trigger phrases include 'X と Y と Z を比較して', '形態素解析ライブラリを比較', 'MeCab と Sudachi どっちがいい', 'どのツールを使うべき', 'compare japanese tokenizers', 'mecab vs sudachi vs janome', 'which embedding model should I use', '日本語NERライブラリの比較表', 'pros and cons of japanese OCR tools'. For a single ranked list use search; for alternatives to one specific tool (or contribution candidates) without a multi-axis table, use discover."
argument-hint: [tool-name | topic]
allowed-tools: Bash WebSearch WebFetch

Compare Japanese NLP resources for the user's query across a few criteria, as a table.

Claude Code and Codex

This skill is shared by the Claude Code and Codex versions of the plugin. The steps are the same in both tools; only these details differ:

• **Query** — Claude Code: the arguments of `/awesome-japanese-nlp-resources:compare`, appended at the end of this skill as `ARGUMENTS: …`. Codex: the user's message that invoked `$awesome-japanese-nlp-resources:compare`, minus that `$…` mention. If the skill was picked automatically rather than invoked by name, use the user's request as the query.
• **Plugin root** — Claude Code: `${CLAUDE_PLUGIN_ROOT}`. Codex: the directory two levels above this `SKILL.md` (use its absolute path).
• **Shell** — run the commands below with Claude Code's `Bash` tool or Codex's shell tool. Copy each Python script in full and run it as written, changing only its placeholders (`RESOURCES_PATH`, `SEED`, the keyword list) — don't shorten it, drop passes, or alter its scores and thresholds.
• **Web** — Claude Code: `WebSearch` to search and `WebFetch` to read a page. Codex: the built-in web search tool (search, then open the page). Do not use the `gh` CLI, `curl`, or other network commands from the shell.
• **Commands** — write any command you show the user in the current tool's form: `/awesome-japanese-nlp-resources:<skill>` in Claude Code, `$awesome-japanese-nlp-resources:<skill>` in Codex.

Candidates come from the bundled data (Step 3), topped up from the web only as Step 4 describes. If the data file can't be read (for example, shell commands are blocked or fail to start), say so and link https://github.com/taishi-i/awesome-japanese-nlp-resources instead of building the comparison from memory alone.

Instructions

Step 0 — Validate input

If the query is empty or blank, **stop immediately** and output (in Codex, write the commands with `$` instead of `/`):

Usage: /awesome-japanese-nlp-resources:compare <tool-name | topic>

Examples:
  /awesome-japanese-nlp-resources:compare mecab
  /awesome-japanese-nlp-resources:compare 形態素解析
  /awesome-japanese-nlp-resources:compare japanese sentence embedding models
  /awesome-japanese-nlp-resources:compare OCR

Pass a tool name (to compare it against its closest alternatives) or a topic/function (to compare the leading options for that task).

---

使い方: /awesome-japanese-nlp-resources:compare <ツール名 | トピック>

例:
  /awesome-japanese-nlp-resources:compare mecab
  /awesome-japanese-nlp-resources:compare 形態素解析
  /awesome-japanese-nlp-resources:compare 日本語 文埋め込み モデル
  /awesome-japanese-nlp-resources:compare OCR

比較したいツール名(その代替と比較)、またはトピック/機能名(その分野の主要な選択肢を比較)を引数に指定してください。

Do **not** proceed if the query is empty. Unlike `discover`, this skill has no empty-argument default — a comparison needs something to compare.

Step 1 — Classify the input: seed mode or topic mode

**Seed mode** — names ONE specific existing tool/library/model (a full GitHub/Hugging Face URL, `owner/repo`, or a bare tool name, e.g. `mecab`, `fugashi`). Pass it through as `SEED` and proceed to Step 3 in seed mode — the comparison set will be the seed plus its closest peers.

**Topic mode** — a descriptive/functional phrase with no single specific name (e.g. `形態素解析`, `japanese sentence embedding models`, `OCR`). Proceed to Step 3 in topic mode — the comparison set will be the leading local matches for the topic.

Step 2 — Locate the data file

PLUGIN_ROOT="${CLAUDE_PLUGIN_ROOT}"  # Codex: replace with the plugin root, two levels above this SKILL.md
RESOURCES_PATH="$PLUGIN_ROOT/data/resources.json"
[ -f "$RESOURCES_PATH" ] || RESOURCES_PATH="$(find "${CODEX_HOME:-$HOME/.codex}/plugins" "${HOME}/.claude/plugins" -type f -name resources.json 2>/dev/null | grep "awesome-japanese-nlp-resources/" | head -1)"
echo "RESOURCES_PATH=$RESOURCES_PATH"

Write the resulting absolute path into the Step 3 script — shell variables may not persist between commands.

The plugin also ships `data/multilingual_resources.json` (same item format) listing multilingual libraries, models, and datasets (GitHub repositories) that also support Japanese, from `docs/multilingual.md`. The scripts below load it automatically when it exists; its items have categories like `Multilingual (Speech recognition)`.

Step 3 — Find comparison candidates (inline Python)

**Do not read `resources.json` directly** (no Read tool, `cat`, or `head`). Run one of the two scripts below, substituting `RESOURCES_PATH` (Step 2) and, for seed mode, `SEED` (from Step 1).

**Seed mode** — locate the seed and score peers by shared category, shared semantic labels, and shared description tokens (IDF-weighted):

python3 << 'EOF'
import json, re, math, os
from collections import Counter

RESOURCES_PATH = "RESOURCES_PATH"   # from Step 2
SEED_RAW       = "SEED"             # from Step 1

with open(RESOURCES_PATH) as f:
    data = json.load(f)
multilingual_path = os.path.join(os.path.dirname(RESOURCES_PATH), "multilingual_resources.json")
if os.path.exists(multilingual_path):
    with open(multilingual_path) as f:
        data +=