jp-number-grounding
Verify that numbers in AI-written Japanese summaries appear in their source, and extract/normalize Japanese quantities.
jp_number_grounding_check: comparesourceandsummary(optionaltolerance) and returnok,problems,counts, and per-fact statuses such assupported,approx_supported,unit_mismatch,unsupported, and weekday mismatches.jp_number_extract: list quantities from Japanese text with kind, text, normalized value, unit, approximation flag, and issues.Normalizes kanji numerals, 万/億, full-width digits, Japanese era years, dates, 午前/午後 times, 割/分 rates, and counters; skips number-like words such as 一宮 and 統一.
Read-only, offline, deterministic, and idempotent; checks presence of values, not semantic correctness; Japanese text only.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@jp-number-grounding この議事録の要約の数字が元の文字起こしと合ってるか確認して"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
JP Number Grounding
Check that every number, amount, date and time in an AI-written Japanese summary actually appears in its source. Kanji numerals, 万/億, full-width digits, Japanese era years and 午前/午後 are normalized first. MCP server, CLI and Python library. Standard library only, no network, MIT.
Meeting minutes (議事録) and summaries are one of the most common jobs given to language models in Japan. A Japanese municipal survey counted 755 use cases for minutes summarization, second only to greeting drafts (cited here). The known failure: a model turns a proposal into a decision, or adds a deadline nobody said.
Checking the numbers by hand is slow because Japanese writes one value many ways. A script that searches the source for the digits in the summary misses most of them:
source summary naive digit search this tool
三千五百万円 3,500万円 "3500" not found supported
令和7年4月1日 2025/4/1 "2025" not found supported
午後2時 14:00 "14" not found supported
1割5分 15% "15" not found supported
参加者12名 12人 found supported
問い合わせ30件 30人 found (wrong) unit_mismatch
次回は5月18日 期限は5月1日 "5","1" found (wrong) unsupported
2025年4月1日(火) 2025年4月1日(水) found (wrong) W_WEEKDAY_MISMATCHOn the 30 bundled cases this tool flags the right cases 30/30; the naive digit search gets 14/30. The cases and the rules come from the same author, so treat this as a regression suite, not proof.
What it normalizes
kanji and mixed numerals: 三千五百万, 3千5百万, 1億2000万, 二〇二五, full-width 12
money, rates and counters: 円, ドル, %, パーセント, 割/分/厘, 人 = 名, 件, 社, 回, か月 = ヶ月 = カ月, 拠点 = か所, kg, km ...
dates: 2025年4月1日, 2025/4/1, 令和7年4月1日, 令和元年, R7.4.1, 4月1日, 4/1, 2025年4月, and the weekday in (火)
times: 14:00, 午後2時, 10時半, 午前12時
era years: 令和6年度 = 2024年度
approximations: 約, およそ, 程度, 前後 ... allow a tolerance (default 10%)
words that only look like numbers are skipped: 一宮, 三菱, 統一, 一部, 万一, 同一人物
Status per fact: supported, approx_supported, year_unverified (source gives no year), unit_unverified (one side has no unit), unit_mismatch, unsupported. ok is false when any fact is unsupported or unit_mismatch, or a weekday does not match its date.
It checks that values are present, not that they are used correctly. A summary can quote 15% from the source and attach it to the wrong item.
Related MCP server: japan-data-mcp
MCP server (for AI agents)
{"mcpServers": {"jp-number-grounding": {"command": "python", "args": ["-m", "jpnumground.server"], "cwd": "/path/to/jp-number-grounding"}}}Tool | Input | Returns |
|
|
|
|
|
|
Both tools are read-only and offline. Suggested agent rule: after writing minutes or a summary from Japanese text, call jp_number_grounding_check and fix or remove every unsupported fact before sending.
MCP protocol: the stdio server speaks both the classic initialize flow (2025-06-18 and earlier) and the stateless 2026-07-28 revision (server/discover, per-request _meta). Tested in tests/test_mcp_2026.py.
MCP の新しい版(2026-07-28・server/discover)と従来の initialize の両方で動きます。
CLI and Python
python -m jpnumground transcript.txt minutes.txt # JSON, exit 1 when a fact is unsupported
python -m jpnumground --extract minutes.txt # list quantities
python -m jpnumground.bench # score on the 30 casesfrom jpnumground import check, extract
check("予算は三千五百万円", "予算は3,800万円")["problems"]
# [{'text': '3,800万円', 'normalized': '38000000円', 'status': 'unsupported', 'issues': []}]
extract("令和7年4月1日 午後2時")[0]["normalized"] # '2025-04-01'Limits
It checks that each number, amount, date and time in the summary also appears in the source. It does not check meaning: a correct number in a wrong sentence passes.
Numbers written only in words it does not know (rare counters, dialects, slang) may be missed. Unknown forms are reported, not guessed.
Japanese text only. It is a checking aid for minutes and summaries, not a replacement for a person reading them.
Tests
python -m unittest discover -s tests # 26 tests日本語
AIが書いた要約・議事録の数字・金額・日付・時刻が、元の文章(文字起こし・資料)にあるかを確かめます。漢数字・万/億・全角・和暦・午前/午後・割/分をそろえてから比べるので、「三千五百万円」と「3,500万円」、「令和7年4月1日」と「2025/4/1」は同じとみなし、言っていない期限(unsupported)や単位の取り違え(30件→30人)、日付と曜日の食い違いを返します。値があるかを確かめる道具で、意味の正しさは確かめません。標準ライブラリだけ・外部通信なし・MIT。
License
MIT. See LICENSE.
Available Tools
2 toolsjp_number_extractExtract and normalize numbers, amounts, dates and times from Japanese textARead-onlyIdempotent
List every quantity in a Japanese text with its normalized value: kanji numerals, 万/億 units, full-width digits, Japanese era years (令和/平成/昭和, R7), dates, 午前/午後 times, 割/分 rates, counters (人, 名, 件, 社, 回, か月 ...). Skips words that only look like numbers (一宮, 統一, 一部, 万一). Returns facts with kind, text, normalized, value, unit, approx, issues.
Inputs: text (required).
jp_number_grounding_checkCheck that numbers, amounts, dates and times in a Japanese summary appear in its sourceARead-onlyIdempotent
Before you send a summary, meeting minutes (議事録), report or answer written from a Japanese source, pass both texts here. Every number, money amount, percentage, count, date and time in the summary is normalized (三千五百万円 = 3,500万円 = 35,000,000円, 令和7年 = 2025年, R7.4.1 = 2025/4/1, 午後2時 = 14:00, 1割5分 = 15%, 12名 = 12人) and looked up in the source. Returns ok, problems, counts and per-fact status: supported, approx_supported (marked 約/程度 and within tolerance), year_unverified, unit_unverified, unit_mismatch (30人 vs 30件), unsupported (not in the source: often an invented deadline or a wrong figure). Also flags a weekday that does not match its date (W_WEEKDAY_MISMATCH). Deterministic, offline. It checks presence of values, not meaning. 要約の数字・日付・金額が原文にあるかを確かめる。
Inputs: source (required), summary (required), tolerance.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 2 tools
The two tools have clearly distinct purposes: one extracts and normalizes quantities, while the other verifies a summary against a source. Despite sharing normalization logic, their boundaries are explicit and an agent can select correctly without ambiguity.
Both names follow the same predictable pattern: jp_number_ + action. The shared prefix communicates scope, and extract/check are consistent verb choices.
Two tools is slightly below the typical 3-15 range, but the server's scope is narrow and both tools earn their place. A third tool is not obviously needed for extraction and verification workflows.
The surface covers the core lifecycle for this domain: extracting normalized facts and checking a summary against a source. Minor gaps exist around semantic verification and batch workflows, but the presence-only limitation is explicitly stated and agents can work around it.
Related MCP Connectors
Exact Japanese text & date tools: char count, business days with JP holidays, era, tax, units.
Free Japanese public-data tools for AI agents: holidays, business days, wareki, tax, postcodes.
121Japan data tools for AI agents: calendar (rokuyo), address, name splitting, corporate number lookup
Deterministic signed verification of numeric & financial claims for AI agents & spreadsheets.
Related MCP Servers
- AlicenseNot gradedqualityCmaintenanceProvides Japanese tax and invoice utilities such as consumption tax calculation, withholding tax, invoice number validation, and tax rate summarization, enabling AI assistants to perform these operations locally without external APIs.MIT
- AlicenseAqualityDmaintenanceEnables AI assistants to access and compare Japanese public data from e-Stat, corporate registration, real estate, and invoice APIs, automatically converting codes into human-readable formats with sources and timestamps.1466 PyPI13MIT
- AlicenseNot gradedqualityCmaintenanceEnables an agent to verify every number in its answer against the tool outputs of the same session, storing recorded outputs and flagging fabricated values that no tool produced. It exposes check, record, and reset tools so agents can recompute, correct, or drop unvouched numbers before finalizing.MIT
- AlicenseAqualityBmaintenanceEnables AI agents to validate a Japanese qualified invoice against the six required items, reduced-rate markings, per-rate totals, and once-per-invoice-per-rate tax rounding, while flagging the common line-by-line rounding mistake, and to normalize and verify T + 13-digit registration numbers. Read-only tools also compute expected consumption tax for a given amount, rate, price basis, and rounding method.3MIT