Japan Diet Kokkai Proceedings Scraper — NDL Speeches

Extract speech transcripts from the National Diet Library (NDL) Kokkai proceedings API — over 1 million speeches from 1947 to the present, covering both chambers and all committees of the Japanese parliament. Returns speaker name, reading (yomi), party affiliation, position, speech text, and meeting metadata. Data is public domain under Article 13 of Japan's Copyright Law.

What does the Kokkai Diet Proceedings Scraper do?

• Queries the NDL Kokkai proceedings API for speeches from the House of Representatives (衆議院), House of Councillors (参議院), or both chambers
• Filters by speaker name (partial match), committee name, session number, and date range
• Returns full speech text plus speaker yomi (reading), party group, and official position
• Returns PDF links, speech URLs, and meeting URLs for document-level access
• Exits with zero records and a logged error if no filter is set — the NDL API requires at least one search parameter

What data does it extract?

Field	Description
speech_id	Unique speech identifier
issue_id	Meeting issue identifier
session	Diet session number
chamber	衆議院 (House of Representatives) or 参議院 (House of Councillors)
committee	Committee or meeting name
issue_number	Issue number within the session
meeting_date	Meeting date (ISO 8601)
meeting_date_wareki	Meeting date in Japanese imperial calendar format
speech_order	Order of this speech within the meeting
speaker	Speaker name in Japanese
speaker_yomi	Speaker name reading (hiragana)
speaker_group	Political party or group affiliation
speaker_position	Official position or title of the speaker
speech_text	Full transcript text of the speech
speech_url	URL to the individual speech record
meeting_url	URL to the full meeting record
pdf_url	URL to the meeting PDF where available
search_query	The search query used to retrieve this record
source_api_endpoint	NDL API endpoint used

How to use it

At least one filter (searchQuery, speakerName, nameOfMeeting, chamber, sessionNumber, dateFrom, or dateTo) is required. If no filter is set, the NDL API returns an error — the actor will log the error and exit with zero records rather than crashing.

Field	Type	Default	Description
searchQuery	string	—	Full-text search query across speech text (Japanese or romaji)
speakerName	string	—	Speaker name (partial match). Japanese characters recommended for accuracy.
nameOfMeeting	string	—	Committee or meeting name filter
chamber	string	—	衆議院, 参議院, or 両院 (both chambers)
sessionNumber	integer	0	Diet session number. 0 = all sessions.
dateFrom	string	—	Start date in YYYY-MM-DD format
dateTo	string	—	End date in YYYY-MM-DD format
maxItems	integer	—	Maximum speeches to return

Use cases

• Quantitative finance and monetary policy research — Track central bank governor testimony: set speakerName to 植田和男 for current BOJ Governor Ueda's speeches, or 黒田東彦 for former Governor Kuroda. Filter by date range and committee to isolate MPM-adjacent parliamentary appearances.
• Policy trend analysis — Search searchQuery for specific policy terms (fiscal policy, social security, defense) across sessions and chambers to track legislative debate evolution over decades.
• Political science research — Analyze party-affiliation patterns in speech frequency and committee participation using speaker_group and committee fields across the full 1947–present corpus.
• Journalism and transparency — Retrieve transcripts of committee hearings on specific legislation by meeting name and date range; link to meeting_url and pdf_url for source verification.
• Natural language processing — Build Japanese political speech corpora for NLP training using the 1M+ speech dataset — all public domain under Article 13 of Japan's Copyright Law.

FAQ

How do I filter for a specific politician's speeches? Use speakerName with the politician's name in Japanese characters. For BOJ Governor Ueda: speakerName: 植田和男. For former Governor Kuroda: speakerName: 黒田東彦. Partial name matches are supported — the NDL API searches within the speaker name field.

What happens if I run the actor without any filters? The NDL API requires at least one search parameter and returns an error for unrestricted queries. The actor detects this, logs the error with guidance, and exits with zero records rather than crashing the run. Add at least one of: searchQuery, speakerName, nameOfMeeting, chamber, sessionNumber, dateFrom, or dateTo.

Is the speech text complete? Yes. The NDL API returns the full verbatim transcript of each speech as recorded in the official Diet stenographic record. The speech_text field contains the complete text. Some historical sessions (especially early postwar) may have shorter records due to original transcription gaps in the source archive.

Results are available for export in JSON, CSV, and Excel formats from the Apify dataset tab.