繁體中文
香港紀事館由語言模型協助建立結構與文稿，但統計必須對回官方文本。目前主要來源是香港政府一站通、政府統計處《香港統計數字一覽》、政府新聞公報、律政司、《香港年報》歷史章及立法會時間線。
人口、面積、本地生產總值等數字會標明統計時點，以及該數字是臨時、修訂，還是長期正常值。政府網站摘要與統計處刊物如有輕微差異，條目會並列兩者，而不是 silently 選一個。
十八區採用政府統計處 2025 年陸上非住院人口，並註明此口徑與年中人口估計不同。公眾假期與法定假日分開列出，因為法律名單並不相同。入境處國籍人數是特定一天身在香港的點算，不是居民總數。郊野公園範圍內的特別地區面積已計入郊野公園，不會再加一次。歷史時間線只收官方已標明日期的節點；年報內的歷史人口約數會標明「敘述／估計」，不可與現行年中人口混用。
升學就業分成兩層：已核對條目只寫法定註冊、執照或公務員招聘文本核對過的職業，並列出可以達到同一目標的多條路徑；清單條目只列職位、工作範圍同「入行關卡係邊一類」，未核對前不會填上薪酬、學歷或年資數字。全部 400 個職業另外有「工作性質」標籤（對人服務、數據分析、體力手作、戶外、輪班、自由身等），只描述工作方式，同樣不含門檻數字。私人機構薪酬若沒有單一官方階梯，不會填上估計數字。兒童遊戲室的派籌時間以康文署場地頁為準；官方沒有公布「要提早多久排隊」，本站不另行估計。戶外共融遊樂場名單是「部分」場地，該頁沒有開放時間就不寫。
本站公開 HTML、Auto Chat（/ask）、llms.txt、llms-full.txt 及 JSON API。人用對話檢索；其他系統可先下載 catalog.json 再對題。同一個名嘅官方數字如果有兩套口徑（例如人口、病床、假期），會喺口徑對照逐組列明兩邊範圍同「唔可以相加」嘅理由，值一律由已登記 fact 派生。發現錯誤應以官方原文為準，並在本庫修正。收錄規模同每類條目的來源層級，見覆蓋範圍頁。
English
Hong Kong Story uses a language model to help structure and draft pages. Statistics still have to match official texts. The main sources so far are GovHK, Hong Kong in Figures, government press releases, the Department of Justice, the Hong Kong Yearbook history chapter, and the LegCo legislature timeline.
Figures for population, land area and GDP carry a reference date, and say whether they are provisional, revised, or long-term normals. If a government summary and a statistical booklet differ slightly, both are shown.
District pages use the 2025 land-based non-institutional population from the Census and Statistics Department, and state that this is not the mid-year estimate. General holidays and statutory holidays are listed separately because they are different legal lists. Immigration nationality counts are presence on a given day, not a resident stock. Special-area hectares inside country parks are already in the park total. History milestones are dated official nodes; yearbook population roundings are labelled as narrative estimates.
Careers are split into two tiers. Researched entries record occupations only where a statutory register, licence, or civil-service recruitment text can be checked, and they keep multiple lawful study routes; listed entries give the title, the work covered and which class of entry gate applies, and carry no pay, academic or experience figures until they are checked. All 400 occupations also carry “nature of the work” tags — people-facing, data, physical, outdoors, shift, freelance and so on — that describe working style only, again with no thresholds. Private-sector salaries are not estimated when there is no official ladder. Play-room ticket times follow the LCSD venue page; how early a queue forms is not published, so it is not guessed. The outdoor inclusive-playground list is partial; hours are omitted unless a park page publishes them.
The site publishes HTML, Auto Chat at /ask, llms.txt, llms-full.txt and a JSON API. People can ask in the chat; other systems should start with catalog.json. Where one measure carries two official bases — population, hospital beds, holidays — the metric definitions page sets out both scopes and why they must not be added, with every value derived from a registered fact. Official primary texts win over this site if they disagree; the repo should then be corrected. See coverage for the scale of the database and the source tier of each collection.
