LargitData — Enterprise Intelligence & Risk AI Platform

Last updated:

Bringing AI into Council Interpellation: The Complete Before-During-After Workflow

Every council session sends agency staff into interpellation hell: digging through past records, compiling sentiment issues, guessing which topics members will pursue, and chasing commitments afterwards. AI has clear leverage at all three stages — pre-session issue assessment and debate preparation, real-time transcription and structuring during sessions, and post-session task tracking. This article covers the concrete practices at each stage, the data foundations required, and how to connect the three into a closed loop.

Bringing AI into Council Interpellation: The Complete Before-During-After Workflow資訊圖表配圖,呈現AI 知識中心的重點概念

Quick Answer: How Can AI Help with Council Interpellation?

AI enters council interpellation in three stages. Before the session, the system compiles members' past interpellation records, recent sentiment hotspots and policy controversies into "what might be asked and how to answer" preparation materials. During the session, speech recognition transcribes proceedings in real time and automatically structures each member's questions and the agency's commitments. After the session, commitments and assignments enter a tracking list, managed item by item and compared against subsequent sentiment. Connected, the three stages turn interpellation prep from last-minute cramming into routine, data-grounded work.

Before the Session: Issue Assessment and Debate Preparation

The core pre-session question is: who will ask what this session? AI cross-references three sources:

  • Past interpellation records:Each member's past interpellation topics, follow-up patterns and constituency concerns, to estimate the likelihood of continued questioning.
  • Recent sentiment hotspots:Media and social issue volume in the weeks before the session — controversies running hot in public sentiment almost inevitably enter interpellation.
  • Policy progress and weaknesses:Delayed projects, contested policies and items with low budget execution rates are all predictable interpellation targets.

From these, preparation materials are produced for each high-risk issue: likely question phrasings, factual and data evidence, a recommended answer structure, phrasing to avoid, and traceable sources. Staff work shifts from compiling from scratch to reviewing and supplementing.

1. 備詢資料的標準欄位

備詢資料要能在台上被翻閱,欄位就必須固定。機關若讓每個局處各自用自己的格式撰寫,長官在質詢台上得先花時間找資訊在哪一段,這正是備詢資料最常失效的原因。建議統一為七個欄位,無論議題大小、無論由哪個主政單位撰寫,版面一致、順序一致;AI 負責把外部輿情與內部知識填入前六欄的初稿,承辦人負責檢核與修正,長官只需記住欄位順序。以下是跨機關通用的七項欄位:

  • 議題名稱與風險等級:以一句話界定議題範圍,並標示高、中、低風險,決定準備深度與長官預讀順序。
  • 可能的提問方式(含追問路徑):不只列出主問題,也要推演對方在得到某種答覆後最可能的第二問與第三問。
  • 已確認事實與數據依據:只放經主政單位確認過的事實與統計,未確認的資訊另行標註,避免在台上被追問來源時失據。
  • 建議答覆架構:先講結論、再講依據、最後講下一步,三段式架構讓答覆在時間壓力下仍然完整。
  • 不宜使用的說法:列出容易被斷章取義、與既有立場衝突或超出權責範圍的措辭,這一欄往往比正面稿更關鍵。
  • 相關歷史承諾與辦理情形:把過去在議事場合承諾過的事項與目前進度並列,避免出現前後不一致的答覆。
  • 原始資料連結:每一項事實都要能一鍵回到原始新聞、公文或報表,讓長官與幕僚可以自行查證。

2. 開議前準備時間軸

備詢準備失敗多半不是能力問題,而是啟動太晚。把準備工作往前拉,並且切成四個檢查點,就能讓資料在開議當天已經過至少兩輪人工檢核。以下是建議的時程安排,機關可依會期長度調整,但四個檢查點的先後順序建議維持:

  • 開議前一個月:彙整會期間的輿情熱點與歷史質詢紀錄,交叉比對後產出高風險議題清單,並指定每個議題的主政單位。這個階段的重點是涵蓋率,寧可清單長一點,也不要在開議後才發現漏項。
  • 開議前兩週:各主政單位依七欄位格式完成備詢資料初稿,並完成第一輪人工檢核。AI 產出的事實與數據必須逐項核對原始資料,確認無誤才進入下一階段。
  • 開議前一週:更新最新的輿情變化,補入這段期間新爆發的議題,並調整既有議題的風險等級。輿情在開議前一週往往最活躍,這一輪更新不能省。
  • 開議前一日:產出當日重點提醒:今日排定的質詢單位、對應的高風險議題、昨日至今的輿情異動,控制在一頁以內,供長官與隨行幕僚快速掌握。

During the Session: Real-Time Transcription and Structuring

The pain point during sessions is record-keeping: oral exchanges move fast, manual notes miss things, and reviewing recordings afterwards is slow. Speech recognition (ASR) with meeting intelligence tools transcribes interpellation verbatim in real time and auto-structures it into a member–question–answer–commitment list. For the team on duty, real-time structuring has a tactical value too: back-row staff can push relevant materials to the official at the podium before the member's follow-up lands.

1. 即時轉錄的技術要點

議事場景是中文語音辨識最困難的情境之一,導入前必須先認清三個挑戰。第一是專有名詞:人名、地名、法案名稱、機關簡稱與預算科目在通用模型的訓練資料中出現頻率低,容易被辨識成同音的日常詞彙。第二是台語與華語夾雜:質詢過程中語言切換頻繁,且常出現同一句話中夾雜兩種語言的情況。第三是多人交錯發言:質詢與答詢互相打斷、旁邊還有議事人員的說明,語者分離的難度高於一般會議。實務上的解法是建立機關詞庫,把該機關的組織名稱、業務術語、正在推動的計畫名稱與常出現的法規名稱匯入,讓辨識模型在解碼時優先採用;同時對關鍵段落保留人工校對流程。要理解辨識準確率如何量測,可參考ASR 語音辨識的模型架構與 CER 指標說明。需要特別強調的是,即時逐字稿的定位是會中掌握與會後整理的工作稿,正式紀錄仍應以人工確認版為準。

2. 幕僚即時支援工作流

會中真正拉開差距的,是後排幕僚的反應速度。傳統做法是幕僚憑記憶判斷議題、再翻紙本資料,等到找齊往往已經錯過追問的時間點。導入即時逐字稿後,工作流改為:系統辨識出當前討論的議題關鍵字,自動比對會前備妥的備詢資料,把對應那一頁推送到後排幕僚的螢幕,幕僚確認後再遞交給備詢長官。同時,系統在會中就把內容整理成議員、提問、答覆、承諾的結構化清單,承辦人只需在旁修正與補註,散會時清單已接近完成,不必再從錄影檔重聽一次。

After the Session: Commitment Tracking and Outcome Comparison

"We'll look into it," "we'll provide the data after the session" — commitments made at the podium, if not systematically managed, become next session's follow-up interpellation: "you said you'd handle it — did you?" Post-session, AI does two things: extracts commitments and assignments from the transcript into a tracking list with lead units and deadlines, and keeps comparing sentiment on related issues so the agency knows whether its response actually resolved the concern.

1. 承諾事項追蹤清單的欄位

承諾追蹤清單之所以常常失效,是因為只記了一句摘要。摘要無法在下個會期還原當時的語境,也無法回答「到底承諾到什麼程度」。清單建議固定六個欄位,讓每一筆承諾在半年後仍然可以被完整重建:

  • 承諾內容原文段落:保留逐字稿原文而非摘要,並附上時間戳記,必要時可回放對照。
  • 主政單位:明確到局處層級,跨單位事項要指定一個主辦、其餘列為協辦。
  • 承辦窗口:記錄職稱與分機而非僅記姓名,人員異動時清單仍然可用。
  • 辦理期限:若台上未明確給出期限,由主政單位在會後三日內自訂並回填。
  • 目前狀態:以研議中、辦理中、已函復三種狀態列管,狀態變更需留下時間紀錄。
  • 對應輿情變化:追蹤該議題在承諾後的討論聲量與情緒走向,判斷回應是否確實化解關注。

這份清單就是下個會期的必考題庫。開議前一個月產出高風險議題清單時,凡是狀態仍為研議中或辦理中、且輿情聲量未下降的項目,應直接列為高風險;已函復但輿情仍有爭議的項目,也要備妥後續說明。換句話說,會後的列管品質直接決定下一輪會前準備的起跑點——清單維護得好,下個會期的準備工作等於已經完成一半。

一個會期的 AI 支援全貌

把三個階段攤在同一條時間軸上,可以看出 AI 與人工的分工邊界始終一致:AI 負責蒐集、比對與初稿產出,人工負責確認、判斷與對外表述。這條界線在導入初期就要跟各局處講清楚,避免承辦人誤以為系統產出的資料可以直接上呈,也避免長官誤以為所有內容都已經過查證。下表以一個完整會期為例,列出各時間點的 AI 產出與對應的人工工作,機關可以直接拿來當作導入時的分工說明。

階段 時間點 AI 產出 人工工作
議題盤點 會前一個月 輿情熱點與歷史質詢交叉比對,產出高風險議題清單與風險等級建議 確認議題範圍、指定主政單位、調整風險等級
資料撰擬 會前兩週 依七欄位格式產出備詢資料初稿,附原始資料連結 逐項核對事實與數據、補充內部辦理情形、刪修不宜使用的說法
每日更新 開議期間每日 前一日輿情異動摘要、新增議題提示、當日質詢單位對應資料 判斷是否需要臨時加開議題、決定長官預讀順序
現場支援 質詢當日 即時逐字稿、議題關鍵字比對、備詢資料推送、承諾事項初步標記 確認推送內容正確、遞交長官、現場修正標記
交辦列管 會後一週 萃取承諾與交辦事項,建立含主政單位與期限的追蹤清單 確認承諾範圍、指定承辦窗口、核定辦理期限
持續追蹤 休會期間 持續輿情比對,標示聲量未下降或再度升高的列管項目 定期檢視清單狀態、更新辦理情形、決定是否提前處置

地方議會與立法院:導入設計的差異

閉環架構通用,但資料設計必須依層級調整。地方議會的議題重心落在選區:一條路的施工進度、一件陳情案的處理速度、一處公共設施的維護狀況,這些議題在地方社團與在地媒體的擴散速度極快,往往在正式陳情之前就已經形成輿情。立法院的議題重心則落在制度:法案的審查進度、預算凍結與解凍的條件、跨部會權責分工,討論場域集中在全國性媒體與政策社群,議題週期較長但牽動面較廣。兩者對追蹤欄位的需求也不同——地方議會的列管重點是個案是否確實解決、承諾的完工時間是否達成;立法院的列管重點則是條文修正的進度、報告的提出時程與跨部會協調的狀態。機關在設定議題分類、輿情來源權重與追蹤欄位時,應依自身面對的層級調整,下表整理四個主要差異面向:

面向 地方議會 立法院
核心議題型態 選區個案與陳情案,與在地輿情高度連動 法案審查進度與跨部會分工,制度面議題為主
質詢型態 縣政總質詢與部門質詢,口頭答詢比重高 委員會與院會質詢,另有書面質詢與函復程序
追蹤重點 個案是否解決、承諾的完工或處理時間是否達成 條文修正進度、報告提出時程與跨部會協調狀態
輿情來源側重 地方社團、在地媒體與區域性論壇 全國性媒體、政策社群與專業評論

Connecting the Three Stages into a Closed Loop

Each stage delivers value on its own, but the real payoff is the connection: the post-session commitment list is the "guaranteed exam questions" for next session's preparation; session transcripts accumulate into historical data on members' attention patterns; pre-session sentiment assessment decides which issues need live support during the session. This loop is the five-stage AI copilot model — sense, understand, assess, recommend, track — unfolded in the parliamentary setting.

On data foundations: pre-session assessment needs external sentiment monitoring (such as InfoMiner) and historical parliamentary data; in-session transcription needs Chinese speech recognition; post-session tracking needs to connect with the agency's internal case-management mechanisms. For adoption, pilot with a single session and a single committee, validate the workflow, then scale.

FAQ

The challenges in parliamentary settings are proper nouns (names, places, bill titles) and code-switching with Taiwanese. In practice, the recognition model is customized with the agency's term dictionary, key passages keep human proofreading, and the official record remains the human-confirmed version. The purpose of real-time transcription is live awareness during the session and fast structuring afterwards — not directly replacing the official proceedings record.
AI assessment aims at coverage, not hit rate: making sure every high-risk issue is prepared, not precisely predicting each question. In practice, a high-risk list built from past records plus sentiment heat covers most actual interpellation topics; the real value is letting staff spend limited time on the issues needing the deepest material, rather than spreading effort evenly across every possibility.
The workflow is the same; the data emphasis differs. Local councils focus more on constituency issues and the sentiment around petition cases, while the Legislative Yuan involves tracking bill progress, caucus negotiations and cross-ministry assignments. Issue taxonomies and tracking fields need adjusting by level, but the before-during-after closed-loop architecture is universal.
備詢攻防資料應比照機敏文件保護。實務上有三層設計:以角色權限控管將存取範圍限定在幕僚圈,一般同仁看不到攻防內容;所有查詢與下載都留有紀錄,可事後追溯;對敏感度最高的機關可選擇地端部署,資料不出機房。另一個常被忽略的重點是——監測設定與研判結果本身就是機敏資訊,機關關注哪些議題、把哪些項目列為高風險,同樣需要納入權限控管。
可以。公開的議事錄與媒體報導本身就足以建立第一版基礎,不需要等內部資料完全數位化才啟動。紙本紀錄可透過 OCR 逐步數位化後納入,依會期由近而遠回溯即可。建議的起步方式是先以近期輿情研判與公開議事資料開始運作,歷史資料隨著每個會期的逐字稿與追蹤清單自然累積;通常在兩到三個會期之後,機關內部的資料量就足以支撐更精準的議題研判。

Make Next Session's Preparation Data-Grounded

Contact us about pilot programs and scenario demos for AI-supported council interpellation.

Contact Us