This task classifies segments of the Australian Federal Parliament Hansard. Each input is a single Hansard segment as defined by the XML structure (e.g. a speech, interjection, question, or reply). CLASSIFICATION GOAL For each segment, estimate how the speaker uses their time across the secondary categories below. Think holistically about the entire segment. Ask yourself: What is the speaker trying to ACHIEVE with this segment? What functions is this segment serving (policy explanation, partisan attack, abuse, process, PR, ceremony, etc)? Roughly what share of the segment is serving each function defined by the secondary categories? You are not labelling sentences. You are identifying how to attribute the function of the segment across the secondary categories provided below. SECONDARY CATEGORIES {secondary_block} ASSIGNMENT RULES For each segment, you MUST: SECONDARY CATEGORY SHARES Provide a numeric share for EVERY secondary category ID. All shares must be between 0 and 1. The shares must sum to 1 (up to normal floating point rounding). If a category is not present at all, assign it 0. OMITTED IDS If you omit any secondary category ID in your JSON, it will be interpreted as having share 0. It is safer to include all IDs explicitly. RATIONALE Provide a short natural language rationale explaining: which secondary categories carry substantial share, and how you decided to divide the segment across categories, referring to the segment's functions (not just quoting the text). ASSIGNMENT PRINCIPLES CLASSIFY FUNCTION, NOT TOPIC Ignore what policy area is being discussed. Focus on what the speaker is trying to do (explain policy, attack opponents, enforce rules, tell a feel-good story, deliver a tribute, etc). THINK IN MODES, NOT SENTENCES Read the entire segment first. Identify the main function of the text. For each secondary category, ask: "How much of the segment's overall function is aimed at doing THIS?" WEIGHT BY SPACE AND PROMINENCE A function that appears in many sentences should usually get more weight. However, a short but central attack or slogan can get meaningful weight if it dominates the tone and purpose. Combine: how much text fits a category, and how central that category is to the overall purpose of the segment. RESOLVE OVERLAPS USING FUNCTIONAL TIE BREAKS When the segment mixes functions, allocate more share to functions that are: more frequent across the segment, and more central to the speaker's overall purpose. SPECIAL CLASSIFICATION GUIDANCE SHORT SEGMENTS (<10 words) Default to interjection: unless the segment contains explicit Chair/Speaker language ("Order!", "The member will resume their seat", formal rulings), assume it is a member interjection (abuse_disorderly_interjections). One-word or fragmentary utterances ("Pork-barrelling", "Legislated", "Shame!") are almost always disorderly interjections, not process. "Resume your seat" or similar from a backbencher directing speech at another member is abuse, not process - only the Chair enforces procedure. If the function cannot be inferred at all, hedge across multiple interpretations. PROCESS vs OTHER CATEGORIES Process categories (4A, 4B, 4C) are NARROW. They cover the mechanics of running parliament, not the substance of what happens in parliament. Process is the MEDIUM, rarely the FUNCTION. Ask: "Is the speaker's purpose to manage parliamentary procedure, or are they using procedural forms to do something else?" Reserve process categories for: Actual points of order about rules, relevance, or behaviour Division calls, quorum calls, scheduling motions Formal stage announcements ("The question is...", "Bill read a second time") Chair/Speaker rulings and directions Do NOT use process for: Questions in Question Time (even though they follow procedural forms) Segments that happen to mention procedure as part of another function Speeches that discuss how parliament should work (that's policy or theatre) Members directing each other (that's abuse or theatre) POLITICAL THEATRE vs POLICY WORK When a segment contains both partisan framing AND policy content, weight the FRAMING over the CONTENT. The key question: "Why is the speaker explaining this policy?" If the answer is "to attack opponents" or "to claim credit", then political theatre dominates even if substantial policy detail is present. Policy work dominates only when: The segment is genuinely explanatory or educational in purpose Scrutiny focuses on implementation details rather than blame Evidence is presented to inform rather than as ammunition If adversarial party positioning, slogans, or electoral framing dominate and policy detail serves to support the attack, give more weight to Political theatre categories (pt_*). LEADING QUESTIONS (DOROTHY DIXERS) Government backbenchers asking friendly questions to ministers serve multiple functions simultaneously. The question format is procedural, but the PURPOSE is typically to create an opportunity for political theatre - credit-claiming or opponent-bashing. Consider what the question sets up: if it invites the minister to highlight achievements, weight toward pt_credit_claiming. If it invites contrast with opponents, weight toward pt_partisan_attacks_blame. The policy content is often a vehicle for these purposes rather than genuine inquiry. The question itself (as distinct from the answer) typically has significant process weight because it must perform the procedural function of being a question, but this should not dominate. ABUSE CATEGORIES Direct personal abuse of individuals should always contribute to Abuse & hostility categories (abuse_*), even when embedded in policy or partisan content. abuse_direct_personal: attacks on named individuals' character, competence, integrity abuse_generalised_contempt: hostile dismissal of groups ("those opposite", "this rabble") without substantive content abuse_disorderly_interjections: heckling, shouted interruptions, fragments that disrupt rather than argue If a segment is primarily a disruptive interjection, use abuse_disorderly_interjections even if the words could theoretically be partisan (e.g., shouting "Pork-barrelling!" at a speaker). OUTPUT FORMAT For each segment, you MUST respond with a SINGLE LINE of valid JSON. No markdown, no questions, no commentary, no extra keys. Below is an example of an output, using the variable names provided above: {"secondary_category_shares":{"policy_substantive_explanation":0.0,"policy_implementation_scrutiny":0.0,"policy_evidence_expertise":0.2,"pt_partisan_attacks_blame":0.5,"pt_credit_claiming":0.0,"pt_electioneering_campaign":0.0,"abuse_direct_personal":0.3,"abuse_generalised_contempt":0.0,"abuse_disorderly_interjections":0.0,"process_points_of_order":0.0,"process_divisions_quorums":0.0,"process_formal_announcements":0.0,"ceremony_member_community_pr":0.0,"ceremony_tributes_condolences":0.0},"rationale":"Short explanation of how the segment was divided across categories."} RESTRICTIONS The JSON MUST be syntactically valid. Do not include trailing commas. Do not include comments. All numeric values must be bare numbers (no strings). The output MUST be a single line. If the text is empty or non-substantive, still produce valid JSON with a secondary distribution that sums to 1, concentrating on the most appropriate background categories. {extra} here are the categories id: policy_substantive_explanation parent_id: policy_work label: 1A Substantive policy explanation description: Explaining what laws, regulations or programs do, how they work, why they are designed that way, and what impacts or values they are intended to advance (including technical detail). inclusion_criteria: • Describes the content of a bill, regulation, scheme or program. • Explains how something will work in practice. • Sets out intended impacts or values. • Provides legal or technical detail to make policy understandable. exclusion_criteria: • The main function is party blame or praise with minimal policy detail (use 2A or 2B). • The main function is personal abuse (use 3A). • The content is purely procedural formula with no explanation (use 4C). • The content is pure PR or ceremony with no concrete policy description (use 5A or 5B). id: policy_implementation_scrutiny parent_id: policy_work label: 1B Implementation & performance scrutiny description: Examining how well existing laws, programs, agencies or services are working in practice, including delivery problems, capability, delays, failures or successes. inclusion_criteria: • Describes outcomes or performance of an existing policy or service. • Points to delivery gaps or failures. • Highlights practical constraints or capability issues. • Evaluates success or failure in practice. exclusion_criteria: • The main function is to blame or praise a party without real implementation detail (use 2A or 2B). • The content is about design or intentions rather than real-world performance (use 1A). • The content is insult or contempt without clear reference to performance (use 3*). • The content is about votes, motions or scheduling rather than real-world implementation (use 4*). id: policy_evidence_expertise parent_id: policy_work label: 1C Evidence & expertise description: Presenting or discussing data, research, inquiries, statistics, expert bodies or independent reports as support for a policy position or evaluation. inclusion_criteria: • Cites statistics or quantitative evidence to justify a policy stance. • Refers to independent reports, reviews or inquiries as reasons to act. • Quotes or summarises expert opinion to support or oppose a policy. • Uses historical or factual context clearly as part of a policy argument. • Where both policy explanation and evidence are present, weight this category according to how central the evidence is. exclusion_criteria: • Facts or history are listed with no clear link to a policy argument (usually 5A or 5B, depending on context). • Numbers are used mainly as partisan ammunition (use 2A or 2B). • There is no genuine evidence, only assertion or slogans (use 1A or 2* depending on function). id: pt_partisan_attacks_blame parent_id: political_theatre label: 2A Partisan attacks & blame description: Criticising another party or parties primarily to damage them, with limited substantive policy detail. inclusion_criteria: • Blames another party for negative outcomes in a broad or sloganistic way. • Describes opponents as untrustworthy, chaotic or extreme in a party-focused frame. • Uses partisan talking points with little concrete policy content. exclusion_criteria: • The main function is to attack individuals personally (liar, incompetent, clown etc) (use 3A). • There is genuine performance scrutiny with concrete implementation detail (use 1B). • The main focus is campaigns, elections, voting appeals or polls (use 2C). id: pt_credit_claiming parent_id: political_theatre label: 2B Credit-claiming & self-congratulation description: Praising one’s own party or government for achievements or outcomes, often with a promotional or celebratory tone. inclusion_criteria: • Claims party or government credit for outcomes. • Presents a positive story about the party’s record or values. • Spins events to maximise credit for one’s own side. exclusion_criteria: • The focus is mainly the local member or electorate, not the party (use 5A). • The tone is neutral and focused on substantive impacts rather than praise (use 1A or 1B). • The content is clearly part of campaign messaging or election appeals (use 2C). id: pt_electioneering_campaign parent_id: political_theatre label: 2C Electioneering & campaign talk description: Campaign style language primarily about elections, voting or campaign themes. inclusion_criteria: • Explicitly tells people how to vote or describes the electoral choice. • References campaign slogans or manifesto themes as such. • Talks about polls, electoral prospects or campaign dynamics as the main point. exclusion_criteria: • Elections are mentioned only as factual background to policy discussion (use 1A or 1C). • The content is mainly partisan blame or praise outside an explicit election frame (use 2A or 2B). • The content is personal abuse unrelated to elections (use 3*). id: abuse_direct_personal parent_id: abuse_hostility label: 3A Direct personal abuse description: Direct attacks on specific individuals’ character, integrity, competence or intelligence, including mocking nicknames and ridicule. inclusion_criteria: • Calls a named member or clearly identified person a liar, corrupt, dishonest, etc. • Demeans individual competence or intelligence. • Uses mocking nicknames or ridicule directed at a person. • Is mainly insult rather than argument. exclusion_criteria: • The target is a broad group such as “those opposite” or “this rabble” (use 3B or 2A depending on context). • There is real policy criticism with only incidental sharp language (use 1* or 2*). • The content is a short shouted disruption with no clearly identified individual target (use 3C). id: abuse_generalised_contempt parent_id: abuse_hostility label: 3B Generalised contempt description: Hostile or contemptuous language directed at groups, parties or institutions as a whole, without substantive policy content. inclusion_criteria: • Describes “those opposite”, “this rabble” or similar in contemptuous terms with no substantive argument. • Dismisses a party or institution as a whole. • Uses inflammatory generalisations about groups rather than specific people. exclusion_criteria: • There is concrete performance or policy criticism alongside the hostility (use 1B or 2A). • The hostility is targeted at specific individuals rather than groups (use 3A). • The unit is really just partisan blame framed as party versus party with some content (often 2A). id: abuse_disorderly_interjections parent_id: abuse_hostility label: 3C Disorderly behaviour & interjections description: Heckling, shouted abuse, disruptive comments or unintelligible interjections recorded in the text, whose main function is to disrupt, inflame or insult rather than argue. inclusion_criteria: • Short interjections captured in Hansard as disruption. • Outbursts that interrupt another speaker. • Crowd noise, heckling or similar brief fragments. • Unintelligible, indistinct or inaudible interjections clearly representing disturbance (for example [inaudible interjection], [indistinct], [members interject]). exclusion_criteria: • The text is clearly part of a continuous speech, even if hostile (use 3A or 3B). • The content is a formal point of order or ruling about behaviour or procedure (use 4A). id: process_points_of_order parent_id: process_rules label: 4A Points of order & rulings description: Objections and decisions about relevance, language, behaviour or procedure, focused on enforcing or interpreting rules. inclusion_criteria: • Raises a point of order. • Responds to a point of order or gives a ruling. • Calls for withdrawal of remarks or addresses behaviour under standing orders. exclusion_criteria: • The “point of order” form is used to make a policy or partisan point (use 1* or 2*). • The content is a simple call or formula with no rule element (use 4C). • It is just disruptive shouting with no procedural content (use 3C). id: process_divisions_quorums parent_id: process_rules label: 4B Divisions, quorums & scheduling description: Calling and managing divisions, quorums, adjournments, order of business and allocation of time. inclusion_criteria: • Calls or announces a division or quorum. • Announces results of divisions. • Manages scheduling or adjournments. • Refers to the order of business or allocation of time in a procedural way. • If the main topic of discussion is order of business and allocation of time exclusion_criteria: • The content makes a substantive case for or against changing the schedule (use 1* or 2*). • It is a pure stock formula marking a bill stage rather than scheduling or results (use 4C). id: process_formal_announcements parent_id: process_rules label: 4C Formal announcements & bill progress description: Standard formulae and announcements that move items along or mark bill or amendment stages, without substantive policy explanation. inclusion_criteria: • States the formal question or stage. • Announces that a bill has passed or reached a stage. • Uses standard formal wording to move from one item to another without additional content. exclusion_criteria: • There is any real explanation or argument about the bill or motion (use 1* or 2*). • The content is about rules or behaviour rather than stages (use 4A). • The content is ceremonial or tribute phrasing rather than business progression (use 5B). id: ceremony_member_community_pr parent_id: ceremony_recognition label: 5A Member & community PR description: Positive stories and recognition that primarily promote the member or their community, without a clear policy demand or scrutiny function. inclusion_criteria: • Highlights the member’s presence or activity. • Celebrates local achievements and events in a feel-good tone. • Tells constituent or community stories mainly for positive sentiment. • Provides neutral background or factual context as part of a PR-style narrative, not a policy argument. exclusion_criteria: • The story is clearly used to argue for or against a policy, or to demand action (use 1B or 1C). • The focus is party-level branding or record, not the member or community (use 2B). • The tone is solemn tribute or condolence for deaths or national events (use 5B). id: ceremony_tributes_condolences parent_id: ceremony_recognition label: 5B Ceremonial tributes & condolences description: Obituaries, condolences, acknowledgments of country, honours and other formal or solemn commemorations. inclusion_criteria: • Expresses condolences or sympathy for deaths, disasters or tragedies. • Offers formal tributes or recognises service or honours. • Gives acknowledgments of Country or similar ritual statements. • Speaks about days of remembrance or solemn national occasions in a ceremonial tone. exclusion_criteria: • The tribute is mainly used to push party spin or self promotion (use 2B or 5A). • The focus is the member’s role at a ceremony rather than the person or event being honoured (often 5A). • The content moves into substantive policy argument about what should be done (use 1*).