Linnan juhlat 2025 alkaa: aika, lähetys, kutsut ja esiintyjät

Let me also check the gates more.

**GATE 2 — CONTRACT COVERAGE:**
The contract has `must_cover_questions` from JSON. But I don’t see that in the provided HTML content. The HTML covers:
– “Mihin aikaan Linnan juhlat alkavat 6.12.2025?” — yes
– “Ketkä saavat kutsun Linnan juhliin ja milloin kutsut tulevat?” — yes
– “Kuka esiintyy ja juontaa Linnan juhlissa 2025?” — yes
– “Mitä Linnan juhlat ovat ja mikä on niiden historia?” — yes
– “Mitä Linnan juhliin pukeutua – pukukoodi” — yes

The contract JSON has a list with each h2 and must_cover_claims. I need to check if all those claims are covered. For example, in the first h2 “Mihin aikaan Linnan juhlat alkavat 6.12.2025?” the must_cover_claims include “Linnan juhlat 2025 alkaa 6. joulukuuta 2025.” etc. The HTML seems to cover them. But I need to compare: the contract for the first H2 has claims like “Linnan juhlat 2025 alkaa 6. joulukuuta 2025.” — the HTML says “Linnan juhlat 2025 alkaa 6. joulukuuta, ja suora lähetys…” Good.

Also, the JSON includes a list of `must_cover_claims` for each h2. I need to ensure all are present. Let me quickly scan:

For first h2, claims:
1. “Linnan juhlat 2025 alkaa 6. joulukuuta 2025.” — HTML has “Linnan juhlat 2025 alkaa 6. joulukuuta” — yes.
2. “Suora lähetys Presidentinlinnasta alkaa Yle TV1:ssä klo 18.45.” — HTML has “suora lähetys Presidentinlinnasta käynnistyy Yle TV1:ssä ja Yle Areenassa kello 18.45” — yes.
3. “Lähetys suoratoistona Yle Areenassa alkaa samaan aikaan.” — HTML has “Yle Areena aloittaa suoran lähetyksen samaan aikaan klo 18.15” — but wait, the HTML says 18.15 for Areena, but the claim says “samaan aikaan” (with TV1 18.45). Actually, the HTML has contradictory information: “suora lähetys Presidentinlinnasta käynnistyy Yle TV1:ssä ja Yle Areenassa kello 18.45” and later “Lähetys alkaa TV1:ssä ja Areenassa klo 18.15” — there’s a conflict. But the contract says for gate 4, I need to check against verified facts.

The research notes include facts like: “Ylen mukaan Linnan juhlien suora lähetys alkaa TV1:ssä ja Areenassa tavallista aiemmin, heti iltauutisten jälkeen klo 18.15.” So the claim “lähetys alkaa klo 18.15” is sourced. But the claim “suora lähetys alkaa klo 18.45” is also there. Actually, the contract says “Lähetyksen aloitus Yle TV1: klo 18.45” in stats line. The research notes claim “Yle kertoi, että lähetys alkaa vuonna 2025 puoli tuntia tavallista aikaisemmin.” So 18.15 is the network start, and 18.45 is when the actual reception starts? The article says: “Yle TV1:llä suora lähetys alkaa poikkeuksellisesti jo klo 18.15, mutta varsinainen juhlavastaanoton lähetys käynnistyy klo 18.45.” So both are true. That’s fine.

But for gate 4, I need to check verified facts. The input JSON includes a list of claims with `confirmed_claims` etc. But that’s in the JSON after the article. However, the article itself has claims with source anchors. I’ll treat the JSON as research data and not necessarily the authoritative verified facts. The gate says: “Parse verified_facts from the input JSON. For each entry with iso_date_or_value: if HTML has conflicting value for same entity+attribute, replace with verified value.” The JSON has a “facts” array with claims, but no “iso_date_or_value” field. Actually, the JSON structure includes a list of claims with source_url etc. But the gate says “GATE 4 — FACT LOCK VS VERIFIED_FACTS”. I need to see if there is a `verified_facts` key in the JSON. Looking at the provided JSON, there is a key `”verified_facts”`? Wait, the JSON includes `”facts”` array with claims. But the research notes mention “research_confidence: low”. The JSON also has `”facts”` with claims and source info. But I don’t see an explicit “verified_facts” key. However, the instruction says “Parse verified_facts” — I’ll treat the “facts” array as containing claims with `source_tier` and `confidence`. Since there is no `iso_date_or_value` field, I’ll skip date-based replacement. But I should check for conflicts.

Given that the research confidence is low, I should be careful. The article has a lot of claims with sources.

**GATE 5 — FAQ DEDUP:**
Check FAQ items: There are 5 FAQ items (details). The first one is “Mihin aikaan Linnan juhlat alkavat?” which is a variant of H2 “Mihin aikaan Linnan juhlat alkavat 6.12.2025?” — that’s 80% overlap? Actually, let me compare: H2: “Mihin aikaan Linnan juhlat alkavat 6.12.2025?” vs FAQ: “Mihin aikaan Linnan juhlat alkavat?” — lexical similarity high. But the FAQ item is a broader version. Since the FAQ item is identical to the H2 except for the date, it might be considered a duplicate. But the gate says “Remove FAQ item ONLY if >85% lexical overlap with H2/H3 OR exact same question.” Let’s calculate overlap: “Mihin aikaan Linnan juhlat alkavat 6.12.2025?” vs “Mihin aikaan Linnan juhlat alkavat?” — token overlap: “Mihin aikaan Linnan juhlat alkavat” is in both, but the second has an added date. Overlap ratio = common tokens / total tokens in first? Using word overlap: common words: “Mihin”, “aikaan”, “Linnan”, “juhlat”, “alkavat” — that’s 5 of 7 in first (excluding 6.12.2025). Actually first has “6.12.2025” as one token. So common tokens = 5, total in first = 7 (including date and question mark? no punctuation). So overlap ≈ 71% for the first, and for the second: common=5, total in second=6 (since “6.12.2025” is one token in first but not in second). Wait, second has 6 tokens: “Mihin”, “aikaan”, “Linnan”, “juhlat”, “alkavat”, “?” but ignore punctuation. So common=5, total=6? Actually second has “Mihin aikaan Linnan juhlat alkavat” = 5 tokens plus question mark. So overlap = 5/6 ≈ 83% for second, and for first = 5/7 ≈ 71%. Both below 85%, so it’s kept. Good.

But the gate says “Remove FAQ item ONLY if >85% lexical overlap”. Since it’s below, keep.

Also, gate says “Minimum 5 FAQ items” — we have 5, good.

**GATE 6 — LINK HYGIENE + DIVERSITY:**
Internal links: The article has one internal link: `https://mediaportaali.fi/tekniikka/vappu-helsinki-2025-tapahtumat/` — that’s an internal domain? The research notes mention internal link plan. The article has a link at the bottom: `Vappu Helsinki 2025` — actually, the HTML has: `Lisää luettavaa: ` — this is invalid. There’s also a stray `` after the URL? Actually, the HTML shows: `` — yes, that’s good. But I see some links don’t have `rel`? For example, the link in the intro? No, intro has no links. There are links like `` — good.