with “Đã xác nhận” and “Điều chưa rõ”. The confirmed list has three
already has a source anchor (NIH, MedlinePlus, NHS). OK.
GATE 4 — FACT LOCK VS VERIFIED_FACTS
The input has “Verified facts (JSON): []” – empty array. So no changes.
GATE 5 — FAQ DEDUP
We have 7 FAQ items. Check if any has >85% lexical overlap with H2/H3. The H2s: “Kẽm có tác dụng gì đối với cơ thể?”, “Ai nên bổ sung kẽm?”, “Liều dùng kẽm như thế nào?”, “Kẽm có trong thực phẩm nào?”, “Kẽm có tác dụng phụ không?”. The FAQ items:
1. “Kẽm có thể uống cùng lúc với canxi không?” – no overlap.
2. “Kẽm dạng nào dễ hấp thu nhất?” – no overlap.
3. “Có nên uống kẽm mỗi ngày?” – there is H3 “Có nên uống kẽm mỗi ngày không?” under Liều dùng. That is >85% overlap? The H3 is exactly the same question but with “không?” at the end. The FAQ item is “Có nên uống kẽm mỗi ngày?” – exactly the same. So we must remove FAQ item 3 because >85% lexical overlap. The H3 already answers it. So remove that FAQ item. That leaves 6 FAQ items. Minimum 5, so fine.
4. “Kẽm có giúp mọc tóc không?” – no overlap.
5. “Kẽm có trị mụn hiệu quả không?” – no overlap.
6. “Kẽm và sắt có nên uống cùng nhau?” – there is H3 “Kẽm và sắt có nên uống cùng nhau?”? Actually the H3s under “Kẽm có tác dụng phụ không?” include “Kẽm có tương tác với thuốc gì?” which covers but not exact. The FAQ item is more specific. Overlap maybe not >85%. Keep.
7. “Trẻ em mấy tuổi có thể uống kẽm?” – no overlap.
So after removing item 3, we have 6. Good.
GATE 6 — LINK HYGIENE + DIVERSITY
Internal links: The article already has two internal links at the bottom: “Protein – Vai trò…” and “Panadol Đỏ”. They are inside a
. The internal links must strip target/rel. Currently they have no target/rel, so fine. External links: All external links have target=”_blank” rel=”noopener nofollow”. Check: the NIH links have “target=”_blank” rel=”noopener nofollow” – good. MedlinePlus, NHS also good. MEDLATEC, Nhà thuốc An Khang, Nhà thuốc Long Châu: they also have target=”_blank” rel=”noopener nofollow”. Good. Each external URL at most once: Check NIH: used many times but same URL – that’s fine, max 3 per domain? The gate says “Max 3 per domain.” But domain ods.od.nih.gov is used many times (probably >10). That violates max 3 per domain. We need to reduce to at most 3 external links per domain. The article has many NIH citations. We need to keep only up to 3 NIH links and replace others with other sources or remove. However the gate says “Max 3 per domain.” We need to comply. Let’s count NIH links: They appear in snapshot cards (3 times), table (4 times), body paragraphs (many times). We have to consolidate. We can use the same URL but different fragments? No, same domain. We must reduce to max 3 appearances of ods.od.nih.gov. But we have to keep sources for claims. We can replace some NIH citations with other sources from research notes: MedlinePlus, NHS, etc. For example, many claims about roles can be cited to MedlinePlus. We’ll do that. Similarly for nhathuoclongchau.com.vn appears twice? Actually appears once in FAQ. That’s fine. medlatec.vn appears twice (in blockquote and in body). That’s 2, under limit. nhathuocankhang.com appears once. NHS appears a few times (in table, in body). We need to check NHS domain: nhs.uk – appears at least 3 times? In table, in body for pregnancy, and in FAQ. That’s 3. So we must reduce to at most 3. Already 3, but we need to ensure not more. Let’s count: table row “Phụ nữ mang thai: 11-12 mg/ngày” has NHS link. Body paragraph under “Phụ nữ mang thai” has NHS link. FAQ item 6 has NHS link. That’s three. So ok. But NIH is over. We’ll replace many NIH links with “MedlinePlus” or “NHS” as appropriate. However the gate says “Max 3 per domain.” We can use the same URL but that still counts. So we need to reduce the number of anchor elements pointing to ods.od.nih.gov to 3 or less. Currently there are: snapshot cards (3 links), table (4 links), body paragraphs (multiple). We’ll reduce to 3 total. Keep the most important ones: e.g., the first snapshot card “Tăng cường hệ miễn dịch” (NIH), the table row for “Nhu cầu nam giới” (NIH), and one body paragraph. The rest we can replace with MedlinePlus or NHS. But we must ensure that the replaced sources are accurate. The research notes have MedlinePlus for some claims. We’ll swap.
GATE 7 — JSON-LD
Already have two