YÖK Ulusal Tez Merkezi — Turkish National Thesis Search

YÖK Ulusal Tez Merkezi (the Higher Education Council's National Thesis Center) is the single mandatory repository for every Turkish graduate thesis — yüksek lisans (master's), doktora (doctorate), tıpta/diş/eczacılıkta uzmanlık (medical/dental/pharmacy specialty), and sanatta yeterlik (art proficiency). It is the authoritative source for two faculty workflows this skill serves:

1. Literature discovery — finding existing Turkish dissertations for a

review, by topic, danışman (advisor), university, or year.

1. *Pre-proposal özgünlük (originality) checks* — before a supervisor

approves a student's thesis topic, scanning whether the same work already exists or is embargoed-but-registered.

This skill is discovery + metadata harvesting + citation, not bulk full-text download (full text is online-view-only — see access model below).

When to Use This Skill

YÖK Ulusal Tez Merkezi'nde "sürdürülebilir mimari" konulu tezleri ara
Bir öğrencim "sanal prodüksiyon" üzerine doktora yapmak istiyor; daha önce yapılmış mı?
Find every doctoral dissertation supervised by Prof. X at IEU
Cite this YÖK thesis in APA 7 (Turkish)
Is this thesis topic novel, or has it already been done in Turkey?

→ Construct a Detaylı Tarama (detailed search) query, run it through the yoktez-mcp connector (or drive the form), dedupe results into a {Tez No, year, university, danışman, title, izin} table, and — when asked — emit a Türkçe APA-7 citation.

Does NOT Trigger — route these elsewhere

This skill answers "does this Turkish thesis exist / has this been done, and how do I cite it?" — it does not write the thesis, judge journals, or score careers.

The data path: yoktez-mcp

There is no official public API for Ulusal Tez Merkezi. The maintained, verified community path is the `saidsurucu/yoktez-mcp` MCP server (MIT licensed). Prefer it over scraping.

Install (uv-first, matches the local toolchain):

uvx --from git+https://github.com/saidsurucu/yoktez-mcp yoktez-mcp

Hosted connector: https://yoktezmcp.fastmcp.app/mcp

It exposes exactly two tools:

These map one-to-one onto the public Detaylı Tarama form fields: Tez Adı (title), Yazar (author), Danışman (advisor), Konu (subject), Anahtar Kelime (keyword), Özet (abstract), Tez No (thesis number).

If the connector is unavailable, the search form at https://tez.yok.gov.tr/UlusalTezMerkezi/ can be driven directly (a real browser is required — see references/endpoints.md for the gated-viewer caveat). Never invent a JSON/REST endpoint — none is published.

A small offline helper, scripts/yok_tez_query.py, builds a normalized query spec and the matching search_yok_tez_detailed argument dict from a plain description (handy for the stemming/boolean rules below) without any network call. See Search craft.

Search craft (apply these automatically)

The full rule set with worked examples is in references/search_syntax.md. The essentials:

1. Search Turkish roots, not inflected forms — the engine auto-stems Turkish,

so query sürdürülebilir rather than sürdürülebilirliğin. Diacritics (ç ğ ı İ ö ş ü) need not be normalized by the user.

1. Boolean operators are `ve` / `veya` / `içermesin` (and / or / must-not-contain),

not AND/OR/NOT. Phrase-match by keeping words adjacent.

1. Run BOTH a Turkish and an English query. English terms only hit the

English title/abstract/keyword fields, so an all-Turkish query silently misses theses written/indexed in English, and vice-versa.

1. Search across `Konu`, `Tez Adı`, and `Özet` for topic coverage — a title

may not name the concept the abstract develops.

Originality / supervision check (the repeatable recipe)

When a supervisor asks "has this thesis topic already been done?", do not run a naive open-access-only search — that misses the most important near-duplicates. Instead:

1. Build paired TR+EN queries (above) across Konu, Tez Adı, and abstract_text.
2. Constrain thesis_type (e.g. Doktora) and a recent year_start–year_end

window if the user gave one.

1. Set `permission_status` to "Tümü" (All) so İzinsiz (restricted,

abstract-only) theses still surface — embargoed work is exactly what a naive search hides, and it is still prior art for a proposal.

1. Return a deduplicated table: `{Tez No, year, university, danışman, title,

izin}`, sorted newest-first, for the supervisor to scan.

1. State explicitly that this is registry coverage, not a plagiarism / text-

similarity score (that is iThenticate/Turnitin territory, out of scope here).

Access model (set correct expectations)

• Full text is online-view-only for İzinli (permitted) theses — viewable in

the browser, no photocopy / bulk download. The get_yok_tez_document_markdown tool works only on permitted theses.

• *Abstracts (özet) are always available*, even when full text is embargoed.
• **İzinsiz (restricted) theses** show metadata + abstract only; the full text

is reachable in print via TÜBESS / inter-library loan through a university library.

• Pre-2006 closed theses can be opened by the author submitting a *Tez

Yayımlama İzin Belgesi* (thesis-publication permission document).

• Legal basis: Law 7100 Art. 10 and Higher Education Law 2547 Additional Art.

40 — theses are electronically accessible by default unless an authorized gizlilik (confidentiality) decision applies. YÖK's FAQ states no fixed maximum embargo length; do not assert a "12-month cap."

Full detail in references/access_legal.md.

Caps and bulk harvest

• Hard cap: 2000 results per search. For broad topics, split the query by

year range, university, or institute to stay under it, then merge and dedupe locally.

• For large/repeated harvests, the cleaned community mirror tezara.org

("Tez Arama ve Metaveri Analizi Platformu") offers structured export and sidesteps the per-search cap; see references/endpoints.md for what is and isn't verified about it.

• Bulk Turkish↔English query expansion, advisor-name normalization, and CSV

dedup are good fits to offload to a local LLM (qwen3-coder / gemma4) rather than burning API calls — this work is bulky, repetitive, and privacy-neutral.

Citing a YÖK thesis (Türkçe APA-7)

Map the YÖK Tez No directly to APA's Yayın No. (publication number). The scripts/yok_tez_query.py --cite mode formats both states; rules and more examples in references/citation_apa7.md.

• Published / permitted (has a Yayın No.):

Soyad, A. (Yıl). Tez başlığı (Yayın No. _NNNNNN_) [Tez türü, Üniversite Adı]. YÖK Ulusal Tez Merkezi.

• Unpublished / restricted:

Soyad, A. (Yıl). Tez başlığı [Yayımlanmamış tez türü tezi]. Üniversite Adı.

Optionally also emit BibTeX @phdthesis / @mastersthesis with note = {YÖK Ulusal Tez Merkezi, Tez No. NNNNNN}.

Self-check before reporting

• Did you run both a Turkish and an English query? An all-Turkish search is

not complete.

• For an originality check, did you set permission_status = Tümü so İzinsiz

theses surfaced? Restricted ≠ nonexistent.

• Did any broad search hit the 2000 cap? If total ≈ 2000, split and re-run —

you are almost certainly truncated.

• Are citations using Tez No → Yayın No. with the correct

published vs unpublished template?

• Did you state that registry coverage ≠ a text-similarity / plagiarism score?

References

• references/endpoints.md — verified vs. unverified

endpoints, the yoktez-mcp tool surface, the gated TezGoster viewer, tezara.org.

• references/search_syntax.md — auto-stemming,

ve/veya/içermesin, TR+EN pairing, field-targeting, worked queries.

• references/access_legal.md — İzinli/İzinsiz,

TÜBESS, pre-2006 opening, Law 7100 / 2547 legal basis, embargo facts.

• references/citation_apa7.md — Türkçe APA-7

thesis templates and the Tez No → Yayın No. mapping.

Part of the AlterLab Academic Skills suite.