Methodology
How ManaGraf works — and where it stops
ManaGraf is a research workbench for evidence-based study of Qur'anic concepts. Its single principle: usage is the arbiter of meaning. The system compiles and visualizes evidence and attaches raw data to every claim; judgment always belongs to the researcher.
Lineage of the method
Farāhī — naẓm and coherence
Words mean within the Qur'an's own weave. In ManaGraf this becomes context-first views: every occurrence is shown with its verse.
Izutsu — semantic fields
A concept carries a worldview together with its neighboring concepts. In ManaGraf this becomes the field graph built from co-occurrence and embedding neighborhood — edges persist only with human approval.
Akdemir — diachronic view
Meaning moves through time. In ManaGraf this becomes revelation-chronology views: Meccan/Medinan distribution and the sura-by-sura course.
Four limits — stated plainly
These methods are powerful but bounded. ManaGraf does not hide the bounds; each panel's ⓘ marginalia states the relevant one.
Root fallacy
Etymology does not determine meaning in use (Barr's critique). Hence etymology sits in a background panel; usage evidence always leads.
Circularity risk
Deriving a 'worldview' from a word-map can smuggle the mapmaker's judgment into the map. Field-graph candidates are therefore only candidates; relation verdicts are human.
Corpus limits
Revelation order and Meccan/Medinan classification are disputed; morphological tagging is human-made. ManaGraf uses one scholarly tradition (Tanzil chronology + QAC) transparently.
Ideological narrowing
Reducing concepts to a single political-theological frame kills the method. The entry template therefore requires a 'counter-readings' section.
Source inventory (closed-world)
| Source | Role | License/Note |
|---|---|---|
| Qur'an text + Turkish translation | Verse texts (6,236) | internal |
| Quranic Arabic Corpus v0.4 | Root/lemma morphology (~130K words) | GPL terms, attributed |
| Tanzil sura metadata | Meccan/Medinan + order | verbatim redistribution |
| Tanzil translation collection (8) | Translation comparison matrix | non-commercial (workshop) |
| al-Mufradāt — Rāghib al-Iṣfahānī (d. 502H) | Classical Qur'anic lexicon (1,695 entries) | public domain (OpenITI edition) |
| Lisān al-ʿArab — Ibn Manẓūr (d. 711H) | Classical Arabic lexicon (9,153 entries) | public domain (OpenITI edition) |
| Hadith corpus — the Six Books + Riyāḍ al-Ṣāliḥīn | Related narrations + hadith search (22,891) | Arabic originals public domain; no authenticity grading shown |
| Ṣaḥīḥ Asbāb al-Nuzūl (Ibrāhīm M. al-ʿAlī) | Occasion-of-revelation reports (witness layer; grading cited from the source) | modern compilation — workshop use; re-evaluated before public launch |
| Tafsir corpora — Elmalılı (YEK) · al-Rāzī · al-Ṭabarī · Ibn Kathīr · al-Qurṭubī | Verse/concept tafsir witnesses (~33K passages) | public-domain originals; matching is semantic |
| Locally produced verse embeddings | Similarity neighborhood | locally produced |
Citation policy
AI produces drafts only, and every sentence must carry a (sura:verse) or [Source: ...] citation; drafts with more than 10% uncited sentences are auto-rejected. No AI text publishes without human approval. Every compilation carries a provenance chain of its source versions.
ManaGraf is not a verdict machine. It offers well-ordered evidence to researchers who ask good questions — the rest is scholarly discipline.