What MySCI does

Every known Malaysian journal, with its publisher, owning institution, ISSNs and lifecycle position. 0 journals, of which 0 have a proven OAI endpoint.

Proves an endpoint answers, which formats it can actually deliver, and records the diagnostics — including where the repository is off-spec.

Derives candidate OAI URLs from a journal's homepage and proves them. Refuses any URL that addresses a whole OJS installation rather than one journal.

Samples an endpoint's records and refuses it if their setSpec belongs to another journal. What a repository says about itself is not evidence of what it serves: one endpoint declared six sets, all its own, and answered with records from twenty-three other journals. 0 of 0 endpoints have been proved single-journal; 0 were refused and disabled.

Full and incremental harvests with opaque resumption tokens, raw payload retention, idempotent upsert and resumable runs.

oai_dc, NLM and JATS, composed so a richer format fills gaps without a sparser one blanking what it cannot express. 0 journal(s) are harvested in a format that carries affiliations.

Why Almost every OJS installation here offers only Dublin Core, and several advertise JATS while returning an error when asked for it. Verification harvests the richest format a repository will actually deliver, not the richest it lists.

Each article states which source, format and harvest it came from. Raw payloads are retained and can be reprocessed without contacting anyone.

DOI registration checks and disagreement detection. Never overwrites publisher metadata.

Reference lists, external identifiers and citation counts, stored as evidence with their source.

Matches affiliation strings to institutions, and declines rather than guessing when the match is not clear.

Why Only formats that carry affiliations can be resolved at all, and Dublin Core — what almost every journal here offers — has no element for one. The ceiling is the corpus, not the matcher: affiliations exist for 0 of 0 authorships. Unresolved strings queue for a curator rather than being guessed at.

Confirming author identity against the ORCID registry.

Why An ORCID in publisher metadata is a claim by the publisher, and checking it against the registry is worth doing — but only once there are enough of them to matter. Almost no article in the corpus carries one, because oai_dc cannot express it.

Author names and affiliations exactly as the publisher supplied them, with contact details removed from what is displayed.

An authorship is linked to a person only on identifier evidence. Most are unlinked, which is the correct state.

Why Names alone are never sufficient here: patronymics, inconsistent romanisation and shared name stems make name matching unreliable, and a wrong merge attributes one researcher's work to another.

Automated disambiguation of researchers across the corpus.

Why Requires identifier coverage the corpus does not yet have. Built on names alone it would merge distinct researchers, and those merges are hard to detect and harder to undo.

Recognition by MyCite, Scopus and others recorded as separate claims with the source each was read from — never merged with MySCI indexing.

CiteScore, SJR, JIF and JCI values, where a source has supplied them.

Why Stored but never displayed. These are licensed measures belonging to other organisations, and republishing them is not ours to do.

Per-field completeness, measured only for fields the source format can express, with no composite score.

Core, Enhanced and Citation Ready levels, assessed per journal from what its metadata actually contains.

Every journal carries the level its newest assessment reached, so the register can be filtered, sorted and counted by what its metadata actually contains. 0 of 0 journals holding articles have been assessed.

Why Measuring one journal answers a publisher's question. Which journals meet the requirements, which do not, and what is stopping them is a different question, and it could not be asked at all without loading a snapshot per journal.

Requiring a journal to reach a metadata quality level before MySCI will index it.

Why The assessment that would drive it is built; the gate is not switched on. Two decisions are outstanding: which level becomes the bar, and what happens to journals already indexed below it. No journal assessed so far has reached Enhanced, so that bar applied today would remove every journal in the index.

Deferred by the project owner — “we can set it up later”.

A composite score and tier for each journal.

Why A single number that ranks journals would be read as a quality verdict on the scholarship. MySCI has not defined what it would measure, and publishing one before that is decided would give the number an authority nobody intended.

Every entry in an article's bibliography, resolved to an indexed article where one matches.

Why No publisher in the index deposits references yet, so the ones held here came from OpenAlex. Each records which source supplied it, so publisher-deposited references can be told apart when they arrive.

Publisher deposit deferred by the project owner — “we will get it later”.

Links between two articles MySCI indexes, shown on article pages as navigation a reader can follow.

Why 0 relationships exist so far, because both ends must be indexed here and the corpus is small. Coverage is the constraint, not the method.

How many works cite a given article.

Why Counts over the MySCI corpus and counts from external aggregators measure different things over different corpora, and must never be added. No field anywhere holds a combined total. The counts are stored and accumulating history so that a measure, once designed, has a trend behind it.

Held back by the project owner — “leave all the metrics for now, we are going to design our own later”.

Collaboration and citation networks across the corpus.

Why A network drawn over the 0 journals harvested so far would describe those journals, not Malaysian scholarship, while looking like it described the latter.

A rebuildable search index projected from PostgreSQL, with versioned indices, an alias switch and count validation before anything goes live.

Boolean operators with brackets, phrases, field search, year ranges, wildcards and genuine ordered and unordered proximity — parsed to a syntax tree, never passed to the engine as raw text.

British and American spellings find each other, and singulars find plurals — without altering a single stored title or abstract.

Malay words are protected from English stemming, so a search for a Malay term returns the records that contain it.

Why Malay does not form plurals with a final “s”, so the English plural stemmer was truncating real words — turning “kes” (case) into “ke” (to), which is one of the commonest words in the language. Routing by the language field is not possible, because publishers overwhelmingly declare Malay articles as English. Instead a curated lexicon of 149 Malay words, drawn from this corpus, is held back from the stemmer. A Malay word outside the lexicon is still truncated; a real Malay analyzer is a separate decision.

Filters across bibliographic, journal-status, metadata and citation dimensions, with a bibliometric summary of exactly the filtered set.

Retrieving conceptually related work when the terminology differs, and fusing that with lexical ranking.

Why The sequencing condition has been met — lexical retrieval is correct and tested, so a semantic layer would no longer be hiding an error underneath it. What blocks it now is the deployment: this OpenSearch has no k-NN plugin, so a vector field cannot be stored, and no embedding model is installed. Both are decisions about what MySCI runs, not code that can be written around them. Advanced Boolean search would keep its own path regardless: it must never be silently expanded.

h-index, co-authorship networks, bibliographic coupling, co-citation and fractional counting over a result set.

Why Needs worker infrastructure and canonical-side computation rather than search aggregations. The instant tier — counts, distributions, coverage — is built and returned with every search.

Journals, articles, authors, institutions and publishers, with full-text search, exact DOI lookup and phrase matching.

Read-only JSON over the whole corpus, with OpenAPI documentation generated from the code.

Staff queues for decisions MySCI declined to make on its own, ordered by how many records each one settles.

What is indexed, what limits coverage, and how every measure on the site is calculated.