Answers to the
questions editors
actually ask.

A keyword search returns reviewers whose profile strings match yours. Reviewer Select extracts the underlying concepts from your abstract — methods, models, biological systems, instruments — and matches them against a 1024-dimensional vector space of reviewer expertise. The result: adjacent specialists, methodological fits, and interdisciplinary matches you would never surface with boolean logic.

A transformer language model fine-tuned on academic literature across STM disciplines. It returns two structured outputs per manuscript: subject areas (broad disciplines) and weighted concepts (0.00–1.00 importance scores). Both feed directly into ranking — editors can review and refine concepts before search runs.

Milvus stores the article and concept embeddings as indexed vectors (IVF_FLAT + HNSW hybrid). Approximate-nearest-neighbour search returns the most semantically aligned articles in under 200 ms across 8.4 M reviewer vectors — fast enough for an interactive editorial workflow.

Yes — and you should. Concepts can be removed, re-weighted, or added (free text or curated taxonomy). The quality of extraction directly determines the quality of matching, and the editor is always in the loop. Edits also feed back into your journal's preference model over time.

From the time the abstract is pasted to when a ranked, evidence-backed shortlist is displayed: under 60 s. Concept extraction averages 2–4 s, vector search runs in ~200 ms, and reviewer aggregation + scoring completes within 30 s for typical query sizes. The editor’s review of concepts is the only step that varies..

Remarkably well. Concepts are first extracted independent of any single ontology, then resolved against a multi-discipline vector index. For a manuscript spanning two or three fields, you receive distinct reviewer clusters — one for each disciplinary lens — so you can compose a panel that covers every angle of the work.

Add them by ORCID, DOI, or full name and inspect their concept overlap directly. If the overlap is genuinely thin, the panel shows why and suggests adjacent reviewers from the same citation neighborhood. Section 7 of the results view (View Unmatched Results) lists articles below the threshold for further diagnostics.

Reviewer profiles are re-embedded from indexed publications and pre-prints on a weekly basis. New ORCID records and recently published authors enter the pool within 7 days of indexing. For your own reviewer database, the ingestion runs nightly via the integration API.

Concept extraction supports 60+ languages and all major STM disciplines as well as humanities and the social sciences. The vector index spans 8.4 M reviewers across 220+ disciplinary clusters. Manuscripts in non-English languages are matched against multilingual embeddings — no need to translate beforehand.

Optionally, and only as a tunable input. Never as a hidden weight. Editors can configure a seniority floor (e.g. post-doc and above) or a prestige weighting tied to journal-level metrics. But the default ranking is concept driven. Every score breakdown shows exactly what contributed. Institutional prestige, h-index, career stage, and network proximity never silently influence rank order — they appear as filterable informational metrics only.

Through three signals: shared institutional affiliations within a configurable lookback window (default 3 years), prior co-authorship (any depth), and declared conflicts of interest from ORCID and PubMed disclosures. Each conflict of interest surfaces on the reviewer card before the invite, never after. Editors can also upload their own exclusion lists.

Yes. Reviewer Select is GDPR-compliant, processes only publicly-available bibliographic data plus what your organization explicitly uploads. It also supports right-to-erasure requests. Data residency in EU, US, or UK regions is selectable per tenant. An SOC 2 Type II report available under an NDA.

Yes. Reviewer Select is GDPR-compliant, processes only publicly-available bibliographic data plus what your organisation explicitly uploads, and supports right-to-erasure requests. Data residency in EU, US, or UK regions is selectable per tenant. SOC 2 Type II report available under NDA.

Yes. Every score is decomposable: editors see the contributing concepts, their weights, the matched articles (with DOIs), and the cosine similarity of each. Every rank change in a session is logged for audit. Enterprise tenants receive monthly bias and coverage reports across protected demographics.

Editorial Manager, ScholarOne, Editoria, and OJS via native integrations. A REST API + webhooks handle anything else — shortlist export, invite-status sync, and rank refresh on revisions. The integration guide ships with sample payloads and a sandbox environment.

Yes. Upload a CSV/JSON of internal reviewers, link them to ORCID, and they're embedded alongside the public corpus. You can also operate in strict mode, where only your reviewers are returned — public results disabled.

Subscription by submission volume, with tiered packages for single journals, mid-size portfolios, and enterprise publishers. Funding agencies and university research offices have dedicated plans tied to call cycles. Request a quote with your annual submission count and the team will get back the same day.

A 30-day pilot with one of your active manuscript queues is standard. The team configures journal-specific policies, ingests your reviewer pool, and runs side-by-side comparisons against your existing process. No commitment until pilot review.

Every tenant gets an editorial success lead who has worked in scholarly publishing. Setup, integration, training, and quarterly model-review calls are included. Enterprise plans add a 24-hour SLA, a private model fine-tune option, and on-site workshops.