Skip to main content
Manusights homeLibrary

Methodology and corpus

A shortlist is useful only when its evidence and limits are inspectable.

This page documents the live venue corpus, how candidate ordering works, the evidence rule every result must pass, what is retained, and what the Journal Fit Predictor has not been validated to predict.

Live pipeline

Validate, order, ground, disclose.

  1. 01

    Validate input

    Require a substantive abstract and reject placeholder, promotional, or instruction-injection text.

  2. 02

    Load the fixed corpus

    Provide 1,321 candidate records with names, observed-topic summaries, research fields, and source identifiers.

  3. 03

    Order candidates

    A language model compares the submitted title and abstract with those corpus records and proposes five candidates.

  4. 04

    Ground every result

    The server rejects a candidate unless its quoted evidence is an exact phrase found in that venue record.

  5. 05

    Return a decision record

    The interface presents rank, rationale, quoted evidence, source link, index flags, and portable exports.

Important semantic boundary: The Journal Fit Predictor orders candidates from corpus topic overlap. It does not calculate acceptance probability, predict editorial decisions, evaluate manuscript quality, or label venues safe, realistic, or aspirational.

Live corpus

One disclosed candidate universe.

The live build contains 1,291 journals and 30 conferences. Journal records originate from OpenAlex; conference records are curated because conferences are primary venues in parts of computer science.

Journal inclusion requires an ISSN and substantial OpenAlex publication and citation history. This favors established venues and leaves genuine specialist or newer journals outside the corpus. Absence is not a judgment about quality.

The scope field is synthesized from topics observed in the venue's publication history. It is useful for retrieval, but it is not copied from the official aims and scope. Authors must inspect current publisher pages and recent articles before submitting.

The DOAJ flag describes a listing in the source snapshot. It does not certify every article, current status, editorial standards, or suitability for a specific manuscript.

The tool does not show MEDLINE status. An earlier flag was built from the NLM list of journals cited in PubMed, which is not the list of journals currently indexed for MEDLINE, so we withdrew it. Check MEDLINE status in the NLM Catalog.

Download live JSONDownload live CSV

Data dictionary

What every public field means.

slug
Stable Manusights venue identifier.
name
Venue name in the source snapshot.
scope
Synthesized summary of observed publication topics. Not the official publisher aims and scope.
fields
Up to five topic labels derived during corpus construction.
kind
Journal or conference.
source
OpenAlex for journals or curated for conferences.
external_ref
OpenAlex source identifier or conference source URL when available.
indexed.doaj
Snapshot flag indicating a DOAJ listing. This is not an acceptance or quality prediction.

Version record

The DOI archive and live corpus are not the same artifact.

The current web tool uses corpus edition 2026.04.27-live2, built 2026-04-27, with 1,321 venues.

The Zenodo record 10.5281/zenodo.19831122 archives version 1.1.0, which contains 1,030 venues. It is a citable historical snapshot, not the data currently queried by the web tool.

The versioned source repository is available on GitHub. A future deposition is required before the live 1,321-venue build can claim DOI coverage.

Refusal states

Unsupported certainty is a product failure.

Short or empty abstract
Return a validation error rather than manufacture a ranking from inadequate text.
Non-abstract or prompt injection
Reject content that does not describe a research study or attempts to control the ranking instruction.
Ungrounded model evidence
Discard a candidate whose quoted excerpt cannot be found in the selected corpus record.
No grounded candidates
Return an error instead of silently replacing unsupported results with plausible-looking venues.
Weak corpus overlap
Show a coverage caution. This is an internal ordering signal, not a calibrated confidence probability.

Data handling

Exact behavior, not a vague privacy promise.

The title and abstract are sent to the configured language-model provider to produce the shortlist. Provider-side API handling is governed by that provider's applicable terms. Manusights does not use submissions to train its own model.

A SHA-256 content hash and computed result may be cached server-side for up to seven days. Saved tool-run history keeps input length and whether a title was supplied, but not the raw title or abstract. Operational email does not contain manuscript text.

The browser stores the current title, abstract, result, and share token in session storage so the page can restore the run in that browser tab. Closing the session clears that browser-scoped copy under normal browser behavior.

Share links contain computed shortlist data, not the submitted title or abstract. Anyone with a share URL can view its result, and share pages are marked noindex.

Validation status

Useful retrieval, not a validated predictor.

Manusights has not published a prospective benchmark against editor shortlists, author decisions, or eventual submission outcomes. Rank quality, inter-rater agreement, field-level recall, and calibration remain unproven.

A credible validation study would pre-register a time-split sample, define field coverage, compare against independent expert shortlists and strong baselines, report top-k recall and disagreement, and publish excluded cases. Until that exists, the tool must be described as a shortlist workspace.

Good alternatives provide useful comparison points: JANE documents its PubMed-similarity method; B!SON documents its open-data matching and source updates; publisher finders disclose the portfolio they search.

How to cite the live tool

Name the version and corpus edition you actually used.

Manusights. (2026). Journal Fit Predictor (version 2.0, corpus edition 2026.04.27-live2) [Web application]. https://manusights.com/tools/journal-fit (accessed YYYY-MM-DD).

Keep going

What do you need help with next?

Find another practical resource or follow a guided publishing path when you want deeper support across the process.