{{ help_step(1, "Automatic Scoring", "When your research encounters a journal for the first time, it is automatically scored using real bibliometric data. Scores are cached so subsequent lookups are instant.") }}
{{ help_step(2, "Tiered Data Sources", "Scoring uses a tiered approach: (1) Predatory blacklist check, (2) OpenAlex snapshot with h-index data for ~217K journals, (3) DOAJ open access status, (4) LLM analysis as a last resort for unknown journals.") }}
{{ help_step(3, "Quality Scale (1-10)", "9-10 = Elite (Nature, Science-level), 7-8 = Strong Q1, 5-6 = Moderate, 3-4 = Low, 1-2 = Predatory. Predatory journals are auto-removed from results.") }}
{{ help_step(4, "Trust Signals", "Badges show: DOAJ listing (open access verified), DOAJ Seal (top 10%), Scopus indexing, and predatory flags.") }}
{{ help_step(5, "Data Storage & Refresh (admin)", "The reference database and raw source files are stored at ~/.local/share/local-deep-research/journal_data/ on Linux/macOS, or %APPDATA%\\local-deep-research\\journal_data\\ on Windows. The data is shared across all users on this server, so a forced refresh affects everyone. To update to a newer OpenAlex snapshot: stop the server, delete the files in that directory, and restart β the next search or dashboard visit will re-download in the background. Typically an admin task; normal users don't need to refresh.") }}
{{ help_tip("Note: H-index has field-dependent bias β mathematics journals naturally have lower h-index than biomedical journals.") }}
{% endcall %}
Thanks to the open-data projects powering this dashboard
All datasets are downloaded fresh from upstream β nothing is bundled.
Journal quality scoring is built on freely-licensed bibliographic data from
OpenAlex (CC0 β sources & institutions),
DOAJ (CC0 β open-access verification),
JabRef abbreviations (CC0), and
Stop Predatory Journals (MIT).
Every dataset is fetched directly from the upstream provider on demand β please consider supporting them if you find this useful.
Journal Quality Filter
threshold = β¦
12345678910
See the How It Works tab for the full per-level breakdown.
π
Loading journal data...
Downloading...
π¬
No Journal Data Yet
Journal quality scores are automatically collected as you run research.
Start a research session with academic sources (ArXiv, OpenAlex, Semantic Scholar, NASA ADS) enabled.
Every time your research touches an academic paper, LDR resolves the source venue (journal or conference)
and assigns it a quality score from 1 to 10. The score is cached in your local DB so
the same journal is never scored twice. The Journal Reputation Filter uses these scores to
automatically remove predatory venues from your results and to surface high-quality sources first.
Tiered scoring pipeline
Each journal flows through these tiers in order. The first tier that returns a score wins, so the
expensive tiers (institution lookup, LLM) only run when the cheap ones can't answer.
Tier 0 β Local cache. Already scored in your per-user DB? Return instantly.
Tier 1 β Predatory blacklist. Match against
Stop Predatory Journals
(~2.5K entries). A hit forces score 1 and the result is dropped from your search.
Tier 2 β OpenAlex sources. ~280K journals + conferences with h-index, 2-yr mean
citedness, citation count, publisher, ISSN. Quality is derived from h-index thresholds
(β₯150 β 10, β₯75 β 8, β₯40 β 7, β₯20 β 6, β₯10 β 5, else β 4) and a Q1βQ4 quartile is computed
from cited_by_count percentiles per source type.
Tier 3 β DOAJ. ~22K verified open-access journals. DOAJ Seal (top ~10%) β 8,
regular DOAJ listing β 5. Used when OpenAlex doesn't have a row.
Tier 3.5 β Institution affiliation salvage. For preprints (arXiv, bioRxiv, β¦)
without a journal, fall back to scoring the author's institution using ~120K OpenAlex Institutions
records (h-index β₯ 250 β 6, β₯ 50 β 5, else β 4). Capped β never beats a real venue match.
Tier 4 β LLM analysis (optional, off by default). A last-resort SearXNG + LLM lookup
for journals not in any of the bundled datasets. Enable it under
Settings β Search β Journal Reputation β Enable LLM Scoring.
Results are persisted to the cache so the slow path runs at most once per journal.
Quality scale (1β10)
9β10 Elite (Nature, Science, NEJM, top conferences)
Caveat: h-index has field-dependent bias β mathematics journals naturally have a lower
h-index than biomedical journals. Quartile (Q1βQ4) is normalized within source type to partially compensate.
Trust signals
DOAJ β listed as a verified open-access journal
DOAJ Seal β top ~10% of DOAJ journals (best-practice publishing)
Scopus / Core β indexed in OpenAlex's core set
Predatory β appears on the Stop Predatory Journals blacklist and is not rescued by the whitelist (DOAJ-listed or h-index above the `PREDATORY_WHITELIST_HINDEX` threshold)
Threshold setting (how aggressive is the filter?)
Drag the slider to change search.journal_reputation.threshold directly.
Changes save automatically and apply to the next search.
12345678910
1 β Filter is effectively off. Nothing is removed (predatory journals score 1, but threshold 1 means "β₯ 1" so they pass).
2 (default) β Drop only predatory (score 1). Keep everything else, including default/unknown venues.
3 β Same as 2 (no scores fall in this gap β values 2, 3, 9 are never emitted).
4 β Same as 2 (no scores fall in this gap).
5 β Also drop default/unknown venues (score 4). Repository aggregators, garbled refs, brand-new or niche venues with no h-index data.
6 β Also drop moderate journals (score 5). Conferences and DOAJ-only OA journals start to disappear.
7 β Keep only good journals (h-index β₯ 20).
8 β Keep only strong Q1 journals (h-index β₯ 40 OR DOAJ Seal).
9β10 β Elite only. Nature, Science, NEJM tier (h-index β₯ 75 / β₯ 150). Aggressive β most search results will be filtered out.
The default of 2 is intentionally conservative: it removes flagged predatory venues
(we have positive evidence of fraud) but doesn't silently delete sources just because
we don't have bibliometric data on them.
Using the filter in research
The Journal Reputation Filter is enabled per-search-engine in
Settings β Search β Journal Reputation. It's designed for
academic engines (ArXiv, OpenAlex, Semantic Scholar, NASA ADS, PubMed). When active it:
Scores every result through the tiered pipeline above
Drops results below the threshold (default 2 = predatory only)
Records the score on the result for downstream ranking and citation export
Triggers a one-time dataset download on first use (~7 min, ~25 MB) β afterwards everything is local