Searching and the DOI lookup

Two tools find a paper: the search, over the papers that have a page (those with their authors' code, with code on request, or with data only), and the DOI lookup, which answers for every paper OSCR has read, with or without a page.

The search box on every page

The box in the dark band at the top of every page searches the catalogue. Its menu chooses where: all fields, the title, the DOI, the journal, or the code repository (an address such as github.com/owner/name, or a part of it). Pressing Search opens the search page with your words; the menu's choice becomes a field of the query language below (title:(…), for example), which you can then refine.

The search page

The search page runs a search only when you ask: when you submit the form, choose a filter, change page or open a shared address. Typing never searches — every search counts against a daily quota shared by every reader (below). The address is the whole state of a search (/search/?q=eeg&modality=eeg&sort=newest): copy it to share or keep a search; Back and Forward replay the searches you made.

The query language

Words are searched in the titles, keywords, MeSH terms, authors, journals, code repositories, tools, identifiers and, when the paper's licence allows it, abstracts. A result must contain every word.

A query holds at most 500 characters, 24 words or phrases and 4 prefixes. What cannot be read as written — an unclosed quote, a prefix too short — is corrected, and the page says how. Words are matched whole, ranked with the title first, then identifiers, keywords, MeSH terms, authors, repositories and tools, the journal, and the abstract last.

Abstracts are searched only when the paper's licence is CC BY, CC0, CC BY-SA or CC BY-NC, the same rule that decides whether its page shows them (the licence policy). A result never shows an abstract or an excerpt: a title, a journal, a date, a status and links.

Advanced search

The page's "Advanced search" is a form — all of these words, this exact phrase, any of these words, none of these words, in the title, an author, a journal, a tool used in the code, a keyword or MeSH term, a code repository, an identifier, publication dates and a status — that writes the query in the language above and shows it before it runs, so you can learn the language from it.

Filters, dates and sorting

Beside the results (below them on a phone), the filters list the values found among them, with their counts: the status, the year, the modality, organism, population and subfield (the taxonomy's categories), the tools found in the code, the code's languages, the journal, where the datasets are, where the code is (GitHub, Zenodo, OSF…), the code's licence family, the paper's type and licence family, whether matches were computed, and whether the paper is open access. Choosing two values of one filter finds either; choosing values of two filters finds papers with both.

Dates take a year, a month or a day (2020, 2020-03, 2020-03-15), both ends included. Results are sorted by relevance when there are words, by date otherwise; you can choose newest, oldest or most cited.

Results, pages and exports

• Results come 20 to a page, as the catalogue lists papers: by day when sorted by date, ranked otherwise, with the words you searched highlighted.
• A search looks at its first 500 results. When more papers match, the total says "more than 500", the filters' counts are those of the first 500, and the pages stop there: narrow the query, a filter or the dates.
• The links under the results download the first 500 results as CSV or JSON: DOI, title, journal, date, status, the paper's page, its code repositories and their licences, its datasets and its citation count.

The daily quota

The search runs on Cloudflare's free plan, which allows a fixed number of dynamic requests a day for the whole site, and a fixed number of database rows read. When they are spent, the page says it plainly: the search has used its daily quota, try again tomorrow (the quota starts again at midnight UTC). Browse, the lists, every paper's page and the DOI lookup keep working: they are static files, which cost no request. A search you repeat — Back, Forward, the same link — is served by your browser for ten minutes without asking again.

Other messages: "unavailable" when the database does not answer (try later), "not available yet" before it is set up.

The DOI lookup

The lookup answers for every paper OSCR has read in its scope, including the many without a page (no code, or only metadata read). Paste a DOI — with or without https://doi.org/ — and it says whether the paper was read, on which day, what was found, and links to its page when it has one. A DOI it does not know was not read: either the paper is not open access in Europe PMC, or the harvest has not reached it yet, or it is outside the scope. You can submit it.

The lookup runs in your browser: it fetches one small file among 256 (chosen by the DOI's SHA-1), costs no request to the site's quota, and works when the search's quota is spent. It needs JavaScript.

Other ways to find a paper

• Every paper, by date, a hundred to a page, and the home page's latest days.
• Browse: the categories, and the lists of authors with an ORCID iD, journals, institutions, tools found in the code and datasets cited. Each has its own page, listing its papers.
• A paper's page lists up to ten similar papers, by the tools, categories, datasets, references and authors they share.