Guides

Document filters

The vertical-specific filters for patents, news, funding rounds, job openings, research papers, market reports and trends.

Each document category takes its own small set of filters, nested under filters like any other.

  • None of them support geo, TextFilter or the company keywords shape. Those are company/person only.
  • Every date here is a DateFilter, wrapping its bound under period.

patent#

FilterShapeNotes
keywordsKeywordQuery
publication_date, application_dateDateFilterPublication and filing are different events
statusA bare stringgranted, not_granted or all (default). Not a filter object
application_typeTermsFilterpatent, utility, utility_model, design, plant, provisional, pct, spc
officesTermsFilterUSPTO, EPO, JPO, CNIPA, KIPO, UKIPO, DPMA, INPI, CIPO, IPA, INPI-BR, ROSPATENT, IPOS, IPI, WIPO, PCT, EAPO, OAPI, ARIPO, GCCPO
classification_codesTermsFilterCPC/IPC codes with prefix semantics — H01M matches its subgroups. match_all is honoured
applicantsEntityFilterThe applicant as named on the document
applied_by_companyEntityFilterThe resolved company the applicant maps to, so a domain matches filings under any of its legal names
application_ids, publication_ids, family_numbersTermsFilterExact identifiers
  • utility and utility_model are deliberately distinct: a utility patent is an ordinary patent, a utility model is a much narrower instrument.
  • WIPO and PCT both resolve to the same authority; ROSPATENT spans two.
{
  "category": "patent",
  "filters": {
    "offices": { "include": ["US"] },
    "publication_date": { "period": { "gte": "2024-01-01" } },
    "status": "granted"
  }
}

news_article#

FilterShapeNotes
keywordsKeywordQuery
published_dateDateFilter
publicationsEntityFilterThe outlet — an entity, because outlets need resolving ("TechCrunch" vs techcrunch.com)
startupsEntityFilterCompanies the article is about
countries, citiesTermsFilterThe place the article discusses, not the outlet's home. Matched exactly, so casing and spelling must match what is stored

funding_round#

FilterShapeNotes
funding_sizeRangeFilterNormalised USD; there is no currency field
stagesTermsFilterPre-seed, Seed, Angel, Series A, Series B, Series C, Early stage, Late stage, Growth, Venture, Debt, Grant, Accelerator, Incubator, Crowdfunding, Strategic, Acquisition, Merger, IPO, Post-IPO, ICO
announced_dateDateFilterlast_updated here is record ingest time, not a data refresh
companiesEntityFilterThe company that raised
investorsEntityFilterAny-of only — "rounds where both A and B participated" is not expressible
company_country, investor_countryTermsFilterDistinct: a German company can raise from a US investor
company_sizeRangeFilterEmployee count of the company that raised
keywordsKeywordQueryMatches the company's own profile text; this vertical has no round-level free text

job_opening#

FilterShapeNotes
seniorityTermsFilterExactly seven, most senior first: C-Suite, VP, Director / Head, Manager, IC, Entry, In Training. Director and Head are ONE value. Renamed in the 2026-08 taxonomy revamp — a value from an older integration (Senior, Head, C Suite, Intern, President) matches nothing
departmentTermsFilterOne of 23 Function labels: General Management, Sales, Marketing, Revenue Operations, Customer Success & Support, Product, Design / UX, Engineering / Software, Data, Analytics & AI, IT & Infrastructure, Security, Finance, Human Resources / People, Legal, Risk & Compliance, Operations, Supply Chain & Logistics, Medical & Health, Education, Consulting / Professional Services, Research & Development / Science, Real Estate & Facilities, Skilled Trades & Field Services, Unclassified. Only a minority of postings carry one, and an unclassified posting is absent from a filtered result rather than misfiled
sub_departmentTermsFilter225 values, populated only where department is
employment_typeTermsFilterfull_time, part_time, contract, temporary, internship, apprenticeship, seasonal, freelance, volunteer, other. A posting may carry more than one
work_arrangementTermsFilterremote, hybrid, onsite. Any-of only — match_all is ignored, since a posting cannot require every arrangement at once. onsite means neither of the other two is set
locationTermsFilterThe role's own country/state/city, not the company HQ
companyEntityFilterThe hiring company, resolved by name/domain/LinkedIn URL
salaryRangeFilterAnnualised USD even for hourly postings. Only a fifth of postings carry pay, so this filter excludes every posting that omitted it
posted_dateDateFilter

This is the one document category where the top-level include_domains / exclude_domains parameters work, because the posting carries the hiring company's domain.

Where the old seniority values went (2026-08 revamp)

seniority is a pure career ladder. It used to be a blended taxonomy, corporate rungs beside occupational classes, but the 2026-08 revamp separated those out:

  • Educator, Healthcare Professional, Utility Worker and Speciality Service Provider are Functions now and belong on department.
  • Founder, Owner, Partner, Investor and Advisor (governance) and Consultant / Freelancer (employment type) are not filterable on any axis.

academic_report#

FilterShapeNotes
publication_dateDateFilter
paper_typeTermsFilterarticle, conference-paper, report, review-article, communication, protocol, …
publisherTermsFilter
funded_byTermsFilterThe funding body named on the paper
citation_count, reference_countRangeFilterCitations received, references made

The narrowest surface of the nine, and there is no keywords field: write the subject into query instead.

market_report#

FilterShapeNotes
keywordsSingleSetKeywordQuery — not the general shapeExactly one positive set, and no negative group at all. Two sets or any negative group is a 422, though KeywordQuery accepts both elsewhere
publisherTermsFilterThe report's vendor
industryTermsFilterThe report's industry classification
published_dateDateFilterRoughly half the corpus has no publication date, so filtering on it silently drops those reports

trend#

FilterShapeNotes
keywordsKeywordQueryThe only filter this category has today

It matches the trend's own text (name, description, analysis), never its stored keyword definition. "Does this trend discuss X" and "is this trend defined by X" are different questions, and only the first is exposed.