Company keywords

The company_keywords filter searches a company's extracted text for the terms you specify. It's a text filter (relation match_expression), built for themes and signals ("sustainability", "ISO 9001", "veteran-owned") rather than for products.

What it searches

company_keywords runs against the text Veridion extracts for each company:

Business tags — machine-learning-extracted keywords that capture the company's core activities.
Company description — a machine-generated summary of the company's activity, synthesized from its website.
Meta information — title tags, meta descriptions, and language metadata from the homepage HTML.

Which of these are searched depends on strictness (below).

Accepted relation

match_expression only. The value is an expression object with a mandatory match block and an optional exclude block. Each block has an operator (and or or) and operands:

and: every operand must match.
or: any one operand may match.
Operands are either search terms or nested operator / operands blocks, so you can build complex expressions.

A search term is a single word or an exact phrase. An exact phrase matches only as written: "sustainable business" matches "sustainable business practices" but not "sustainable practices for your business."

{
  "attribute": "company_keywords",
  "relation": "match_expression",
  "value": {
    "match": { "operator": "or", "operands": ["sustainability", "net zero", "carbon neutral"] },
    "exclude": { "operator": "or", "operands": ["consulting", "agency"] }
  },
  "strictness": 3
}

Nested expressions

Nest blocks for "this AND (that OR the other)". For example, veteran-, women-, or minority-owned IT or software companies, as an and of two or groups:

{
  "attribute": "company_keywords",
  "relation": "match_expression",
  "value": {
    "match": {
      "operator": "and",
      "operands": [
        { "operator": "or", "operands": ["veteran-owned", "women-owned", "minority-owned"] },
        { "operator": "or", "operands": ["IT services", "software"] }
      ]
    }
  }
}

Strictness: which text is searched

strictness widens or narrows the text the filter looks at. The default is 3.

1: business tags only. Highest precision, fewest results.
2: business tags and meta information. A balance of volume and accuracy.
3: business tags, meta information, and the company description. The most results, though accuracy can taper off toward the end of a long result list.

Tips

Group variations with or (plurals, synonyms, alternate spellings) to raise both volume and accuracy; companies matching more terms rank higher.
Use and to narrow broad terms and cut false positives.
Use exclude to drop unwanted senses of a term.

Evidence

On a match, evidence returns matched_snippets — the text that matched (one snippet per or, one per operand of an and). See Explainability (evidence).

Where it's accepted

A text gate in supplier_discovery and segmentation, and a claim check in company_verification.