TrueCalling glossary

What is Semantic Talent Discovery? Semantic candidate search, explained

Semantic Talent Discovery is an AI candidate-search method, used by TrueCalling, that interprets the contextual meaning behind a job brief (skills, seniority, sector, hiring context) rather than matching exact keywords or job titles, surfacing relevant passive candidates that Boolean search routinely misses. TrueCalling applies it to 1.2B+ professional profiles (TrueCalling internal data, 2026).

Definition reviewed on

How does semantic candidate search work in TrueCalling?

  1. How is a plain-English brief turned into search criteria?

    Semantic Talent Discovery starts by reading the job description the way a recruiter would. In TrueCalling, a language model extracts structured criteria from the text: title keywords, must-have skills ranked by importance, nice-to-have skills, city and country, minimum years of experience, industry and job function. A description written in French, German or any other language is read as written; the criteria are translated into English for the profile search and shown to the recruiter in their own language. The model also checks that the city and the country belong together and corrects a mismatched pair instead of running a broken search. The recruiter can edit every criterion. The limit is the source text: if a description never names a skill, no model can infer it reliably, which is why a clear job brief still matters.

  2. How are equivalent titles and related skills found?

    Semantic Talent Discovery has to match meaning, not spelling, and TrueCalling does it in two ways. For job titles, a language model generates equivalent titles for the role, and the search covers all of them without the recruiter typing each variant. For skills, each skill is turned into a 256-dimension vector, an embedding (TrueCalling internal data, 2026), and skills whose vectors sit close together are treated as neighbours, so related skills match even when the wording differs. If that lookup fails, the search falls back to plain keyword matching. A Boolean string needs every synonym spelled out with OR, which is where experienced sourcers still miss variants; our guide to boolean search for recruiters covers where Boolean remains the better tool. The limit: wider matching lowers precision, so must-have skills carry the search and deserve care.

  3. How many profiles does one search read, and how are they ranked?

    Semantic Talent Discovery reads widely and ranks, rather than returning only exact matches. In TrueCalling, one search on a job description analyses up to 2,000 profiles from a pool of 1.2B+ and keeps the 500 best-scored results (TrueCalling internal data, 2026). Each profile receives a candidate fit scoring result from 0 to 100 with matched and missing skills, so the list is ordered by fit to the description, not by keyword hits. The search runs on the server and continues if the recruiter closes the page. When too few profiles come back, it can be widened by dropping lower-priority skills one at a time. The limit: a profile database only knows what people have published, so niche or offline profiles may be missing, and the ranking still needs a recruiter's review. For the full method, see the AI sourcing guide.

How does semantic candidate search differ from Boolean search?

Legacy approachSemantic Talent Discovery (TrueCalling)
Recruiter writes a Boolean string per search.Criteria extracted from the job description by a language model.
Each title variant typed with OR.Equivalent job titles generated and searched automatically.
Exact keyword match only.Related skills matched through embeddings, with keyword fallback.
Results unranked or sorted by keyword hits.Up to 2,000 profiles read, ranked by a 0–100 fit score.

FAQ: Semantic Talent Discovery

Do I still need Boolean strings with semantic search?

Not to start a search. Semantic Talent Discovery builds its criteria from the job description, including equivalent job titles. Boolean remains useful for precise exclusions and on platforms that only accept Boolean syntax, so teams can use both: semantic search to find the pool, Boolean to narrow it.

Can semantic search find candidates who use a different job title?

Yes, within limits. TrueCalling generates equivalent job titles for the role and searches all of them, so a search for one title also covers its common variants. It cannot find a person whose profile never mentions the relevant work: titles and skills still have to appear somewhere.

Which languages can a job description be written in?

Any language the language model reads well. TrueCalling translates the extracted criteria into English for the profile search and shows them in the recruiter's interface language. City names are kept in their local spelling, such as München or Milano, to match how profiles record them.

What happens after semantic search finds the profiles?

They are scored. In TrueCalling, the search reads up to 2,000 profiles for a job description, then TrueFit 360 gives each one a score from 0 to 100 with matched and missing skills. The recruiter reviews the ranked list, builds the shortlist and approves any outreach.

See Semantic Talent Discovery in action

30-minute demo of TrueCalling against one of your open roles: sourcing, scoring, and outreach end-to-end.