Federated Content Search for Lexical Resources (LexFCS): Specification

A LexFCS field type, a LexCQL query or an XML fragment.

An XML element with the Generic Identifier Element that is bound to an XML namespace denoted by the prefix prefix.

An XML attribute with the name attr.

The literal string as the content of an element or value of an attribute.

Implicit use of the default lemma field.

Explicit use of the lemma field.

Combined search on different fields.

Virtual field lang of the lexical entry and relation modifier lang for values of field translation.

Virtual field any for searching on all fields within a lexical entry.

The search term is tokenised by the endpoint and it is evaluated whether all tokens are present in the tokenised field value while ignoring the given token order. For instance, a query for Wege matches "Alle Wege führen nach Rom", as the token "Wege" occurs within it. Likewise, a multi-word query such as "Rom führen" matches the same phrase because both tokens are present. Endpoints MAY take measures to enhance recall, e.g., lemmatisation, case folding, or diacritic normalisation. For example, if an endpoint implements lemmatisation, a query for "mouse green" MAY match inflected forms like "green mice".

The regexp relational modifier and the default special characters are evaluated only with respect to individual tokens.

Endpoints MAY fall back to an untokenised matching if tokenisation is not supported.

This relation is used for exact matching against the complete untokenised value of a field. A query for lemma == Wege MUST NOT match the string "Alle Wege führen nach Rom", but MUST only match records where the entire untokenised field value is identical to "Wege".

By default, endpoints MUST NOT perform any recall enhancement, e.g., lemmatisation, case folding, or diacritic normalisation, to ensure exact matching. However, relation modifiers (ignoreCase, ignoreAccents) MAY be used to override this behaviour.

The regexp relation modifier will match against any substring of the untokenised field value. Regular expression anchoring (^, $) can be used to change this behaviour.

This relation evaluates exact inequality. It matches all records where the untokenised field value is not identical to the query string. For example, lemma <> "car" matches any record whose untokenised value is not exactly "car". It is the only inequality relation supported in LexCQL.

This relation is used to search for results based on entities or concepts that are defined in external vocabularies using Uniform Resource Identifiers (URIs). It thus resolves term values to entities rather than literal strings.

Most grammar fields use Universal Dependencies (UD) vocabularies as default. In these cases, the corresponding namespace can be omitted. As an example, the query case is Acc uses the default namespace https://universaldependencies.org/u/feat/Case and is equivalent to the query case is "https://universaldependencies.org/u/feat/Case#Acc". All field types with a default vocabulary are given in Table 7, “Field Types and Universal Dependencies Vocabularies”.

Endpoints MUST be able to support full URIs or resolve namespace prefixes to their canonical form for matching. Endpoints MAY support other namespace prefixes.

Entity references are considered to be using the default namespace if they do NOT contain a colon (:) character. Otherwise, they are either full URIs or using namespace prefixes that MAY need resolving for matching.

Matches lexical records where the lemma contains the token "car". Endpoints MAY determine their own balance between precision and recall (e.g., applying case-insensitivity or unicode normalisation).

Matches lexical records where the definition contains the token "person" (e.g., matching a definition like "describes a person who…​").

Matches lexical records whose lemma exactly matches the string "car", including canonically equivalent characters.

Matches lexical records whose lemma is not exactly the string "car".

Matches lexical records whose part-of-speech corresponds to a noun within the Universal Dependencies part-of-speech tagset.

Matches lexical records whose part-of-speech corresponds to a noun within the Universal Dependencies part-of-speech tagset. Search term is shortend by using the default namespace.

Matches lexical records whose part-of-speech corresponds to an intransitive verb according to Wikidata using the (non-default) namespace "http://www.wikidata.org/entity/" with the namespace prefix "wd".

By default, special characters within search terms are interpreted as wildcards.

Specifies the language of the search term. A match SHOULD only occur with lexical field values that contain an explict language specification matching this language, or if the default language of <lex:Entry> matches. It is RECOMMENDED to use and support an IETF BCP 47 compliant language code.

Instructs the endpoint to either ignore or respect string casing during matching, overriding the endpoint’s default casing behavior.

Instructs the endpoint to either ignore or respect diacritics and character variants, overriding the endpoint’s default behavior (where respectAccents is RECOMMENDED as the default). This modifier controls the handling of Unicode combining characters, historical letter forms (e.g., ſ U+017F), ligatures (e.g., fi U+FB01), or multi-character equivalents (e.g., ß U+00DF). When ignoreAccents is active, endpoints SHOULD apply Unicode compatibility normalisation (NFKD) and strip combining marks (e.g., fi → fi), fold historical letter forms to their modern equivalents (e.g., ſ → s), and apply case-folding rules for characters such as ß → ss. When respectAccents is active, endpoints MUST preserve the full distinction of characters as encoded, including diacritics, combining marks, and compatibility-equivalent or historical letter forms. No accent stripping, compatibility folding, or character-level normalisation (beyond standard Unicode normalisation for comparison) are applied for matching purposes.

Interprets the search term as a POSIX regular expression. Support for extended regular expression features is OPTIONAL and endpoint-specific. By default, regular expressions are unanchored, meaning they match against any substring of the index value unless the expression explicitly uses anchoring constructs such as ^ (start of string) or $ (end of string). This overrides the interpretation of the default special characters mentioned above.

Matches records where the lemma contains the token "car" and any token starting with "s" (e.g., "safety car", "car service").

Matches records where the lemma contains the literal tokens "car" and "s*".

Given search term in French.

Given search term to be evaluated regardless of case (incl. "Handy", "handy", "haNDY").

Given search term to be evaluated as "café", "cafe" or other accent-normalised variants.

Given search term is a regular expression describing any value beginning with "cat" using an anchored regular expression.

Given search term in English regardless of case.

Represents the intersection of two result sets. A record MUST match both connected search clauses to be included in the final results.

Represents the union of two result sets. A record MUST be included in the final results if it matches either of the connected search clauses.

Combination of two search clauses with the operator "AND".

Using parentheses to change evaluation order of a complex query.

Combination of search clauses with two different relations.