LCP's query language DQD (Descriptive Query Definition) follows the idea of Entity-Relationshop models. Entities sets are defined by logical formulae on properties and relations between them. The query engine then searches for those constellations inside the corpus or corpora selected. Along the lines of first order logic, quantors can be employed to enforce the existence or non-existence of constellations.

Every query needs to specify at least one result set, which is either a (plain) list of entities comprising the query matches, a statistical or a collocational analysis.

In a text corpus with standard annotations (viz the CoNLL-U format), a simple query for the occurrence of the word “dogs” would look like this:

Segment s
 
Token t
    form = "dogs"
 
Result => plain
    context
        s
    entities
        t

Entities are defined in the Corpus Template. Though aleatory names can be used, we use Document for documents, Segment for sentence segments and Token for tokens in the examples. Entities are expected to start with an uppercase character.

Attributes are also defined in the Corpus Template and thus their naming is free for a corpus creator to define. As a standard set, we use form for word forms, lemma for lemmas, upos for Universal part-of-speech tags, and morph for Universal features. Attributes stored as Meta Data is mapped to entity attributes unless a native attribute with the same name exists. In that case, the meta attribute needs to be explicitely referenced.