in with

KorAP: Annis QL

ANNIS Query Language (Annis QL or AQL) is a query language of the ANNIS corpus search system designed particularly to deal with complex linguistic corpora with multiple annotation layers (e.g. morphology) and various annotation types (e.g. attribute-value pairs, relations). The concept of AQL is similar to searching node elements and edges between them, where a node element can be a token or an attribute-value pair.

KorAP supports the following keywords by using the default foundries:

node
a node element
tok
a token
cat or c
a constituent
lemma or l
a lemma annotated node
pos or p
a part-of speech annotated node
m
a morphologically annotated node

KorAP does not support in-query metadata constraints in AQL yet, namely the prefix "meta::". In KorAP, metadata constraints should be separated from search queries and be given as corpus queries defining virtual corpora.

Node elements

Simple tokens

"liebe"

Attribute-value pairs

tok="liebe"

Namespaces in AQL are realized as foundry and layer combinations in KorAP. They can be used to query tokens having a specific layer annotated by a specific parser (foundry), for example coordinating conjunctions (part-of-speech layer) from the TreeTagger foundry.

tt/p="KON"

Regular expressions

tok =/m.*keit/

Sequence queries

Two consecutive tokens

"der"."Bär"

Finite verbs indirectly followed by an adverb, where any number of tokens may occur in between.

pos="VVFIN" .* pos="ADV"

Negation

Negation, such as negated tokens, is only supported in KorAP in a sequence query.

"Katze" . pos != "VVFIN"

Pointing relations

Pointing relations describe direct relationships between two node elements, for instance dependency relations.

Querying all "SUBJ" dependency relations

node ->malt/d[func="SUBJ"] node

Querying "SUBJ" dependency relations where the source node is token "ich"

"ich" ->malt/d[func="SUBJ"] node

Querying "SUBJ" dependency relations where the source node is token "ich" and the target node is a perfect participle

"ich" ->malt/d[func="SUBJ"] pos="VVPP"

Using references

Node elements may be refered to by using # and the position number of the element. For instance,

"ich" & pos="VVPP" & #1 ->malt/d[func="SUBJ"] #2
"ich" & pos="VVPP" & #1 . #2

Unary operators like arity or tokenarity are not yet implemented in KorAP.