Typically, a concordancer searches bilingual texts (on the web) for words from a list (a glossary) and returns references listing sentences containing the word and translations of these sentences.
We extend this type of (string-based) search to multiword expressions and phrases where the search is robust with respect to morphology (singular, plural, inflected verb forms etc.), ontological hierarchies and word order (i.e. it can search for phrase templates with constraints about the fillers of the arguments).
The extended concordancer integrates a string based concordancer in a pipeline approach as a first step and filters the results returned using linguistic analysis of the input and comparison to the search template constraints.
We cooperate with SketchEngine (cf. Kilgarriff), which among other things provides a concordancer, and integrate the corresponding results as sketched above. However also results of other concordancers may be used (AntCoc for instance), provided corresponding agreements about use conditions are made.
Depending on the purpose the sketched procedure may be cascaded in order to optimize the tradeoff between analysis cost and validity of the results.