- Status: Proposed
- Type: Specific
- Work Package: WP3
- Research Coordinators: INT Lexicographers; ELEXIS project members
- Coordinators for CLARIAH: Jesse de Does, Vincent Vandeghinste
- Participating Institutes: INT
- End-users: INT Lexicographers; Lut Colman, Carole Tiberius; ELEXIS community
- Developers: (Who is involved in implementing this use-case (if any)? Try to mention name, institute, role/responsibility)
- Interest Groups: (a list of CLARIAH interest groups, such as Text and DevOps, for which this use case may be relevant. See the list of IG's at: https://github.com/clariah/ig/.
- Task IDs: Wp3 search engine extensions: treebanks
Using dependency treebanks to study verbal-headed word combinations (cf. INT word combinations project)
The scala of current corpus environments lacks a middle ground, where the extremes are sketch engine/blacklab/CQP on the one hand (fast and easy search in large amounts of material, flexible possibilities for grouping and result display) and Gretel/PaQu (extensive search possibilities but not yet fast with large corpora) on the other hand.
We propose extensions to blacklab/blacklab-server/autosearch
- to enable fast retrieval by dependency relations
- extraction of relevant statistics by grouping on the relevant slots in the queries
- upload of treebank data created by researchers into autosearch
- exploitation of existing parallel corpora
e.g.
Query something like v:[lemma="praten"] → (n:[pos="NOUN"] → p:[pos="ADP"])
Result presentation:
Grouping:
- Existing UD treebanks
- UD-enriched version of OpenSONAR and other corpora
- extended version of blacklab/autosearch
- UD pipeline for Dutch (to be determined by accuracy/usability)
- Is the researched satisfied?
References to related resources and publications and especially links to related use-cases: