DUUI: spaCy
Splits the document into sentences and tokens and adds lemmas, part-of-speech tags, morphology, dependencies and basic named entities. These annotations provide the foundation for subsequent components.
Models & components
This workshop combines several DUUI components. Each one adds a different layer of information to the same text. The overview below explains what each analysis contributes and links to its implementation and models.
DUUI: Coreference
Connects mentions that refer to the same entity—for example, a person’s name and a later pronoun. Coreferee supplies the German rule-based resolution used for the interactive arrows.
DUUI: Transformers NER
Compares two multilingual transformer-based approaches. They recognize people, organizations, places, events, products and taxonomic names. Viewing the models together makes agreements and differences immediately visible.
GLiNERFlexible label-driven entity recognition.
GLiNER2Multilingual structured information extraction.
DUUI: Time Detection
Detects dates, times and durations and normalizes them as TimeX3 annotations. The workshop compares TEI2GO with Microsoft Recognizers-Text.
DUUI: HeidelTimeExt
Adds rule-based, domain-sensitive temporal annotation with the Text Technology Lab’s extended German HeidelTime resources.
DUUI: GeoNames FST
Resolves location mentions from every NER view against the European GeoNames gazetteer. Although it runs once per NER result internally, the workshop presents it as one component.