Project archive · 2012—2015
↗Open-source semantic text annotation
A multilingual system that found entities in free text, linked them to encyclopedic knowledge, and assigned meaningful classes at multiple levels of detail.
From document to linked knowledge
01 / Method
Entityclassifier.eu combined named-entity recognition with semantic wikification. Its output described what a phrase referred to, what kind of thing it was, and how important it was to the document.
01 · Spot
Lexico-syntactic patterns marked both named and common entities in free text, producing stable character spans for downstream processing.
02 · Link
Wikipedia Search supplied candidate meanings. The most-frequent-sense strategy selected an article and mapped its URL to the corresponding DBpedia resource.
03 · Classify
Targeted Hypernym Discovery mined explanatory sentences in an entity’s Wikipedia article. Patterns such as “X is a Y” supplied a readable class.
04 · Enrich
Entities and hypernyms were cross-linked with DBpedia and enriched with DBpedia and YAGO2S ontology types. The Linked Hypernyms Dataset made these assignments reusable.
05 · Prioritize
Later versions assigned entity salience so consumers could distinguish the document’s subject from incidental mentions.
06 · Publish
Results could be published through the RDF-based NLP Interchange Format, preserving document context, character offsets, entity links, and extracted types.
@prefix nif: <http://.../nif-core#> . <#char=0,12> nif:anchorOf "Marie Curie" ; itsrdf:taIdentRef dbpedia:Marie_Curie .
02 / Languages
03 / Source
The archived Java code separated classification, service delivery, interface, and knowledge loading.
Pattern spotting, Wikipedia linking, hypernym discovery, and ontology-backed classification.
↗RESTA Java REST API exposing the processing pipeline to integrations.
↗APPThe original browser application for submitting and inspecting text.
↗DATAImport tooling for Linked Hypernyms, DBpedia, and YAGO2S resources.
↗04 / Research record
The project connected information extraction, entity linking, hypernym discovery, linked data, and multilingual knowledge resources.
Archive note
Entityclassifier.eu is preserved as an open academic project by Milan Dojchinovski and Tomáš Kliegr. The GPL-3.0 source, datasets, and publications remain available.