http://cleopatra.ijs.si/midas/
- Diego Alves (ESR14)
- Gabriel Maia (ESR3)
- Gaurish Thakkar (ESR15)
- Tin Kuculo (ESR1)
This is one of the demonstrators created during the Cleopatra project. This demonstrator was developed by receiving funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement no. 812997. The goal of the demonstrator is to parse data from Wikipedia corpora in multiple languages, extract named entities through hyperlinks, and align them with entity classes from DBpedia. This will generate a named entity corpus with DBpedia class annotations which we’ll then translate to a new complex tagging hierarchy, UNER. The final objective is to train models on recognizing and tagging named entities with a multi-tiered tagging hierarchy. This will grant us a Named Entity dataset creation workflow which works for languages covered by both Wikipedia and DBpedia, including under-resourced languages. With these datasets, we can then create trained models in those same languages, leading us to a multilingual Named Entity Recognition and Classification (NERC) platform.
Multilingual information retrieval is conducted in the top 10 most spoken languages in the EU:
- EN (English)
- HR (Croatian) More to be developed and deployed in the future. Note: You can generate datasets and train your own models by using the code available in this repository.
Currently, the demonstrator only supports the languages listed above. It has the following UI functionalities:
- Raw text NERC annotation using the UNER hierarchy;
- Selection of Coarse-grained vs Fine-grained annotations;
- Webpage annotation functionality based on the URL;
- Selection of multiple languages (in progress);
- Text file input and NERC-annotated text file output functionalities (in progress).
@article {article, year = {2020},
pages = {72-79},
keywords = {named-entity,
universality, low-resourced languages},
title = {UNER: Universal Named-Entity Recognition Framework},
keyword = { named-entity, universality, low-resourced languages},
publisher = {CEUR Workshop Proceedings},
publisherplace = {Heraklion, Gr \ v {c} ka}
}
This repository makes available:
- The Documentation: Documents, files and publications that serve to explain design decisions and work done towards both the UNER tagging hierarchy and the demonstrator.
- The code for dataset creation: Code that allows you to download dumps from Wikipedia, cross hyperlinks with DBpedia classes and generate a dataset for a given language.
- The UI code: Code to run an Angular user interface, as seen in the demonstraror.
- The server code: Code and trained models that serve the tagging functionality seen in the demonstrator.
- Testing code: Code to create test cases for the demonstrator.