Entity Recognizer
Deep learning based extraction of named entities from text documents.
The entity recognizer is a deep learning-based solution that takes a text document as input and returns a list of instances of pre-defined entities (Person, Location, Organization, Miscellaneous).
It uses bidirectional LSTM networks to generate informative word representations that capture the contextual dependencies between words in a sentence. Additionally, a CRF layer is added on top for a higher tagging accuracy. The models have been built using Flair, a PyTorch-based NLP framework.
This tool includes a multilingual NER model supporting English, German and Dutch.
The provided container is packaged in the following asset deployed in the AI4EU Experiments Platform: https://aiexp.ai4europe.eu/#/marketSolutions?solutionId=e3794e16-0225-4bf1-a99c-b99638a22232&revisionId=f7447500-0c8d-4ca7-be7e-24ce3fefd144
Additional information:
The multilingual entity recognizer has been trained over aggregate CoNLL-2002/2003 data for German, English and Dutch.
In addition, two entity recognizers for German only have been trained on the two biggest available datasets for German: CoNLL-2003 and GermEval2014.
References: Gugu, Andel, Evaluation and re-usable implementation of DL-based approaches for Entity Recognition, 2021, Master Thesis.