Evaluation of Named Entity Recognition for the Danish Language

Project course objectives

A student who has met the objectives of the course will be able to:

• Define own learning objectives for the evaluation of NER systems project.
• Collect scientific knowledge and data related to the project topic.
• Carry out a well-founded delimitation of the project and formulate specific hypotheses and aims.
• Plan and carry out the course of the project in collaboration wit the project supervisors.
• Assess and summarize the project results in relation to aims, methods and available data.
• Carry out the project and interpret results by use of Python or other programming language.
• Structure and write a final short technical report including problem formulation, description of methods, experiments, evaluation and conclusion.
• Presentation of methods and results at meetings with project supervisors.

Short project description

Evaluate performance (NER classification performance, computational complexity: space and time) on the dataset (1) using models (2).

(1) Dataset

• Name: UD-DDT (DaNE)
• Task: NER
• Words: 100,733
• Sents: 5,512
• Annotated with Named Entities for PER, ORG and LOC

Link to dataset.

(2) Models

• BERT NER model
• flair_ner_model

Link to models.

Relevant links

Alexandra Institute - danlp

BERT - original paper

Transformers - original paper

BERT word embedding tutorial

Classification evaluation measurement

Confusion matrix

IOB

How to use Flair

Flair paper

Bidirectional LSTM-CRF Models for Sequence Tagging (paper)

Finetuning of the BERT model