This repository contains the code for the participation on GermEval 2019 Task 1 -- Shared task on hierarchical classification of blurbs.
Our team achieved the 9th place in Sub-Task A and 7th place in Sub-Task B, many improvements could have been made, and the code here can be the starting point for more experiments.
You can see on the images below the results for both tasks achieved by all participating teams.
The paper describing our system:
You can read the papers of both best winners here:
- Multi-Label Multi-Class Hierarchical Classification using Convolutional Seq2Seq
- TwistBytes - Hierarchical Classification at GermEval 2019: walking the fine line (of recall and precision) Fernando Benites
pip install -r requirements.txt
wget https://s3.eu-central-1.amazonaws.com/alan-nlp/resources/embeddings-v0.4/de-wiki-fasttext-300d-1M.vectors.npy
wget https://s3.eu-central-1.amazonaws.com/alan-nlp/resources/embeddings-v0.4/de-wiki-fasttext-300d-1M
python -m nltk.downloader punkt
python -m nltk.downloader stopwords
subtask_a.py
subtask_b_local_classifier.py
subtask_b_global_classifier.py
score.sh