Repository for the paper: Analyzing Structures in the Semantic Vector Space: A Framework for Decomposing Word Embeddings

Link to the paper: Analyzing Structures in the Semantic Vector Space: A Framework for Decomposing Word Embeddings

Please use the following citation:

@inproceedings{hanselowski2019vedec,
          title={Analyzing Structures in the Semantic Vector Space: A Framework for Decomposing Word Embeddings},
          author={Hanselowski, Andreas and Gurevych, Iryna},
          booktitle={arxive/...},
          year={2019}
        }

Disclaimer:

This repository contains experimental software and is published for the sole purpose of giving additional background details on the respective publication.

Download the word embeddings

Download pretrained GloVe Vectors

    wget http://nlp.stanford.edu/data/wordvecs/glove.6B.zip
    mkdir -p data/glove
    unzip glove.6B.zip -d data/glove
    gzip data/glove/*.txt

Download pretrained Word2Vec Vectors

    wget https://s3-us-west-1.amazonaws.com/fasttext-vectors/wiki.en.zip
    mkdir -p data/word2vec

Run experiments

In order to reproduce the experiments from the paper run the following jupyter notebooks:

• Semantic Tree (Paper Section 3.1)

semantic_tree_model.ipynb

• Semantic Space Networks (Paper Section 3.2)

semantic_space_networks.ipynb

• Category completion (Paper Section 4.1)

ssn_category_completion.ipynb
svm_category_completion.ipynb

• Word analogy (Paper Section 4.2) (not yet implemented)

evaluate_on_all_mod.ipynb (not yet implemented)

Contacts:

• hanselowski .a.t. gmail.com
• https://www.informatik.tu-darmstadt.de/ukp/ukp_home/
• https://www.tu-darmstadt.de

License:

• Apache License Version 2.0