drgriffis / knn-embedding

k-Nearest Neighbor second-order word embeddings

Implementation of k-nearest neighborhood methods of generating second-order word embeddings, as described in:

• D Newman-Griffis and E Fosler-Lussier, "Second-Order Word Embeddings from Nearest Neighbor Topological Features." arXiv, arXiv:1705.08488. 2017.

This library contains two components:

1. Nearest neighbor calculation

• Scripts: nn_saver.py, nearest_neighbors.py
• Implemented in Tensorflow
• Uses cosine similarity to identify nearest neighbors

2. Graph generation

• Script: generate_graph.py
• Generates a weighted, directed edgelist file compatible with node2vec

Dependencies

A few custom libraries are included as frozen copies in the dependencies folder:

• drgriffis.common.log -- Logging utilities, from here
• pyemblib -- Library for reading/writing word embedding files (Github link)
• configlogger -- Library for writing runtime configuration to logfiles (Github link)

knn-embedding's Issues

Some questions about input files

Could you please share the two files “data/sample_embeddings.bin” and “data/sample_embeddings.neighbors”？Thank you！！！