bheinzerling / selpref-emb

Selectional Preference Embeddings (EMNLP 2017)

selpref-emb

Selectional Preference Embeddings (EMNLP 2017)

This repository contains joint embeddings of selectional preferences, words, and fine-grained entity types.

Download

• lemmatized, basicDependencies

• lemmatized, enhancedDependencies

• lemmatized, enhancedPlusPlusDependencies

• unlemmatized, basicDependencies

• unlemmatized, enhancedDependencies

• unlemmatized, enhancedPlusPlusDependencies

Usage

The files are in gensim model format, which can be loaded in Python like this:

from gensim.models import KeyedVectors

emb_file = "/path/to/embedding_file"
emb = KeyedVectors.load_word2vec_format(emb_file, binary=True)

The vocabulary consists of:

• verbs and their dependency relation separated by "@", e.g. "sink@nsubj" or "elect@dobj"
• words and short noun phrases, e.g. "Titanic"
• fine-grained entity types using the FIGER inventory, e.g.: /product/ship or /person/politician

Reference

@InProceedings{D17-1138,
  author = 	"Heinzerling, Benjamin
		and Moosavi, Nafise Sadat
		and Strube, Michael",
  title = 	"Revisiting Selectional Preferences for Coreference Resolution",
  booktitle = 	"Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing",
  year = 	"2017",
  publisher = 	"Association for Computational Linguistics",
  pages = 	"1343--1350",
  location = 	"Copenhagen, Denmark",
  url = 	"http://aclweb.org/anthology/D17-1138"
}

selpref-emb's Issues

Cannot reproduce result in the paper

In the paper, it mentioned the cosine similarity between (Titanic, sank@nsubj) is 0.11 while the similarity of (iceberg, sank@nsubj) is -0.005. However, using all 6 models provided, I could not achieve the same number. What is the issue?
emb.similarity('titanic', 'sink@nsubj') emb.similarity('iceberg', 'sink@nsubj')