`t_res.utils.REL.mulrel_ranker` module

class t_res.utils.REL.mulrel_ranker.PreRank(config, embeddings=None)

PreRank class is used for preranking entities for a given mention by multiplying entity vectors with word vectors.

Note

Credit:

Reference:

@inproceedings{vanHulst:2020:REL,
author =    {van Hulst, Johannes M. and Hasibi, Faegheh and Dercksen, Koen and Balog, Krisztian and de Vries, Arjen P.},
title =     {REL: An Entity Linker Standing on the Shoulders of Giants},
booktitle = {Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval},
series =    {SIGIR '20},
year =      {2020},
publisher = {ACM}
}

forward(token_ids, token_offsets, entity_ids, embeddings)

Multiplies local context words with entity vectors for a given mention.

Returns: entity scores.

training: bool

class t_res.utils.REL.mulrel_ranker.MulRelRanker(config, device)

The MulRelRanker class implements a neural network model for entity disambiguation.

Note

Credit:

This class and its methods are taken (minimally adapted when necessary) from the REL: Radboud Entity Linker Github repository: Copyright (c) 2020 Johannes Michael van Hulst. See the permission notice. This is based on the mulrel-nel approach developed by Le and Titov (2018), whose original code is available in the mulrel-nel: Multi-relational Named Entity Linking Github repository, and on Ganea and Hofmann (2017).

References:

@inproceedings{vanHulst:2020:REL,
author =    {van Hulst, Johannes M. and Hasibi, Faegheh and Dercksen, Koen and Balog, Krisztian and de Vries, Arjen P.},
title =     {REL: An Entity Linker Standing on the Shoulders of Giants},
booktitle = {Proceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval},
series =    {SIGIR '20},
year =      {2020},
publisher = {ACM}
}

@inproceedings{ganea2017deep,
title={Deep Joint Entity Disambiguation with Local Neural Attention},
author={Ganea, Octavian-Eugen and Hofmann, Thomas},
booktitle={Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing},
pages={2619--2629},
year={2017}
}

@inproceedings{le2018improving,
title={Improving Entity Linking by Modeling Latent Relations between Mentions},
author={Le, Phong and Titov, Ivan},
booktitle={Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
pages={1595--1604},
year={2018}
}

forward(token_ids, tok_mask, entity_ids, entity_mask, p_e_m, embeddings, gold=None)

Responsible for the forward pass of the entity disambiguation model and produces a ranking of candidates for a given set of mentions:

ctx_layer refers to function f. See Figure 3 in Le and Titov (2018).
ent_scores refers to function q.
score_combine refers to function g.

Returns:: Ranking of entities per mention.

loss(scores, true_pos, lamb=1e-07)

Computes given ranking loss (Equation 7) and adds a regularization term.

Returns: loss of given batch.

regularize(max_norm=1)

Regularizes model parameters.

Returns:: None

training: bool

t_res.utils.REL.mulrel_ranker module

`t_res.utils.REL.mulrel_ranker` module