Neural Cross-Lingual Entity Linking

Authors: Avirup Sil, Gourab Kundu, Radu Florian, Wael Hamza

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The proposed system has strong empirical evidence yielding state-of-the-art results in English as well as cross-lingual: Spanish and Chinese TAC 2015 datasets.
Researcher Affiliation Industry Avirup Sil, Gourab Kundu, Radu Florian, Wael Hamza IBM Research AI 1101 Kitchawan Road Yorktown Heights, NY 10598 {avi, gkundu, raduf, whamza}@us.ibm.com
Pseudocode No The paper does not include any pseudocode or clearly labeled algorithm blocks.
Open Source Code No The paper does not provide any statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes We evaluate our proposed method on the benchmark datasets for English: Co NLL 2003 and TAC 2010 and Cross Lingual: TAC 2015 Trilingual Entity Linking dataset.
Dataset Splits Yes We use standard train, validation and test splits if the datasets come with it, else we use the Co NLL validation data as dev.
Hardware Specification No The paper does not provide specific details about the hardware used to run the experiments, such as GPU or CPU models.
Software Dependencies No The paper mentions various models and systems (e.g., word2vec, LSTMs, LIEL) but does not provide specific version numbers for any software dependencies.
Experiment Setup Yes We run CNNs on the sentences and the Wikipedia embeddings with filter size of 300 and width 2. The non-linearity used is tanh. For both forward (left) and backward (right) LSTMs, we use mean pooling... For the NTNs, we use sigmoid as the non-linearity and an output size of 10 and use L2 regularization with a value of 0.01. ... For the main model, we again use sigmoid non-linearity and an output size of 1000 with a dropout rate of 0.4. ... For the MPBL node, the number of dimensions is 100.