CogniVal: A Framework for Cognitive Word Embedding Evaluation
This provides a more comprehensive evaluation method for NLP researchers, though it is incremental as it extends existing cognitive evaluation approaches.
The authors tackled the problem of evaluating word embeddings by comparing them to human brain data, introducing a multi-modal framework that tests six embedding types against 15 cognitive datasets and finds strong correlations across modalities and NLP tasks.
An interesting method of evaluating word representations is by how much they reflect the semantic representations in the human brain. However, most, if not all, previous works only focus on small datasets and a single modality. In this paper, we present the first multi-modal framework for evaluating English word representations based on cognitive lexical semantics. Six types of word embeddings are evaluated by fitting them to 15 datasets of eye-tracking, EEG and fMRI signals recorded during language processing. To achieve a global score over all evaluation hypotheses, we apply statistical significance testing accounting for the multiple comparisons problem. This framework is easily extensible and available to include other intrinsic and extrinsic evaluation methods. We find strong correlations in the results between cognitive datasets, across recording modalities and to their performance on extrinsic NLP tasks.