What a Nerd! Beating Students and Vector Cosine in the ESL and TOEFL Datasets
This work addresses the challenge of improving word similarity tasks in natural language processing, particularly for educational and language testing applications, though it is incremental as it builds on existing measures like Average Precision.
The paper tackles the problem of unsupervised word similarity measurement by proposing APSyn, a method that outperforms Vector Cosine and co-occurrence on ESL and TOEFL datasets, achieving accuracies of 0.73 and 0.70 respectively, which beats human non-English college applicants and other state-of-the-art approaches.
In this paper, we claim that Vector Cosine, which is generally considered one of the most efficient unsupervised measures for identifying word similarity in Vector Space Models, can be outperformed by a completely unsupervised measure that evaluates the extent of the intersection among the most associated contexts of two target words, weighting such intersection according to the rank of the shared contexts in the dependency ranked lists. This claim comes from the hypothesis that similar words do not simply occur in similar contexts, but they share a larger portion of their most relevant contexts compared to other related words. To prove it, we describe and evaluate APSyn, a variant of Average Precision that, independently of the adopted parameters, outperforms the Vector Cosine and the co-occurrence on the ESL and TOEFL test sets. In the best setting, APSyn reaches 0.73 accuracy on the ESL dataset and 0.70 accuracy in the TOEFL dataset, beating therefore the non-English US college applicants (whose average, as reported in the literature, is 64.50%) and several state-of-the-art approaches.