Empirical Study of Diachronic Word Embeddings for Scarce Data
This work addresses data scarcity in diachronic word embeddings for NLP researchers, but it is incremental as it builds on existing methods.
The paper tackled the problem of learning diachronic word embeddings from scarce temporal data, comparing three models and showing that regularization improves performance, with specific gains in drift detection accuracy.
Word meaning change can be inferred from drifts of time-varying word embeddings. However, temporal data may be too sparse to build robust word embeddings and to discriminate significant drifts from noise. In this paper, we compare three models to learn diachronic word embeddings on scarce data: incremental updating of a Skip-Gram from Kim et al. (2014), dynamic filtering from Bamler and Mandt (2017), and dynamic Bernoulli embeddings from Rudolph and Blei (2018). In particular, we study the performance of different initialisation schemes and emphasise what characteristics of each model are more suitable to data scarcity, relying on the distribution of detected drifts. Finally, we regularise the loss of these models to better adapt to scarce data.