fastText (sub)word Vectors

Computational implementations of semantic knowledge represent the meaning of words as numerical vectors, derived from their usage in (natural) language. This methodology, known as distributional semantics, has seen substantial advancements, such as the extension reviewed in this article: fastText. By incorporating sub-word information into the representation of meaning, fastText overcomes the traditional limitations of early distributional semantic models and, at the same time, it opens the way for broader applications across various domains. The present article discusses available resources, domains of application, functionalities, and fastText's potential pitfalls.