Corpora Compared: The Case of the Swedish Gigaword & Wikipedia Corpora
This work addresses the problem of corpus selection for NLP practitioners, showing that bigger data isn't always better, but it is incremental as it builds on existing embedding evaluation methods.
The study investigated whether factors beyond corpus size, such as domain breadth and noise, affect embedding performance by comparing Swedish Gigaword and Wikipedia corpora, finding that Wikipedia embeddings generally outperformed the larger Gigaword corpus in analogy tests.
In this work, we show that the difference in performance of embeddings from differently sourced data for a given language can be due to other factors besides data size. Natural language processing (NLP) tasks usually perform better with embeddings from bigger corpora. However, broadness of covered domain and noise can play important roles. We evaluate embeddings based on two Swedish corpora: The Gigaword and Wikipedia, in analogy (intrinsic) tests and discover that the embeddings from the Wikipedia corpus generally outperform those from the Gigaword corpus, which is a bigger corpus. Downstream tests will be required to have a definite evaluation.