The advantage of using relational databases for large corpora: Speed, advanced queries, and unlimited annotation

Relational databases can be used to create large corpora that provide both very good search performance and a wide range of queries. This paper outlines how this approach has been used to create theCorpus del Español, which contains 100 million words of text in Spanish texts from the 1200s-1900s. The main databases are composed of n-grams tables (all unique 1, 2, 3, and 4 word sequences) and the associated frequency of all n-grams in each century (historical Spanish) and register (Modern Spanish). These tables are then joined to other tables containing part of speech, lemma, synonyms, and user-defined lists of words and lemma. There is essentially no limit to the amount of annotation that can be added in additional tables (with little or no impact on performance), and the SQL-based queries allow a wide range of searches that are not available with traditional corpora.