Index Structures for the Exploration of Natural Language Corpora
Centrum für Informations- und Sprachverarbeitung Ludwig-Maximilians-Universität München
This study describes the development of a large-scale corpus query system – a specialized search engine used to perform advanced types of pattern search, especially for patterns used by linguists interested in discovering syntactic phenomena in large corpora.
Beginning with a review of traditional search engine algorithms, the main focus then shifts to suffix arrays, a data structure that has been available since 1987, but is not commonly used in large-scale search engines for various technical reasons.
Recently developed algorithms are considered in this study as the starting point for a new attempt to re-introduce the suffix array as a data structure of practical value to corpus-linguistic research. One of the key findings is a technique that combines several suffix arrays using indexed bit vectors and enables the searching of layers of meta information, such as part-of-speech information and semantic labels, in parallel to searching the text. A set of algorithms operating on that data structure is presented, enabling sophisticated pattern matching, such as gap-matching and gap-filling, as well as improved methods of concordance generation. The final chapters present practical examples of how the new system is used to make linguistically relevant discoveries in real corpora.
ISBN 9783862884087. Linguistic Resources for Natural Language Processing 06. 140pp. 2013.