Automation at your Fingertips

Presenters: Felix Ostrowski, Uwe Dierolf

The KIT catalog is based on Ex Libris' resource discovery system Primo. It includes titles not only of the KIT library, but also of other scopes, the university of applied sciences and Baden-Württemberg Cooperative State University and last but not least the records of the KIT repository KITopen. This catalog provides the search interface for KIT records and offers features such as faceting and drill down for convenient search. For years, however, the important function "autocompletion" has been missing. Today, due to the use of search engines, users are accustomed to receive meaningful word suggestions when formulating their search. This makes it possible to avoid typing errors but also 0 hit queries.

Search engines such as Google base their suggestions on previous queries, which works well given their immense global usage. Systems such as Primo typically experience much less usage, making it impossible for the system to derive sensible suggestions from user input. The autocompletion feature of Primo is thus based on collecting user input across many instances. At least in case of the KIT catalog, the results are mediocre. Far too many suggestions lead to 0 hit queries, and many words present in the index are not suggested at all.

This presentation describes an alternative approach that has been implemented for the KIT catalog. XML-Metadata is pre-processed and converted to JSON, which is indexed in Elasticsearch. A jQuery plugin enhances the existing search input by firing queries which combine multiple Elasticsearch features to (a) autocomplete the current word,
(b) autosuggest the next word and (c) autocorrect spelling mistakes.
This results in much more precise search engine like word suggestions.

It can be added to Primo directly, but also to the search slot on the homepage. Suggestions respect the scope (KIT, KITopen etc.) and also active facets, and in any case the suggested words lead to hits.