Since the beginning of 2011, SLUB Dresden and the start-up company Avantgarde Labs have been developing a multilingual semantics-based catalogue search called SLUBsemantics.
SLUBsemantics - Multilingual Semantic Search
In the first phase of its development, SLUBsemantics will allow users to access our four million locally-held titles with the help of automatically generated categorical branches. Search queries will be followed back to their conceptual origins and sorted out based upon contextual knowledge. Search queries, regardless of the language that they are posed in, will return relevant results from our catalog, regardless of the language in which the texts are written or described.
SLUBsemantics' seemingly magical way of working is the result of one simple principle: drawing upon enormous, socially maintained, network-like information structures with the goal of automatic enrichment and connectivity among catalogue entries. For this purpose, we rely, among other things, on Wikipedia, the world’s largest encyclopedia. The software recognizes connections within the content of the search query automatically and produces a structured results list. For example, if the user types in the word “Bank” (in German, this word means both bank and bench), he will see relevant catalogue entries about individual banking institutions, the banking crisis and the Euro, as well as information about seating or sandbanks. If he looks for “capital city Belgium”, all of the hits containing Brussels will be displayed. If the entry reads “Gone with the wind”, he’ll get all of the hits for the film “Gone with the wind”, hits regarding the director Victor Fleming, for Max Steiner, the composer of the film’s music and for Margaret Mitchell, the author of the novel.
The normal index search of the SLUB Catalogue with spell-checking, word standardization and adjustable sorting criteria is a progression from the familiar single search field, but it can only produce results from the things that are filtered out from title descriptions. Titles that are written with descriptions in other languages, or whose descriptions use synonyms of the search phrases, are often overlooked. Other semantic relationships between descriptions are missing completely. By contrast, even in its first phase of development, SLUBsemantics covers all of the local catalogue entries and allows the entry of search terms in three languages: German, English, and Polish. Semantic relationships between topics are shown as branching structures, and the standard settings for graphic display can be switched to a tree-like representation of the search results. SLUBsemantics is currently in the beta phase. In the interest of increased transparency of the search results we will display all of the automatically produced search enrichment terms for all of our catalogue entries in the detailed view of the hits, and these terms will be linked to the corresponding Wikipedia articles. In the long run, we are planning to expand the realm of SLUBsemantics to all of the entries in the SLUB catalogues and to integrate SLUBsemantics into the standard search. Additionally, the search enrichment terms will be corrected and expanded.
Please help us to continue to improve SLUBsemantics and to further its development. For your suggestions or comments, please use our feedback form on the SLUBsemantics website.
Avantgarde Labs, Dresden