Mining Text, Linking Entities – NLB’s Journey

Abstract

Connecting collections across institutions and discovering hidden knowledge has always been our goal. Due to disparate data sources, differing levels of quality and granularity in the data sets and varying terminologies used for concepts and entities, this has proved to be a challenge. Barring standardisation across knowledge repositories, which is an insurmountable task, the next best thing will be to explore how text mining techniques together with linked data technology may be of help to support semantic searches.

Working on text mining techniques to identify works and entities, the National Library Board of Singapore (NLB) has discovered a way to link its resources to other institutions’ knowledge bases. This approach was implemented across the collections of NLB, the National Archives of Singapore (NAS), and the National Heritage Board (NHB) which manages the museums in Singapore. Conducting a search on the respective portals will enable users of these institutions to retrieve contextualised results from all three institutions.

An entity widget was created and deployed on the various portals. An example would be at NHB’s “Roots” page where visitors looking for previously undiscovered heritage connections will be able to supplement their research through the widget. This will link them to NLB and NAS resources defined by events, people, places or organisations identified in NHB’s content. Based on linked data’s resource description framework (RDF) and an entity recognition software, connections between entities are extracted, disambiguated and established. NLB proposes that such a widget can be deployed on the websites of interested institutions.

Using the same technical infrastructure, we highlighted the works of prominent Singapore Chinese pioneers in collaboration with the Singapore Federation of Chinese Clan Associations. The collated results form a knowledge base of these illustrious pioneers who can be traced back to China and their connection to overseas Chinese.

This paper will also address the issues faced with text mining and entity linking, the importance of named entities creation and how data mining can enhance knowledge discovery. NLB’s efforts can help other libraries to connect with similar institutions in their community.