Sari Vainikainen

Utilizing Linked Open Data Sources for Automatic Generation of Semantic Metadata

In this paper we present an application that can be used to automatically generate semantic metadata for tags given as simple keywords. The application that we have implemented in Java programming language creates the semantic metadata by linking the tags to concepts in different semantic knowledge bases (CrunchBase, DBpedia, Freebase, KOKO, Opencyc, Umbel and/or WordNet). The steps that our application takes in doing so include detecting possible languages, finding spelling suggestions and finding meanings from amongst the proper nouns and common nouns separately. Currently, our application supports English, Finnish and Swedish words, but other languages could be included easily if the required lexical tools (spellcheckers, etc.) are available. The created semantic metadata can be of great use in, e.g., finding and combining similar contents, creating recommendations and targeting advertisements.

Utilizing Existing Semantic Knowledge Bases for Creating Annotations on Mobile Devices

This paper discusses how existing semantic knowledge bases can be utilized in creating annotations on mobile devices. The difficulties in this topic arise from the fact that different semantic knowledge bases offer different access methods and in general the access methods are designed for desktop applications instead of mobile applications. To examine these issues we implemented a Java ME application that can send annotated messages to two web applications, Tilkut and Owela, which had been created earlier within the same project at VTT. Our findings highlight the challenges and possibilities that implementing the communication between the semantic knowledge bases and the mobile application offers. We also discuss different ways of easing the annotation process on a mobile device such as retrieving tag suggestions continuously based on user input.

Experiences of semantic tagging with Tilkut

There are two main approaches for adding metadata to web content: free user-generated tags and keywords based on taxonomies or semantic ontologies. There are pros and cons in both approaches, so benefits may be gained by combining these approaches. This paper presents a solution that utilizes both freely defined tags and predefined ontologies. To study suggestionbased semantic tagging we created a social bookmarking application called Tilkut. This paper describes the system and experiences from a small scale user study. Suggestions for more intelligent tagging solutions will be given.