Domains and context: First steps towards managing diversity in knowledge

Abstract

Despite the progress made, one of the main barriers towards the use of semantics is the lack of background knowledge. Dealing with this problem has turned out to be a very difficult task because on the one hand the background knowledge should be very large and virtually unbound and, on the other hand, it should be context sensitive and able to capture the diversity of the world, for instance in terms of language and knowledge. Our proposed solution consists in addressing the problem in three steps: (1) create an extensible diversity-aware knowledge base providing a continuously growing quantity of properly organized knowledge; (2) given the problem, build at run-time the proper context within which perform the reasoning; (3) solve the problem. Our work is based on two key ideas. The first is that of using domains, i.e. a general semantic-aware methodology and technique for structuring the background knowledge. The second is that of building the context of reasoning by a suitable combination of domains. Our goal in this paper is to introduce the overall approach, show how it can be applied to an important use case, i.e. the matching of classifications, and describe our first steps towards the construction of a large scale diversity-aware knowledge base.

Introduction

Semantics is core in many knowledge management applications, such as natural language data and metadata understanding [20], [22], [23], [24], natural language driven image generation [54], abstract reasoning [55], [56], converting classifications into formal ontologies [7], [27], [28], automatic classification [25], [26], ontology matching [17], [18], [19] and semantic search [29]. However, despite the progress made, one of the main barriers towards the success of these applications is the lack of background knowledge. In fact, as underlined by several studies (see for instance [8], [9], [10], [11], [51]) without high quality and contextually relevant background knowledge it is impossible to achieve accurate enough results.

Dealing with this problem has turned out to be a very difficult task. In fact, on the one hand, in order to provide all the possible meanings of the words and how they are related to each other, the background knowledge should be very large and virtually unbound. On the other hand, the background knowledge should be context sensitive and able to capture the diversity of the world. The world is extremely diverse and diversity is visibly manifested in language, data and knowledge. The same real world object can be referred to with many different words in different communities and in different languages. For instance, it is widely known that in some Nordic circumpolar groups of people the notion of snow is denoted with hundreds of different words in the local language carrying very fine grained distinctions [1]. This phenomenon is often a function of the role and importance of the real world object in the life of a community. Conversely, the same word may denote different notions in different domains; for instance, bug as insect in entomology and bug as a failure or defect in a computer program in computer science. Space, time, individual goals, needs, competences, beliefs, culture, opinions and personal experience also play an important role in characterizing the meaning of a word. Diversity is an unavoidable and intrinsic property of the world and as such it cannot be avoided. At the same time, diversity is a local maximum since it aims at minimizing the effort and maximizing the gain [35].

Our approach is to take into account this diversity and exploit it to make explicit the local semantics, i.e. the meaning of words in a certain context, such that information becomes unambiguous to humans as well as to machines. Towards this goal a preliminary step is the creation of a diversity-aware knowledge base. This requires appropriate methodologies for its representation, construction and maintenance. With this purpose, we propose and adapt the faceted approach, a well-established methodology used in library science for the organization of knowledge in libraries [21]. In this paper, we describe the fundamental notions of domain and its components, called facets, which allow capturing diversity and, at the same time, allow for an incremental growth of the knowledge base.

The rest of the paper is organized as follows. In Section 2, we explain the main steps of our approach by taking semantic matching as an example. Semantic matching has been chosen because of its intrinsic importance, witnessed by the large amount of research and publications in this area and also because it was the main motivation which originally led us to the problem of managing diversity. In Section 3, we provide the definitions of domain and facet, we present the corresponding data model and describe their fundamental properties. In Section 4, we provide our definition of context and explain how to build and use it at run-time by selecting from the background knowledge the language and knowledge of the domains which are relevant to the problem. In Section 5, we describe and provide a brief description of the diversity-aware knowledge base that we have been developing. Section 6 focuses on the related work in terms of the notion of context (Section 6.1), methodologies for the construction and maintenance of domain knowledge (Section 6.2), existing knowledge bases and approaches followed for their construction (Section 6.3). Section 7 concludes the paper by summarizing the work done, listing the open problems and outlying the future work.