The increasing interest on ontologies as the backbone technology for knowledge based systems implies the refinement of ontologies development methods. In this paper we propose a new approach to develop an ontology for African Traditional Medicine. The aim of our approach is to build a deep ontology by deepening concepts descriptions to formally represent all the semantics underlying the concepts used in African traditional medicine. We use a description logics language to formalize our approach.
1. Introduction
These last decades witness the rebirth of herbal-based treatments and Traditional Medicine (TM) across the world. Several factors can explain the increasing success of TM: it is a local medicine, less restrictive and very often cheaper than modern medicine (MM). Moreover, it is a credible alternative for low-income households. This has led to its recognition by WHO1 to work with researchers of its member countries to promote the use of TM for health care.
However, while in European and Asian countries TM is well documented, in most African countries, this knowledge is transmitted orally from father to son, from mother to daughter or from master to disciple. Therefore, as time goes by, this knowledge tends to deteriorate or to get lost when a practitioner of African traditional medicine (ATM) has no child or when none of his children keeps the interest on ATM. These are the main reasons behind the development of a knowledge based system (KBS) on ATM to store, maintain and facilitate knowledge sharing on ATM. Such a KBS should be ideally based on an ontology [1] in order to provide shared conceptualizations by formal descriptions of ATM concepts and relations between them.
To our knowledge, the first and the only attempt to build an ontology for ATM was done in [2]. This first attempt has the merit of presenting some important concepts of the field, such as actors involved in ATM (healers, fetishists, soothsayers and magicians), treatment processes and general steps taken to cure a disease. However, this attempt is poor in concepts descriptions. The described ontology [3] looks more like a glossary than a deep analysis of ATM concepts. It mainly contains terms and their “primary” definitions. Furthermore, this ontology does not consider the deep semantics and specificities of ATM. Indeed, in ATM, the deep analysis of some concepts may reveal hidden aspects, appearing at a first glance as mystical considerations, but which can be captured and conceptualized in the ontology. In addition, staying at a lexical level in concepts descriptions allows only making lexical approximations between concepts of ATM and those of modern medicine.
Our work is different from [2] in that we propose a methodology for describing concepts with their various aspects and contexts of use. We provide a framework for the formal representation of the expertise of ATM practitioners and some “mystical” associated aspects. Thus, our approach allows, as much as possible, to give the possibility, on the one hand, to holders of knowledge (traditional healers) to express their knowledge on ATM, and on the other hand, to scientists (ethno-botanists, chemists, etc...) to complete this knowledge by integrating the various scientific interpretations that may be associated to each described concept. Our new approach constructs a “deep” ontology in ATM by considering different contexts of its use.
The rest of the paper is organized as follows. We briefly present ATM and motivate the need for its ontology in Section 2. In Section 3, we present Description Logics (DL), the formalism we use to formalize the ontology. Section 4 is devoted to the presentation of a formal framework to build a “deep” ontology for ATM. Finally in Section 5, we conclude and discuss how this work could be extended.
2. African Traditional Medicine and Ontology
The purpose of this section is not to fully describe ATM, but to present an overview which highlights the importance of the development of an ontology of this domain.
According to WHO, traditional medicine refers to the knowledge, skills and practices based on the theories, beliefs and experiences indigenous to different cultures, used in the maintenance of health and in the prevention, diagnosis, improvement or treatment of physical and mental illness. Traditional medicine covers a wide variety of therapies and practices which vary from country to country and region to region. In some countries, it is referred to as “alternative” or “complementary” medicine (CAM). Traditional medicine has been used for thousands of years with great contributions made by practitioners to human health, particularly as primary health care providers at the community level. TM/CAM has maintained its popularity worldwide. Since the 1990s its use has surged in many developed and developing countries. In Africa, 80% of the population depends on TM for primary health care. In many developing countries, 70% to 80% of the population has used one form or another of alternative or complementary medicine (e.g. Acupuncture2). Such expansion of TM in general and the ATM in particular, has encouraged the creation of dedicated scientific units3 conducting researches on medicinal plants, diseases and treatment processes based on these plants. Scientific publications of these research units do not reflect the actual level of the huge work done because most of the time, research results are only presented on ethnobotanical survey forms. These forms generally contain descriptive information such as specifications of the plants and their useful parts, names of the harvests, target diseases and medicine preparation methods. The increasing number of publications in this field (e.g. [4-6]) as well as the variety of practices are such that improving the integration and the sharing of knowledge and information search on ATM is a real challenge. In addition, socio-cultural differences (different experiences, different training, different cultures, different needs, different perspectives, different languages or jargons, different contexts of use, etc ...) may cause real difficulties in ATM if different stakeholders do not have a common ontological basis.
The aim of introducing an ontology for ATM is to eliminate or at least reduce the conceptual and terminogical confusion and to walk towards a common and shared understanding in order to improve communication, sharing, interoperability and reusability. In order to satisfy these goals, ontologies such as [7] and [8] have been established in the field of medicine, and in particular an ontology for ATM has been proposed in [2]. Our work falls within this framework and aims at proposing a new approach for the construction of an ontology for ATM which considers the particularities and specificities of the domain. Uschold [9] notes that “no unified methodology is suitable for all ontologies, but different approaches are required for different circumstances”. In order to set a possible matching between ATM concepts and those of the modern medicine, our approach allows translating ATM concepts in their semantics and therefore does not limit to lexical descriptions.
Before its processing by a computer, the ontology must be represented in a knowledge representation formalism. We use description logics which provide adequate inference mechanisms for hierarchical concepts descriions.
3. Description Logics
Description Logics (DL) [10] are knowledge representation formalisms used to describe concepts in a given field. A knowledge base (KB) described in DL comprises two components, the TBox and the ABox. The TBox introduces the terminology, i.e., the vocabulary of an application domain, while the ABox contains assertions about named individuals in terms of this vocabulary. The vocabulary consists of concepts, denoting sets of individuals (identified objects of the domain), and roles (binary relationships between individuals). In addition to atomic concepts and roles, all DL systems allow building complex descriptions of concepts and roles. Depending on provided operators, there are several DL languages, the “Attributive Language” (AL) being the minimal one. We summarize here the syntax and the semantics of some DL languages.
3.1. Syntax
Concepts and roles are inductively defined from a set NC of concepts names (atomic concepts), a set NR of role names (atomic roles) and a set of operators. In the folwing, unless otherwise stated, A and B are elements of NC; r and s are components of NR; C and D are concepts descriptions; R is a role description and n is a positive integer. A description of a concept X and a role R in some DL languages is formed by the syntax of Figure 1.
The minimal language AL contains the atomic concept, the universal concept, the bottom concept, atomic negaon, intersection, value restriction and limited existential quantification:
. As indicated by the syntax in Figure 1" target="_self"> Figure 1, we obtain more exessive languages if we add further constructors to AL. The added constructors determine the names of the obined languages:
• $r.C: Full existential quantification, is indicated by the letter e
• C⊔D: The union of concepts, is indicated by the letter U
• C: Full negation is indicated by the letter C
• ≤ nr, nr: unqualified number restriction is indicated by the letter N
• ≤ nr. C, nr. C: qualified number restriction, is indicated by the letter Q
• r–: Inverse of role, is indicated by the letter I Extending AL by any subset of the above constructors yields a particular AL-language. We name each AL-language by a string of the form AL[U][e][N][C][Q][I].
Hence, ALCQI is the language obtained from AL by adding full negation, qualified number restriction and Inverse of role.
Figure 1. Syntax of concepts and roles descriptions in some DL.
3.2. Semantics
In order to define a formal semantics of concepts descriptions, we consider an interpretation I that consists of a non-empty set DI (the domain of the interpretation) and an interpretation function .I, which assigns to every atomic concept A a set AI 
DI and to every atomic role r a binary relation rI 
DI x DI . The interpretation function is extended to concept descriptions by the following inductive definitions in Table 1:
The signification of the sets in the “semantics” column in Table 1 is intuitive. For example, we have the following interpretations:
• ⊤ is the whole domain, i.e. all the individuals in the domain.
• "R · C is the set of individuals who are related through R only to individuals satisfying C.
• $R.⊤ is the set of individuals related through R with other individuals of the domain.
• £ nR · C is the set of individuals who are related through R to at most n individuals satisfying C.
4. Towards a Deep Ontology for African Traditional Medicine
4.1. A New Approach
ATM practitioners (healers, soothsayers, fetishists ...etc.)
Table 1. Semantics of some concepts and roles descriptions in DL.
make a number of assumptions that seems inconsistent with some principles of modern medicine. Terms used by those practitioners could sometimes refer to hidden aspects and therefore should not always be considered at the first degree. The semantics of the domain cannot be captured if these hidden aspects are not taken into consideration in the ontology. For example, consider the following two statements4:
1) “Drink a calabash of the potion every morning.”
2) “This plant should be collected by a young man, early in the morning, before the daybreak. Once the plant is collected, the collector must run straightly at home without stopping on the way, and the plant must be used immediately.”
The first statement introduces the term “Calabash”, a widely used container in ATM for therapeutic indications and conservation of healing potions. A “naive” or a purely lexical description of this concept in the ontology would restrict it to a container made from a fruit. However, a deeper observation of this indication shows that the concept “Calabash” in this context can refer to at least three different aspects: a fruit-based container, a dosage and/or a restriction (the medicine must be stocked exclusively in a calabash otherwise it would not be effective). This third aspect apparently superficial has a scientific basis because in many cases, the fruit forming the Calabash helps in a better preservation of the potion than a metal container where there can be a risk of corrosion or some chemical reactions between the liquid (potion) and the container.
The second statement describes the attitude to adopt when collecting some medicinal plants. It has been shown that plants requiring such an attitude contain essential oils for which the concentration is very high in the early morning and their volatility requires a very short delay between the collection and the usage. That is why some traditional doctors, without knowing this scientific explanation, recommend that such plants be collected by young men who can run fast.
These two situations show that in ATM, the ontologycal description of a concept has to go beyond the shallow conceptualization by structuring the ontology development process into two major steps: the first one consisting of primary descriptions of concepts and, the second one consisting of an incremental and iterative association of, if any, hidden aspects of the described concepts (see Figure 2).
• The first step builds a first ontology called the “primary ontology” where concepts are defined in a “naive” way; concepts are defined at the first degree ignoring any eventual hidden aspects.
• The second step “deepens” concept descriptions in order to obtain the final ontology called the “deep ontology”. This second step involves various specialists (ethnobotanists, chemists …etc.) and aims at iteratively and incrementally introducing the different hidden aspects (also called facets).
Figure 2 shows the process of the description of a concept in the proposed approach. First, the concept is primarily defined, and then, iteratively, the different stakeholders (specialists) involved in the ontology development process explore and add associated hidden aspects (facets). The following section aims at formalizing this process.