|Particular Context||Extension to medical tele-diagnosis|
Telesystematic is a method of remote collaboration by which experts pool their knowledge (interpretation of observation, choice of best characteristics and illustrations) with the objective of improving the robustness of the multi-expert knowledge base, and its speed of construction. The scenarios discussed here are based on the experience gained during two demonstrations effectuated at the MILIA'97 Exposition (The International Multimedia Market at Cannes, France) and at ATM Development'98 (international congress on new technologies of telecommunication at Rennes, France).
In addition to the application of telesystematic to the creation of knowledge bases in the area of natural sciences, remote collaboration demonstrates the applicability of these tools and the methodological approach to the management of complex knowledge with IKBS. This approach is adaptable to analogous needs in industry and small and medium size businesses (tele-instruction and tele- diagnosis).
This methodology may be situated in the general context of management of dispersed information as a decision making business tool for building corporate memories.
Certain experts have acquired a know-how that qualifies them to judge complex situations. The expert is not simply a living encyclopedia that has memorised vast amounts of information; but rather an individual with reasoning powers that can be applied to decision making (identifying the reason for a breakdown, diagnosing sickness, identifying a species etc.) in a certain domain. Often this knowledge is partial and/or dispersed among experts who are individually competent in different related subfields. Difficulty thus arises, in synthesizing their expertise into a global knowledge base.
In order to create a corporate memory for a business enterprise, computer technology serves as a priceless aid in modeling, processing and communicating knowledge. For example, in artificial intelligence the construction of a knowledge base brings together several activities : representation of knowledge, databases, machine learning, data analysis and expert systems. The role of the computer scientist/knowledge engineer is to assist the expert in the process of eliciting his knowledge.
Today, the expansion of Internet by means of Java and high-speed terrestrial/satellite transmissions makes possible new methods of work and new tools for the remote construction of knowledge bases. This is a multi-disciplinary effort including several functions :
- Production is taken over by globally dispersed experts who furnish the knowledge.
- Edition is the role of knowledge engineers who, with the assistance of a software workshop extract the elements of expertise and store them in the knowledge base. The editor is located at a site that centralises the information.
- Validation of the knowledge involves end-users, experts included. These individuals evaluate the results and the robustness of the knowledge base.
- Telecommunication is ensured by operators who, by means of a high-speed network, establish the necessary conditions for quality audio/visual communications between the participants.
The editor and the operators are responsible for the establishment of "collectware" for the storage, management and communication of the expert's knowledge. Both are confronted with the end-users problems to which they must find suitable responses: The quality of the services they furnish must ultimately be adapted to practical usage.
The target for collectware are decision makers that need to establish a shared conceptual model (descriptive model and/or functional model) for experts who have not come to define a common point of view that would enable them to resolve problems over an enlarged conceptual domain. For example, individuals may be connected by video-conference to this collectware in view of creating a common thesaurus of useful vocabulary for identification (breakdowns, symptoms, etc ) and choosing illustrations to guide the end-user towards problem resolution. The conceptual model that they construct becomes the common denominator of their experience and is crucial for the creation of an effective and robust knowledge base.
The objective of this collectware is to facilitate the sharing and communication of specialized knowledge, knowledge.to be disseminated corporately (Intranet) or externally over the Web (Internet). IKBS enables us to apply a strict methodology to the storage, management, communication and transmission over the web,. Its application domains are numerous.
In industrial applications, we are presently witnessing the development of databases (relational or object model) associated with decision making tools (data mining, case-based reasoning). As the names of these techniques indicate, the accent in placed on the efficient management of great amounts of data appearing more often than not, in tabular form (Excel, etc.). The nature of the above knowledge concerns human constructs and activities (manufacturing, banking, insurance, etc.) and is established from a conceptual model and based on deduction.
Inversely, in life science applications, experts must apply an inductive experimental method based on observation, construction of hypotheses, and testing. Therefore, the establishment of a knowledge base for a natural domain is more complex than for an industrial domain: the accent will be generally placed on the nature of the data to represent, which are more often complex (structured, imprecise and polymorphous).
In addition, the knowledge of biology experts evolves through time. It can be called into question from one day to the next. We can therefore not consider the computerized reproduction of expert knowledge as a linear process beginning with the acquisition of knowledge, continuing through its processing and experimentation, ending in validation. Nature has furnished us with such a varied and contradictory field that it is extremely difficult to describe it through eternally valid laws.
Guided by this realization we have developped IKBS; a tool whose originality lies in dealing with this evolution by making it possible to iterate on the definition of new models and to update standing cases.
Knowledge evolves with experience, that is to say by confrontation of a conceptual model of the domain (mental representation) with the observed reality (the cases).
Concretely, this has led to the construction of a descriptive model that makes it possible to depict the observable knowledge of the domain at a given time, and with the aid of a questionnaire that instanciates this model, to enter new observed cases. These examples can all be compared to one another, making it possible to derive new general knowledge by induction (classifications, decision trees) or to produce results by comparison (identification). The produced knowledge puts the descriptive model to the test since it makes it possible to unearth certain errors in the descriptions or incoherence in the original descriptive model.
To summarize : IKBS makes it possible to make knowledge explicit at one moment and update it later on.
In order to obtain robust identifications, we attach the greatest importance to the entry of quality descriptions (the examples or the cases) before the processing, to the definition of observable attributes and the ease by which they can be made to be sucessfully interpreted by non-expert users. We have taken into account the importance of the transmission of knowledge to less knowledgeable individuals through the use of pedagogical interfaces for reaching expertise.
It is for this reason that we have found it indispensable to construct a multi-expert knowledge base that incorporates the dimensions of space (geographical distribution of experts) and time (evolution of expert knowledge) in the process of its creation. This original procedure enables us to construct a collective memory that records and communicates all/or part of the knowledge without recourse to shared time and/or place, and which guarantees a normalization of knowledge (a shared thesaurus) within the chosen domain. For example, the experts can harmonize their vocabulary and choose the best pictures and drawings to construct an illustrated web dictionary acessible to non-experts through the use of a questionnaire.
Systematics is the scientific discipline that studies and describes the diversity of living beings, elucidates the nature and causes of their resemblances and differences, brings to light the family relationships existing among them and contructs a system of classification on the basis of these relationships [Matile et al., 1987]. It is an indispensible preliminary to the study of biodiversity, for not only does it enable us to precisely determine the name of the species that we are studying, but it also enables us to assess substancial amounts of relevant data for environmental protection.
Telesystematics is a concept of remote collaborative research (multi-expert) that addresses itself to the problem of the dissipation and global dispersion of knowledge. For coral, the number of experts that describe, name and differentiate between different species can be counted on one hand. In addition, most of them are now nearing the end of their professional activities. The existing collections are scattered, as well, throughout different national museums. To avoid the disappearence of this invaluable know-how in coral systematics, the computerized tools that we propose aim at exploiting today's intuitive knowledge and existing collections of coral specimens.
Furthermore, we are establishing learning tools and expertise training for non-specialists. For this, a consensus regarding vocabulary choice (thesaurus) and illustrations (pictures, drawings) must be established between experts, in order to better disseminate their know-how and make it operational for beginners.
Remote collaboration makes possible consultation and more objective definitions of the descriptive characteristics of species. It can be effectuated synchronously (live through video-conference) or asynchronously (by emailing/exchanging documents).
The importance of remote collaboration through virtual meetings using high speed terrestrial or satellite connections can be described as follows :
- sharing the expertise of several experts,
- remote consulting of the specimens kept in various museums,
- transmit knowledge to learners.
The observation of the utilization of this method, that takes advantage of IKBS, will make it possible to define shemas of remote collaboration that should prove highly useful when applied to other sectors of experimental sciences.
|Extending to medical telediagnostic|
We propose to effectuate a feasibility study of IKBS methodology applied to different medical specialties (anatomic pathology, dermatology, neuro-surgery etc.) The result of this study will be the constrution of a knowledge base prototype for the domain which seems best-adapted to our method of knowledge management utilizing high-speed networks. It will be necessary to find a group of specialists in a field of medical science who are confronted by the same problems of interpretation in classification, description, and identification of pathologies that confront experts in coral systematics.
[ Home Page | Corals of the Mascarene | IKBS page | Presentation | Top ]
© IREMIA - IKBS team - 11/2000