Imprecise meanings as a cause of uncertainty in medical knowledge-based systems

and Mount Sinai School of Medicine, Department of Cardiology, One Gustave L. Levy Place, New York, N.Y. 10029 There has been a considerable amount of work on uncertainty in knowledge-based systems. This work has generally been concerned with uncertainty arising from the strength of inferences and the weighting of evidence. In this paper, we discuss another type of uncertainty: that which is due to imprecision in the underlying .pnmitives used to represent the knowledge of the system. In particular, a given word may denote many similar but not identical entitles. Such words are said to be lexically imprecise. . · Lexical imprecision has caused widespread· problems in many areas. Unless this plienomenon is recognized and appropriately handled, it can degrade the performance of knowledge-based systems. IIi particular, it can leaa to difficulties with the user inter:face, and with the inferencing processes of these systems. Some techniques are suggested for coping with this phenomenon. INTRODUCTION: LEXICAL IMPRECISION Specialized fields of knowledge can be viewed as having their own languages. These languages� which are subsets of natural language, are known as "sublanguages" f61. Typically, sublanguages have a specialized vocabulary: for example, the vocabulary of medicine as found in a medical dictionary, or the vocabulary of law as found in a legal dictionary. Since sublanguages are generally used in complex situations where there are difficult problems to be solved, and critical aecisions to be made, it is important that the underlying vocabularies be well defined. In particular, the words need to have precise meanings. Oearly, if a given word does not have a precise meaning, then a sentence containing that word may be imprecise as well. If the sentence was intended to convey information relevant to the solution of a problem, then the imprecision might well lead to an incorrect solution. Unfortunately, the vocabularies underlying most sub languages are much less well defined than is commonly believed. For example, a word may be used to denote many similar, but not identical phenomena: In a recent review of the medical literature [4] we found more than a dozen definitions