The Examples of Human Commonsense Reasoning Processes

The Examples of Human Commonsense Reasoning Processes

Xenia Naidenova (Military Medical Academy, Russia)
DOI: 10.4018/978-1-60566-810-9.ch005
OnDemand PDF Download:


In this chapter, we concentrate our attention on analyzing and modeling natural human reasoning in solving different tasks: pattern recognition in scientific investigations, logical games, and investigation of crimes.
Chapter Preview

An Example Of Reasoning Process In Pattern Recognition With The Use Of Knowledge Base

The investigation of human reasoning process in different real situations is the inevitable step foregoing any work on creating automated algorithms modeling this process. For studying, we have taken the process of visual deciphering forest photographic images. In this case, the features of forest regions are investigated under stereoscope by a decipherer (operator or executer), registered on blanks, and analyzed by the use of the decision rules created in advance by the specialists on the basis of previous explorations and experiences.

The attempt of automation of the process of deciphering forest images leads, first of all, to investigating the algorithms used by experts for this goal. We have analyzed, the dichotomous diagram (scheme) elaborated by an expert for Lena - Angara forest region under deciphering the types of forest plant conditions. The scheme is a decision tree the nodes of which are associated with some factors or attributes’ values to be checked. The sequence of checking the factors associated with the nodes of scheme is rigidly determined. It is optimum with respect to the length of branches coming to each of the nodes.

Our studies show that a strict collection of attributes and a strict sequence of their use in all situations do not reflect adequately the processes of specialists’ reasoning during deciphering forest photographs. Decision trees help greatly to increase the productivity of the work of decipherers, but they decrease the number of correctly recognized objects. The strict order of the use of attributes is easier for inexperienced decipherers and more difficult psychologically for experienced ones.

We have compared the two forms of experts’ knowledge representations: 1) the dichotomous scheme familiar to a decision tree, 2) the table of rules reflecting the links between the factors (attributes) and the types of forest plant conditions with the indication of the factor occurrence frequency.

Two decipherers worked: one having experience of work more than 10 years (Executor 1) and the second, which did not have an experience of forest deciphering, but he knew how to carry out the preliminary processing photographs and how to work with the simplest stereo - instruments (Executor 2).

Two regions have been chosen: the basin of rivers Lena - Angara (Region 1) and Khentey - Chikoy region (Zabaykalie and Mongolia) (Region 2). The results of deciphering are represented in Table 1 and Table 2. The data of the ground-based assessment have been taken as true. The trustworthiness of recognition was estimated for 214 parts of Region 1 and for 192 parts of Region 2.

Table 1.
The results of deciphering the types of woodland
      Executer      Region 1      Region 2
      By rules      By decision tree      By rules      By decision tree
      % correct results      % correct results      % correct results      % correct results
      1      88      79      66      69
      2      43      62      49      64

Complete Chapter List

Search this Book: