Techniques for automatic indexing of textual documents have been used for almost two decades by many commercially available systems (e.g., STAIRS, INQUIRE, MEAD DATA CENTRAL, KWIC). However, these systems are built upon the uniformity of the English language and the frequency of occurrence of specific word forms and word stems. In particular, they rely on the fact that the content of a document is directly related to the frequency of occurrence of key words in the text. Automatic key word extraction, therefore, has become a science unto itself with many algorithms being used to sort out the relevant key words from those that don’t really describe the indexed document.

In the paper, Jonák states that automatic indexing is unsatisfactory for full texts because current methods cannot guarantee required efficiency. Instead of full text indexing, Jonák postulates the use of semantic equivalents to the units of text. A semantic equivalent is a set of elementary units of meaning, called semes, which can represent the units of text. Semes are more than just word stems. For instance, a bank, an agency, a library, and a government building are all instances of the seme “AO” (institution).

Jonák’s experiments with semes included comparing the work of expert indexers against the coding of semes for textual documents. The aim was to provide better results than could be done with automatic indexing. However, the tables used to suport the finding are not consistent with the stated results of the experiment. Although the text states that the semes approach did a good job of matching the results of expert indexers, the tables clearly show little similarity between the results of the two approaches. This could be because the document was originally written in Czechoslovakian, and it lost something in its translation.

In short, the idea is interesting, but the results do not support the conclusions. I would not recommend the paper to someone unless he or she were interested in researching the original in its native language.