ISR11 Scientific Report No. ISR-11 Information Storage and Retrieval Information Analysis and Dictionary Construction chapter G. Salton M. E. Lesk Harvard University Gerard Salton Use, reproduction, or publication, in whole or in part, is permitted for any purpose of the United States Government. IV-25 included concepts. For example, the first phrase shown in Fig. 6 carries the concept number 422, and the mnemonic indicator M[OCRerr][OCRerr]GSWI to indicate that this phrase deals in one way or another with magnetic switches. Fig. 6 also shows that the first component of the phrase must consist either of concepts 185 or 624, while the second phrase component must represent concept 225. The indicators after the dollar sign in the output of Fig. 6 carry the syntactic information. In particular, the information given for the phrase [OCRerr] indicates that this particular phrase must be either of syntactic types 7, or 15, or 16. More specifically, there exist four `nail classes of syntactic specifi- ca[OCRerr]ions, corresponding respectively to noun phrases, subject-verb relations, verb-object relations, and subject-object relations. The four syntactic classes are in turn subdivided into apprpxi'nately twenty syntactic types, each of which specifies a particular syntactic relation between the components. The particular relations which apply to a sample phrase, labelled. SY[OCRerr]AX, are shown in Fig. 7. It `nay be seen in the figure, that the first component of the phrase must correspond either to concepts 11 or 158, whereas the second component corresponds to concepts 102, 188, or 170. Also specifiei in Fig. 7 are the four allowable format types namely 1, 3, 4 and 13. These formats are specified in the center of Fig. 7 in the form of syntactic dependency trees. Dependency trees are characterized by the fact that vertical dis- placement along a given path of the tree denotes syntactic dependence, the dependent structures being always listed below the corresponding governing structures. This can be illustrated by using the example of Fig. 7, where the format type 1 specifies that the second component,