Journal of Information Science http://jis.sagepub.com Relational Indexing. Part I J. Farradane Journal of Information Science 1979; 1; 267 DOI: 10.1177/016555157900100504 The online version of this article can be found at: http://jis.sagepub.com/cgi/content/abstract/1/5/267 Published by: http://www.sagepublications.com On behalf of: Chartered Institute of Library and Information Professionals Additional services and information for Journal of Information Science can be found at: Email Alerts: http://jis.sagepub.com/cgi/alerts Subscriptions: http://jis.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. Relational Indexing. Part I J. Farradane University of Received 17 turn, is a reflecis almost comterms and does not express structure; ’not’ causes considerable difficulties, because it is wrongly applied to entities, whereas it should be applied to predicates. Other attempts to control meaning, such as the use of links or roles, have also been too unspecific or arbitrary to be useful. These inadequacies have to some extent been modified by the application of additional procedures, such as the truncation of search terms (so as to cover various grammatical forms, or meaningful derivatives, of words), or the more involved procedure of ’feedback’ and ’nteraction with the questioner, verbally or on-Ij’.le, so that the questioner can ’guess again’ with an altered question, or possibly, by supplying relevance judgments on an initial output, introduce changes in weighting of terms so as to bias a further search output in his favour. Several iterations are often necessary in order to improve results. The author’s system of relational indexing has been subject of several papers in the journal literature (sec bibliography at end), but no comprehensive exposition of the system has previously been made. The following description of the principles of the method, with full examples, and an outline of its computerization, now provides such a definitive statement. Many of the details discussed relate to practical problems raised by students and colleagues. the 1. Introduction essen- tially two indexing requirements: (1) a method of controlling the vocabulary to be used, and (2) a method of structuring terms from the vocabulary to express the meaning between words. Classifications offer a flxed hierarchy of terms, with only implied, and mixed, relations, or, if faceted, show either implied relations (free-facet systems) or a limited number of generalized categories of relations (e.g. Colon). None of these provide sufficient flexibility for detailed indexing of complex subjects such Attempts to use computers to analyse original by means of programs for grammatical or syntactical analysis, or for statistical word frequency and/or word co-occurrence computations, have not demonstrated much better results. Such approaches, and also the use of compound terms in thesauri, show that a need is felt for pre-coordination of terms at the indexing stage. Classifications are fully pre-coordinated systems, but are too static. One special system of pre-coordination (Hans Selye’s &dquo;Symbolic Shorthand System&dquo; for endocrinology and stress literature) uses classification for word control and a special set of symbols to indicate specific technical relations between terms or special attributes of terms, and has been reported to be highly successful; it is however not applicable to other subject fields. Gardin, in his &dquo;Syntol&dquo; system, uses a grammatically derived set of categories of relations, but the results have not been satisfactory. What is needed is a means of expressilig relations which will be of general application in any subject and at any level of complexity. Meaning, texts articles. The usual coordinate thesaurus to control the vocaindexing systems bulary, selected as unconnected terms, and Boolean connectors (in putting a question) to organize the meaning. Most thesauri offer only limited indications of classificatory, or, more particularly, other relations between terms, intended as an aid to the choice of words (descriptors) by the indexer or questioner; the indexer, in the light of his opinion concerning the subject content of a document, must still decide which of the available terms will express the subject (without expressing relations). For many reasons, Boolean logic (though suiting computer operation) is inadequate or misleading for structuring the terms in a question, and in fact expresses accurately only a small part of the relations between terms which we as appear in to tion of July 1979 An information retrieval system involves indicate in language, which, in thought. The Boolean ’and’ pletely unspecific; ’or’ only replaces try Western Ontario, London, Ont., Canada journal use a considered as relations between terms, must therefore be analysed directly, unimpeded by the subject matter or by linguistic considerations. Relational Indexing, as described here, is a means of expressing relations on a basis of the mechanisms of thought, to Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. 267 268 be converted directly into indexing notation. The involved may be very different from case to case, but the framework of possible relations is limited and invariant, and thus by-passes the subject field and linguistic problems, and can be applied at any level of detail or complexity. The principal difficulties in appreciating meaning lie in the complexity and arbitrariness of language, and often also in the imperfections shown by many authors in their use of language. Language may be regarded as a translation of thought into a surrogate for the purpose of communication (see George Steiner: &dquo;After Babel&dquo;, Oxford, 1975), but it has not evolved as a medium of great accuracy, so that attempts are always in progress to standardize terminology, at least for teclnical terms. But there has been no standardization (in common use) for parts of speech other than nouns or verbs, or of the structure of sentences or writing styles. Meaning depends very greatly on the connectives between nouns and verbs, and these connectives are the means of expressing relations. In speech and writing, however, the connectives we use vary greatly, and often arbitrarilv. The standardization of words is still very much an unsolved problem. An authority list, or the more complex thesaurus, does not include definitions (at most it has a few scope notes), and it is assumed that the reader knows the meanings. To make it into a dictionary would be too cumbersome. The prevention of the entry of synonyms, and the provision of adequate cross-references to cover them, is not always well performed. The partial classification, in the form of broader and narrower terms, and of (rather vaguely) related terms, should clarify meaning, but is often misleading. There are no established rules for citing ’related’ terms. The frequent use of compound terms, again arbitrarily, bedevils standardization. Verbs are not well distinguished from nouns. No final solution to these problems can however be offered here. Relational Indexing does nevertheless tend to indicate, in a subtle manner, the types of words needed for adequate expression of meaning, and obviates the need for casually used compound terms. subject concepts in the psychology of thinking. It has sufficient evidence, that these processes are basically much simpler than might have been supposed, and yield a system comprising a limited number (9) of relations (or, more correctly, categories of relations) between concepts. The term ’concept’ is used here to mean any iiiiit concrete ‘tlring’ or abstract ’idea’, of any level of complexity, e.g. wood, chair, furniture, design, or theory. A concept is denoted by its r13r11e, which is an association of a sound or written symbol(s) with the concept. The concept, in itself, is a remembered pattern or structure of sense data, as acquired by sigl, hearing, touch, etc., but its namc, once learned, and defined, obviates the need for deeper analysis. Furthermore, if a word (name), however complex in meaning, it clearly known, it should be used as such in indexing, and should not be substituted by two or more explanatory terms, though the ’lower level’ terms may be used individually in their own right. Language, however, involves many ambiguities, and care is always necessary to ensure that a word is used with one exact meaning in mind. In a thesaurus it is usually assumed that the reader knows the meaning of words, but, without definitions, unforeseen difficulties often arise. For Relational Indexing, only concepts which are nouns or verbs are to be used. For consistency, also, it is recommended that all verbs (actions, operations or processes) be used in the form of the present participle, ending in ’ing’ (if possible); this also avoids confusion between other meanings that occur when a verbal noun is used, e.g. it avoids confusion between ’governing’ and ’government’, when the latter might be taken to mean ’tlie government’, or some more abstract concept of government. Occasionally, however, an action is to be considered by itself, and not applied to something else (at least in the context of a given document), e.g. system of analysis (analysis /; system) or camera for photography (photography /, camera) - see later for the notation where, although ’analysis’ or ’photography’ do really investigated been found, on imply analysing or photographing something, it is the technique as such which is in these cases the main subject. The noun form of the action is then quite correctly usable, but such 2. The basis of relational indexing Since the true basis of meaning exists in our the system of relations to be described here is based upon an analysis of thought processes, as thought, treatment must be careboth the noun and gerund must of course be accessible in the thesaurus or other list.... Adjectives must not be used alone as concepts; they can, however, be added after nouns, with an intervening comma, if necessary (see examples, later), fully standardized, and form (with scope notes) Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. 269 and searched for, or not, as desired. In some cases the use of compound terms (two or more words) is however unavoidable, e.g. ’aromatic compounds’, ’photographic emulsion’, ’boiling point’. In such cases, the use of the noun first, followed by the adjective, would give unsouglt first terms, or terms implying a different meaning. Any such compound term must also be a matter of careful standardization. A term such as ’ultraviolet light’, although it is a well-known combination, is however better standardized as ‘light, ultraviolet’ since it may be useful to be able to search for the word ’light’ alone. A relation exists between two concepts when some meaning is implied between them in the mind. The psychological aspects are unfamiliar and look abstruse at first, especially since they expose some of the vaguer interconnections which thought produces in the absence of exact knowledge. If, however, the relations are studied in terms of practical situations of the expression of meaning between concepts, their use is greatly simplified and memorized, so that recourse to the theoretical basis for elucidation of a difficult subject statement is only very rarely neces- 3. The relations Table 1 shows the relations in the The analysis of the mind has ’mechanisms’ for interconnecting concepts: association and discrimination. Each mechanism develops into three fairly well-defined stages, and a child, as it grows, develops the capacity to use both mechanisms together. Association is developed quite quickly, though not at first with complete assurance; discrimination is much slower in development. These processes which develop in the child remain as the mechanisms of thought in the adult. It is not the mechanisms which vary to produce the great variety of knowledge and thought in different people, but the stocks of concepts which are acquired by the different experiences of individuals, and to some extent also their differences in intelligence, which controls the ability to use the mecha. nisms, and their different memory capacities. It is the nine combinations which result from the two sets of three-stage mechanisms which are the basis of the relations between concepts, and these nine relations (together with their possible negations) have been found in practice to be necessary and sufficient to express meaning in all subject fields. In terms of the basic relations, the mechanisms of our thought are not different in different subject fields. Where our knowledge of a subject (that is, the relation between two concepts) is vague or insufficient (even though two main thinking shows that setting of their origin. The typewriter symbols, which are slighly mnemonic. have been devised for ease in writing meaningful ‘strings’, etc. of terms, e.g. A /- B /; C. The names below the symbols are arbitrary, and are provided only for ease of naming, and so referring to, the relations. In a computerized system the relations can be recorded as the numbers 1 to 9 (and 10 to 18 for their negations, which are rarely required). The names of the columns and rows are, it is hoped, more expressive. Some brief explanation of the psychological basis of the relations may be useful at this stage, but this can soon be replaced by the meanings they express in practical Knowledge of a situations. concept (or datum) proceeds from sary. basically can talk about it), the result will appear in the need to use only the less precise (less clearly associated and/or less clearly discriminated) relations. we even of mere awareness a single sense to association of a concept first with a word and then with another concept or concepts. The earliest concepts, beyond unnamed feelings, which are acquired are those of external objects, concrete ‘things’; abstract ’idea’ concepts are harder to acquire, and depend, of course, upon the prior acquisition of language. Even for adults, however, there are many situations where one’s knowledge is only awareness of a new sensation, or concept, or one knows only that it occurs in the presence of (concurrently with) another concept; with repetition of the experience, an association is formed by memory; sometimes the repetition fails when expected, and one recognizes that the occurrence is only occasional (temporary association). The power of discrimination also starts with mere appreciation of concurrence, without sufficient experience to make distinctions. There is then the stage of recognizing that two concepts have characteristics in common (although other characteristics may be different), so that they are not distinct in conceptualization. Finally, the mind is able to interrelate concepts which are recognized as being distinct. The relations which are perceived between two concepts are produced by the combined effects of the two types of mental mechanism. The meanings of the relations have been derived partly from consideration of the implications of the two mechanisms, but also thing Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. 270 Table 1 by analysis of practical situations and their probable interpretations. Each of these relations is actually a category of relations which permits different shades of meanings as expressed in language, although the basic sense of the relation is unchanged, and this will become clear from the definitions given below. This state of affairs has been considered in further analysis, which has suggested that the different aspects of a relation appear to arise from whether each of the two words connected by a relation is a ’concrete’ or an ’abstract’ concept. If this is true, one would expect four possible variants of meaning of a relation according to whether the word pairs were concrete-concrete, concrete-abstract, abstract-concrete, or abstract-abstract combinations; linguistic usage confirms this. It has however been found quite adequate to use the nine relations (relational categories) as given, and that the use of 36 relations would be far too complicated and unnecessary, even though language shows the use of different connectives in some cases, and the same relation may need to be used with variants of meaning at different positions in a complex subject. It has been found that common speech (or writing) contains too many variations in expression, and attempts to derive relations from natural language have resulted in chaos. to some extent 3.1. The relations 1. Concurrence /0 . Concurrence or thing witten as It also expresses ’duration’, which can be introas a word, e.g. annealing /0 duration /= 2 hr, or omitted, e.g. annealing /02hr. It can also be used to express ’future action’, as will be discussed duced later. 2. F~/ra/c/?c6’ /= . This expresses complete or equivalence in some degree, up to equivalence (which is the case of synonym, not the best way of introbe used for the introduction of OR-term. though this is ducing it). It can proper names, e.g. polyelectrolyte /= Separan. It also expresses the idea of something to be considered as, or to be used as, something else, e.g. sodium /= ion, molasses /= fodder, acetone /= solvent. 3. Distinctness /) This is applicable in the expression of the relation of an imitation or substitute, e.g. man /) statue, information retrieval /) model, mathematical. It is rarely required for the expression of mere awareness of difference, but is sometimes useful to indicate comparative difference, e.g. exemplified The relations can however be natural language terms as follows: with (expressed linguistically also as encyclopaedia of chemistry, even though ’of involves other types of possible meaning). exemplified juxtaposition (mentally) of in The two recorded one occurrences of the term speed must be separately; in such a case, greater clarity might perhaps be achieved if one could add, say, /= 40 km/hr’ after the second entry. another, e.g. A in the presence of B, A /0 B. It also expresses the relation of bibliographic form, e.g. chemistry /0 encyclopaedia 4. Self actiaitv /* The meaning of this relation of concurrence Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. with 271 temporary association has been difficult theoretically, practice. but to assess meanings have been found from It suffices to express the intransitive verb situation, e.g. man/* walking, bird /* migrating. (The is applied to such an intransitive the term action, ’operation’ is used for a transitive This relation must not be used with a transiaction.) term ’process’ tive verb, even though linguistic custom appears to make it suitable, e.g. ’a child receiving a gift’ is not to be indexed as: but is ‘gift /- receiving /; child’ (gift receiving by child). Such constructions will be discussed later with more examples. It has also been found to be suitable to express the dative case of ’to’ (or the ablative ‘from’; but see relation 9, /: , for the sense of ’arising from’) where there is an indirect object of an operation, e.g. children /* French /- teaching, for ’teaching French to children’; the meaning becomes clear on reading the analysis from right to left. (Note how language may be misleading if one says ’teaching children French’.) It should be noted that /* implies ’to’ only for the dative case situation (as read, exceptionally, from right to left), and is to be used sparingly; it must not be confused with ’into’ or ’on to’, which are positional relations, e.g. states, and certain temporary properties. Space may concern just relative position, e.g. shelf /+ position above /+ table (note the difficulty of terminology for such relative position), or actual position, e.g. book /+ shelf, or place, e.g. manufacturing /+ England. Similarly, relative or actual time may be expressed, and the relation of rate or speed, e.g. engine /+ speed. These temporary states also include temperature, electric charge, crystalline form (e.g. salt /+ crystal), solution (e.g. salt /+ solution), etc. The relation also applies to variable properties of amount or size, e.g. number, weight, volume, concentration, pH (acidity), etc. (Note that unique properties of a thing take the relation /( , so that 2 lb of apples is ’apples /+ 2 lb’, but a particular two-pound apple is ’apple /( 2 lb’). The relation of quantity can also be recognized in unusual terminology, e.g. oil /+ loss. porary 6. Action /This is used for any thing or operation acting on, or affecting, another thing or action, e.g. clothes /water /- purifying, decomposing /- preventing. Note the order of the words, with the object of the action placed first. This relation always detergent, expresses present action; past and future action will z be discussed later. 7. Association /; This expresses various forms of association, which may be unspecified, e.g. prison /; disgrace, or the tool for, an operation, e.g. knife. (But see page 272.) etching /; acid, cutting/: It may also be used for abstract properties, e.g. picture /; beauty, or for indirect or calculated (not intrinsic) properties (imposed by man’s thought), e.g. food /; purity, machine /; efficiency. It can also be used to express ’past action’, as will be discussed further below. relation of ’adding sugar to (into) tea’. Similarly, ’converting X-rays to visible light’ is to be analysed correctly as the light arising from the converted X-rays, i.e., for 8. The relation can also mean ’through’ (reading from left to right), e.g. ’flowing through a pipe’ becomes ’flowing /* pipe’. It may be noted that in the more detailed subject ’water flowing through a pipe’ two meanings of the relation may arise within one subject: water /* flowing /* pipe, and it will be seen that the two meanings have similarities. 5. Dimensional /+ This expresses position in space or time, tem- an agent of, or Appurtenance /( This expresses the whole-part relation, e.g. table /( or the organ of a body, e.g. pig /( liver, or an intrinsic ingredient, e.g. tea /( caffeine. (It may be noted that one can distinguish such a case from sugar (added) in tea (now the beverage) by writing ’sugar /+ tea&dquo;, but the two meanings of ’tea’ would have to be leg, separated in a thesaurus.) It also expresses the generic relation, e.g. genus /( species, if it were desired to express this specifically in a subject (and not find it by reference to a thesaurus or classification). Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. 272 This is also the relation for all intrinsic direct physical properties of a particular material e.g. metal /( density, tube /( diameter. 9. Functional or thing, dependence /: This expresses the relation of one thing causing or producing something, which is also describable as a product arising out of (or partly arising out of) an initial thing, e.g. wheat /: bread, computer /: output; author /: book. It thus expresses ’cause and effect’ which is a special case of functional dependence. It is especially applicable in indexing chemical reactions, e.g. sucrose /: sucrose esters /- synthesizing, coal tar /: dyestuffs /- manufacturing. It should be noted that one should not index a material thing as arising from an action, which is an abstract term; the product arises from the object of the action, e.g. raot water language as /- heating /: ’steam from steam, as might be said heating water’. in 3.?. The order of concepts The slash which precedes the ’punctuation’ symbol in each relational code indicates the ’direction’ of the relation. The relation is always to be understood (read) as proceeding from left to right, or, in twodimensional diagrams, which are needed for more complex subjects, from top to bottom (see many later examples), and the second concept is that which is in some way subordinate to, or later in time than, the first concept; the second concept thus adds additional detail (depth of indexing) to the subject than by literal interpretation of some frequently used linguistic expressions, especially in regard to the meaning of prepositions. A common type of expression is ’page in a book’, which really means a (particular) page of a book, since the page is part of the book, i.e. ’book /( page’ (not ‘page /+ book). Similarly, note the different meanings of ’of in the expressions: ’colour of a rose’ (property), ’sport of kings’ (what kings do), and ’book of the film’ (book from which the film was made). Other examples of correct meaning of ‘ot~ have already been shown, as in: chemistry /0 encyclopaedia (encyclopaedia of chemistry), and information retrieval /) model (mathematical model of information retrieval). In speech, very varied prepositions are used for ’position’, e.g. ’living on the third floor at (of) a hotel in London’. Similarly, very varied prepositions may be used in language to express most of the other relations. In some cases language dispenses with a preposition, although a relation is intended, e.g. ’computer output’ for ’output from a computer’. common When the indexer has become accustomed to the relations, as defined, the underlying basic, and correct, relation (meaning) between two concepts will more easily become apparent. The idea of ’operation /; agent’ must not be confused with the situation where one action is really part of another, even when the word ’by’ is the usual linguistic connective. E.g., ’cooking meat by frying’ does not mean the frying is the agent of cooking, but that it is the species of cooking, and should be written as ’meat /- cooking /( frying’. This is the correct analysis even when the second action applies to a different object, e.g. ’purifying water by precipitating impurities’ is: represented by the first concept. The dative case, as noted above, provides the only exception to the ’reading’ rule. It has been found useful, for consistency, to standardize actions (operations) to provide the order of the passire construction of a sentence, i.e. an action follows the thing acted upon, the agent of the action follows the action; similarly properties follow the thing possessing the properties, and so on. Of course, in an intransitive situation, the action (process) follows the originator of the action. With practice, this order soon becomes a simple, meaningful habit. 3.3. l7ie ambiguity Language, as uf prepusitions commonly used, is often misleading, to analyse meaning more basically and it is necessary since the impurities are associated with the water. The same form of analysis applies to the situation of ’methods’ of carrying out an operation, e.g. compound /- analysing /( methods. If, however, the second action is less clearly part of the first action, the relation /; is appropriate (see later examples). Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. I 273 Such distinctions may look confusing, but careful standardizations of sense aid retrieval. 3.4. Blultiple relationships - So far, for the most part, only binary relations have been exemplified. It is however possible for one concept to be related to several terms at once, either by one relation to two or more terms simultaneously (not by the same relations in different directions), or by different relations to two or more terms. For example, a substance may contain, by the relation /(, two or more ingredients at once, or , as in the case of steam from water when heated (see above), one concept may be related, by different relations, to several others. In most subjects, it will be found that one concept is not related, by different relations, to more than four other concepts, but cases have been found where seven different relations at once were possible. In complex subjects, the procedure of perceiving the relation between two initial concepts is followed by adding further terms by suitable relations, so that A / B is extended to A / B / C. until the whole subject is expressed in a complete diagram, as subsequent full examples will show. 3.5. Distinguishing meaning by relations The relations also permit sensitive distinctions of meaning which, carefully used, enable different types of records to be separated. Thus, the use of /( for intrinsic physical properties, /; for indirect or calculated properties, and /+ for variable temporary properties, enables more exact subjects to be separated by the indexing from more diffuse subjects, with greater precision in retrieval. Some properties, such as ’hardness’, which are not quantitative, but are measured by comparative ranking, are nevertheless better treated as intrinsic, e.g. steel /( hardness. Similarly, transitive actions, expressed by /-, can be separated from intransitive actions, expressed by /*, when the verb used is the same; for instance, the on the physical chemistry of the intransitive (self-acting) process: sugar /* crystallizing can be separated from the manufacturing operation: even literature /- crystallizing. Furthermore, a present, ongoing, action indicated by /- can be separated from a past action (which has virtually become an associated property) by using for the latter the relation /; , e.g. potatoes /- washing indicates an action actually occurring, and potatoes /; washing is equivalent to sugar ’washed potatoes’, but still permits retrieval on the &dquo;washing’. It also permits the subject to be enlarged by a further, present, action, as in term which ’storing washed potatoes’. The actual of the indexing ‘diagrams’ can be varied arrangement as long as the direction of the relations is clear, so that one could write: means and perhaps so allow for the addition of further relations and terms in a complex subject; it is however best to use the clearest representation of the relations that is possible for a given subject, as it makes understanding of the indexing diagrams, and their checking, simpler. Good analysis is aided by clear diagrams. In the computer input, however, the diagrams are converted to a set of binary relations. Consideration of the meaning of a llltllre action has led to the conclusion that the notation should be /6 . This has however rarely proved necessary in the indexing of actual literature. It will be noted, however, that in all these cases the gerund verbal form used still remains unchanged. The ease of indexing any complex subject will depend very much on the clarity of expression, in ordinary language, of the subject statement to be indexed. It will also depend on the accuracy of knowledge expressed in the subject. Where that accuracy is lacking, the relation tends to be in the top line of the table of relations; for example, ’association’ implies a vaguer meaning than ’appurtenance’ or ’causation’. Sometimes, also, a relation in the left-hand column may be more appropriate for a vaguer relation than a relation more to the right in the table, e.g. /=. expressing some degree of equivalence, indicates a vaguer relationship than that given by /+ or /( , so that ’sodium /= ion’ is a better representation than ’sodium /+ ion’, since an ion is not exactly a ’state’. Other examples will be apparent later. Once decided, such analyses can be standardized. In a few cases, language involves semantic ambiguities which cannot be resolved, and must virtually be disregarded. For example, a salt solution (salt in a state of being dissolved) is ’salt /+ solution’, but when it is poured (’pouring a salt solution’) the word ’solution’ becomes, by a subtle transformation, an entity Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. 274 still write ’salt /+ solution fact that a direct physical property may vary under different conditions does not make it into an indirect property (/;) or a variable (ad hoc) property (/+), so that /( is still to be used for e.g. elastomer /( resistivity, variable. It will now be clear that the exercise of analysing a complex subject for relational indexing involves a type of (psycho-logical elucidation which may at first appear artificial. Since, however, as is claimed, the relations are those which are the natural basis of thinking, if not of language, the method soon becomes compellingly clear and easy to use. Cases of apparent difficulty will be found to arise either from difficulties in determining the exact meaning of words, or possibly the intended meaning of the whole subject statement, both being sometimes obscured by the idiosyncracies of everyday language. Such cases will be discussed later. It is important not to be misled by the conventions of language, and not to fall into a habit of using a particular relation when a particular connective is used in a text; thus, although /; is often expressed in language by the preposition ’by’ or ’with’, this is not an invariant guide and may be misleading. (a solution), but we must /- pouring’. Similarly, the , , 4. Indexing techniques . 4.1. Boolea?i connections . . are a number of notational rules which the writing of the indexed representation of simplify a subject. No term should be written as related by the same relation to two other terms in different clirections; the other two terms should be brought together, i.e., one should not write There This rule must be strictly observed; otherwise analyses may be produced. It is not that faulty is wrong in meaning, but that non-observance of the rule may obscure the correct analysis of a complex subject of which it is part. The rule will not affect the capability of retrieving a given combination of terms. In such cases, moreover, the application of the Boolean connectors may meaningfully come into play as an aid to the indexer’s intention. If the terms B and C are clearly in close combination, then this can be shown by a curly bracket to imply the Boolean ’and’, e.g. an alloy containing copper and nickel is indexed as: I If the Boolean ’or’ is to be indicated, be used; steel containing nickel indexed as: can a square bracket chromium is or Both Boolean connectors may appear together; for example, steel containing iron and chromium or nickel is: It will be noted that the Boolean connectors do not a ’linear’ relation between terms as do the other relations in the table, but a simultaneous state of connection of two or more terms to a first term by one of the relations from the table. The Boolean connectors are in fact taken from ideas in formal logic which are imposed upon our thinking at a relatively late stage in learning. Their use in this indexing is however quite simple, as an aid in analysis, to clarify the exact terms which are related. These Boolean connectors will imply their meaning only in the direction of the relation next to which they are inserted; if the same meaning is intended on the other side of the two or more terms connected, the connector must be added also on the other side. In some cases, however, the Boolean connectors may apply on one side only, as later examples will show. The Boolean connectors are not needed in the computer input for information retrieval. The Boolean ’and’ is implied in all cases by the input and searching procedures. With adequate standardization of terminology, searching on alternative terms, or ’generic posting’ of higher terms as well as specific terms, should not be necessary. The application of ’logical jumps’ (see later) in searching makes the program very complex if searching on alternative terms in the question is also introduced; it is much simpler to enter alternative questions. This will be considered again in the section on computerization. imply Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. 275 If the Boolean connector is not used, the be left in a vaguer form, as above: indexing and then the terms further relation can be written as a applying to only remarked that Boole originally intended his be applied to statements, not to single In coordinate concepts.) indexing, the Boolean ’not’ has often given rise to difficulties. The negation may be applied to any relation, so that in practice (especially in the computer program) one must allow for eighteen relations, not only nine. The negative relation is different from the positive form, so that the two relations can be applied to one concept in connectors to can one of different directions, e.g., or possibly . 4.3. Afore where the dots indicate that this leaves room for further With careful diagramming, can be shown even when included on the further D is not related to B, and additions in the diagram. such individual relations a Boolean connector is side of the connected concepts. 4.::. complex diagrams The complexity of a relational diagram will of depend on the degree of detail (depth of indexing) represented, and this detail will relate to the complexity of the subject. Some simple subjects yield just a linear string, e.g. course Ring Jiilgrams There are frequent cases where the statement of a subject in ordinary language involves the repetition of word, e.g. ’analysis of a compound by measuring the spectrum of a derivative of the compound’. This repetition is clearly undesirable for retrieval, and is avoided by the creation of a ring in the diagram, e.g., a A ring is ’read’ clockwise for meaning, still taking each relation in its left-to-right or top-to-bottom direction (which is essential for the computer input). Side chains are included as they are reached, but a longer ’main-line’ chain may be reserved for reading last. Again, this is only for checking meaning during indexing; it does not affect the computer input. (See also the example in the computerization section). There are a few occasions when it is desired to show a negated relation, e.g. ’coffee not containing caffeine’. In this case, a bar is written above the relational symbol, e.g. ’coffee /( caffeine. This, it may be noted, shows the falsity of the Boolean ’not’, in coordinate indexing, in application to a concept, instead of to a relation; it is inaccurate because there is no such thing as a ’not-concept’. (It may be which is ’statistics cross-tabulations production by ’minicomputer for producing crosstabulations for statistics’. A sligltly more complex subject is: minicomputer’ or The depth of indexing may depend on the extent to which it is worth while, for retrieval, and whether the detailed additional terms are likely to be sought terms for retrieval. Fig.l 1 Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution. 276 Fig. 4 temperature. The dots are only to show the direction of relational connections; arrowed connecting lines could be used instead. The subject of Fig. 2 is about an alloy sheet of thickness 0.5 mm, which has magnetic properties and contains iron, carbon and silicon, obtained from rimming steel (an unalloyed electrical steel) which has been decarburized by hot rolling and subsequent annealing at 750-800°C for 1-~’ hours. One or two cases have been found where the two dimensions of the paper do not suffice to show the room Fig. 2. Fig. 3 illustration that very much more complex be handled in one diagram, if required, two further examples are given here. The subject of Fig. 1 is a comparison of a hammer mill, grater and rasp for comminuting beet in which the sucrose and pulp content is quantitatively analysed on extraction with ethanol and water, at As an subjects can interconnections clearly, e.g. as in Fig. 3, but this call be drawn as if it was a ring seen from above, as in Fig. 4. The subject is the study of the arrangement of atoms in a crystal by means of an X-ray diffractometer counter for counting X-rays diffracted by the atoms. The value of such an elaborate diagram is that it enables the indexer to check whether the meaning he intends has been correctly represented; in practice he might then decide that a simpler form might be adequate. The diagrams are only an intermediate which alterations can still be made with accuracy. As will be described later, the computer input requires only the statement of each triad, wordrelation-word. taken separately from the diagram. stage, at Downloaded from http://jis.sagepub.com at University of Western Ontario on November 20, 2007 © 1979 Chartered Institute of Library and Information Professionals. All rights reserved. Not for commercial use or unauthorized distribution.
© Copyright 2025 Paperzz