Integrative Levels Classification project scheme monograph references

« Facets

Basic facet syntax

Free relationship allows to list several concepts co-occurring in a knowledge item:

mqvtn wni   "whales [in some relationship with] vessels".

Such way of combining concepts provides no information on the nature of the relationships between them. However, in many contexts specific relationships need to be expessed, like in "whales disturbed by vessels". This function is performed by facets.

While in ILC free relationships are expressed by a blank space, facets are expressed by digits. Digits are thus said to be the facet indicators:

mqvtn360wni     "whales, disturbed by vessels"

Facet relationships are slightly more complex to be managed both by indexers and by information retrieval systems. For many purposes free relationships can be enough, as the links between concepts are irrelevant, or obvious even if not explicitly expressed. In other situations, however, specifying relationships can be important:

An unorganised set of keywords [or corresponding notation] representing a total subject packet could still be filed under each of its significant elements in turn, and it would not strictly be necessary to organise these terms into any particular pattern to allow the machine to identify citations in which certain terms or sets of terms have co-occurred. [...]

But while it might be a relatively simple matter to set up and run such a system, it is doubtful whether its performance would be rated as entirely satisfactory beyond a certain limited range of subject fields. These are those areas of knowledge, often referred to as the hard sciences, in which authors and their readers usually tend to share common and well-established frames of reference, and use the same term or terms when referring to the same objects or phenomena. In the harder disciplines terms also usually suggest, of their own accord, the grammatical roles they are most likely to adopt in the majority of subjects being indexed. In these circumstances it would be hard to justify the extra effort of achieving a fully structured subject statement. Given that certain terms, such as "Flow", "Wings", "Boundary layers" and "Swept back", together constitute the subject of a document, we should reasonably expect a user familiar with the frames of reference which hold within the field of aeronautics to be able to relate one term correctly to another, and so deduce the whole subject correctly from its parts. It would be hard, for example, to construe "Flow" as a kind or part of "Wings", or to regard "Swept back" as an attribute of "Boundary layer".

We cannot assume, however, that these favourable circumstances hold in every indexing situation. We cannot, for example, presume the existence of a single frame of reference if we are classing or indexing across the entire subject spectrum. Nor can we dispense with structure, except at the risk of ambiguity, when we are dealing with subjects at the "soft" end of the subject spectrum, in which the meanings of terms tend to vary according to the different contexts in which they are used.

This applies particulary to studies relating to man and his social activities. [...] Concepts such as "The individual", "Expectations", "Role" and "The State", for example, can be related together in various ways to produce quite different but equally valid subjects. These are also the fields in which the exact meaning of a term often cannot be known until after it has been placed within the context of some other term or terms, and the relationship between them has been made explicit." [Aus76a]

ILC can then be used either in a simpler way with only free relationships, or in a more sophisticated version with facets. Choice depends on how much information is available about the nature of relationships in the subject, as well as on the indexing policy adopted. A mix of the two approaches can also be useful: core combinations of concepts linked by facets, in turn combined freely with others:

mqvtn81n wni5he     "whales, young [in some relationship with] vessels, with steam engine"

The syntax of facets follows the rules of predicate logic [Gno06a]. Relationships are predicates, with the involved concepts as their arguments:

disturb (whales, vessels)
p        a1      a2

Most predicates are asymmetrical, that is, their meaning changes according to the order of the arguments (which form an ordered set). Indeed, if we revert arguments a1 and a2, we get the meaning "disturb of vessels by whales", which is different from "disturb of whales by vessels". Some other predicates are symmetrical, as their arguments can be changed without altering the meaning of the relationship (they are an unordered set), like "similarity of a1 and a2 and a3 and...". Symmetric predicates are not much used in classification though.

As in most knowledge items the theme is an entity like "whales", rather than a relationship like "disturb", it is more usual to express relationships in term of arguments of the theme (nominal construction), listed one after the other:

mqvtn81n360wni     "whales, young, disturbed by vessels"

mqvtn     81            n        360           wni
whales,   of_age_class: young,   disturbed_by: vessels
theme	  p1            a1       p2             a2

This kind of syntax is called infix notation, as opposed to prefix- or Polish notation; it is also used in mathematics, where a relation like "subtraction (eight, three)" is usually written 8 - 3 rather than - 3 8. Infix notation is equivalent to a set of relationships all involving the theme as a common argument:

age_class (whales, young)
disturb (whales, vessels)

Each phenomenon can potentially act as a theme with several facets as its arguments, that is, as attributes able to specify the theme where they are expressed (numbering explained in next section):

whales, of_sex a9, of_age_class a81, in_place a2, ...

gift, of_good a8, to_beneficiary a7, by_giver a6, ...

The possible arguments of a facet, like "fetus, young, adult, old", are also called its foci in Ranganathan's classical faceted classification terminology, or the possible values of that attribute in computer science.

In linguistic terms, our basic syntactical structure can be represented according to transformational grammar, in its prefix-notation version introduced by Fillmore [Fil68; Hut65]:

                subject statement
              /   |           |   \  
            /     |           |     \
          V       a1          a2     a3
        /         |           |        \ 
      give     giver   beneficiary    good

For example,

                subject statement
              /   |           |   \  
            /     |           |     \
          V       a1          a2     a3
        /         |           |        \ 
      give      John        Jill      plants

      "giving by John to Jill of plants"
i.e.  "John giving Jill plants"

The relationship above is also asymmetrical: if "Jill" is in a2, it means that she is the beneficiary, but if she is in a3 it means that she is the giver. This makes it necessary that the expressed arguments are identified as for their role, that is, whether they are a1, or a2, or a3, etc. Natural languages do this basically in two ways: either by the position of each argument, or by a role marker such as a preposition or a case termination. In the former solution, adopted by English, the phrase John giving Jill plants makes it clear that Jill is the beneficiary by expressing it immediately after the verb; in the latter, adopted by French, John donnant à Jill des plantes makes it clear that Jill is the beneficiary by the preposition à.

Notice that the position option requires that all arguments be always expressed, even when they are not needed: to mean "givings to Jill" we should write give (some good, Jill), as give (Jill) would rather have the unusual meaning "giving of Jill". The same problem occurs in computer sciences with databases where each field has a fixed length: to avoid ambiguities, the field must be left blank if no content is available for it, thus wasting space; while databases with fields of variable length can save space, at the cost of using additional markers to identify the content of each field.

In the early versions of Ranganathan's Colon Classification, each successive facet was separated by a comma, making it mandatory to express all facets; later Ranganathan devised different symbols (comma, semicolon, colon, dot, apostrophe) to introduce facets with different meanings and make them optional: he thus introduced role markers in classification, or, as we usually call them, facet indicators. ILC uses digits as facet indicators in the same way as Colon uses punctuation marks.

General facets »

References cited in this section

Fil68: The case for case / Charles J Fillmore = Universals in linguistic theory. P 1-88 / Bach, Harms : ed' – Holt Rinehart and Winston : New York : 1968

Hut75: Languages of indexing and classification. P 56 / WJ Jutchins – Peregrinus : Stevenage : 1975


Integrative Levels Classification. Structure. Facets. Basic facet syntax / Claudio Gnoli – ISKO Italy : <> : 2010.07.22 - 2011.07.29 -

  Integrative Levels Classification project scheme monograph references