A Tale of Two Properties; or, CIDOC CRM P88 and P89
Mark R. Lindner
LIS 590OH Spring 2007
Prof. Allen Renear
CIDOC CRM has four properties of classes (actually relations between two classes) that hold between entities of the class E53 Place. These are P88 consists of (forms part of), P89 falls within (contains), P121 overlaps with, and P122 borders with. The last two are purely symmetrical properties, which is why there is no corresponding verbal distinction that depends on the order of interpretation.
P121 and P122 both represent “purely spatial” properties that seem fairly straightforward (74). The example for P121 could use a bit more clarification or another example or two, especially for non-Americans. Clearly, though, since part of Alaska overlaps with the Arctic “the territory of the United States (E53) overlaps with the Arctic (E53)” (74). These two properties could be influenced by some intriguing philosophical issues within the subject of mereology and, in particular, fiat and bona fide boundaries ala Smith and Varzi. While these would be interesting issues to explore, it may also be the case that there are means to distinguish these issues within the CRM, as these issues could cause problems with certain inferences if not distinguished. Nonetheless, my focus in this paper will be on the first two properties mentioned.
A note about my methodology: While not entirely out of hand, I have just dismissed a primary tool of the logician and modeler. This was not done lightly. Sometimes, though, the expertise of the domain practitioner trumps that of the modeler and philosopher. In our case, the modelers are attempting to model a practitioner’s domain, thus, the practitioner’s domain expertise takes priority, although it likely that the modelers and the practitioners need to engage in further dialogue.
March 6th, in class, we had an interesting discussion re P88 and P89. These are relationships of places to other places and they, respectively, label properties of "consists of (forms part of)" and "falls within (contains)" between Place(s) E53. This is a difficult difference to parse out, made none the clearer by the supplied scope notes and examples. While mereological issues may well play a role in distinguishing these two properties, I believe that bringing them to bear on the situation will only serve to make it more complex than can be dealt with in a few pages. I also fully believe that resources available in the thesaurus construction standards can fully serve to disambiguate these two properties.
At the heart of the confusion is what exactly lies in the difference between “consists of (forms part of)” (P88) and “falls within (contains)” (P89). The scope note for each property is the natural place to begin.
P88 consists of (forms part of): This property identifies an E53 Place that forms part of another Place. It supports the notion that a Place can be subdivided into one or more constituent parts. It implies both spatial and contextual containment relationships between the two Places (64).
P89 falls within (contains): This property identifies the instances of E53 Places that fall within the area covered by another Place. It addresses spatial containment only, and no ‘whole-part’ relationship between the two places is implied (64).
The scope notes give us a first move towards disambiguation. P88 “implies both spatial and contextual containment relationships,” while P89 addresses “spatial containment” only. I take the comment regarding “no ‘whole-part’ relationship” in P89 to contrast with the comment in P88 about “contextual relationship.” To pull apart the contrast in these two concepts I will now turn to the thesaurus standards.
Regarding hierarchical relationships, as a whole, all sources state in some manner that, "Terms are hierarchically related only if both are members of the same fundamental category, that is, they both represent entities, activities, agents or properties, etc. (Aitchison, 55, emphasis in original). Let us be very careful here, based on further examples, none of these sources are claiming that all entities are related by being in the same fundamental category. Only that, as a first cut, things must be in the same fundamental category to possibly participate in a "true" hierarchical relationship. If they do not pass this test, they are at best in an associated relationship.
For example: In a hierarchy of grammatical word forms, the category of helping verbs is subsumed by the category of verbs, both being in the same fundamental category; namely, verbs. Of any two words, if one is a helping verb and the other is a pronoun they simply cannot be in a hierarchical relationship as they belong to different fundamental facets in this scheme; namely, one is a verb and the other a noun.
“Hierarchical” as used within the thesaurus standards and practice implies levels of superordination and subordination, utilizes a logically progressive sequence, and includes inheritance. The allowed hierarchical relationships are generic (or genus-species, i.e., “is-A”) relationship, the hierarchical whole part (partitive, i.e, "is a part of") relationship, which is covered by the domain of mereology, the instance ("is an example of") relationship, and the polyhierarchical, in some cases.
So, what are fundamental categories? That is next but, first, a comment on what I mean by "as a first cut." This is simply to claim that methodologically one begins with the application of a logical principle of division. All of these lists claim to not be exhaustive. The purpose of such division, or faceting, "using only one characteristic (or principle) of division at a time, [is] to produce homogeneous, mutually-exclusive groups” (Aitchison, et. al., 70).
Fundamental categories are the sorts of entities that comprise what might be called “basic divisions of the world.” Entities such as concrete entities, abstract entities, and entities referred to by proper names (ISO 2788-1986 per Aitchison, et. al., 18) or Ranganathan's famous PMEST formula for Personality, Matter, Energy, Space and Time are only two ways of many to divide the world into fundamental facets. They are simply a methodological move "modelers" make to begin putting things together into various kinds of representations of the way we (humans) experience the world.
In general, there is an important distinction to be made regarding whole-part relationships and hierarchical relationships. All three standards (American, British and international) and Aitchison, et. al. claim that most whole-part relations are not hierarchical relations. ISO 2788 defines the hierarchical whole-part relationship as covering “a limited range of situations where the name of a part implies the name of its possessing whole in any context. The terms can then be organized as a hierarchy, the name of the whole serving as the superordinate term, and the name of the part as the subordinate term” (16). The only allowed whole-part hierarchical relations are of four types: (1) Systems and organs of the body, (2) Geographical location (our interest here), (3) Discipline or field of study, and (4) Hierarchical social structure. The geographical example given by all standards is the same:
The example for P88 fits well with this exemplification: “the area covered by the London Borough of Islington in 1976 (E53) forms part of the area covered by Greater London in 1976 (E53)” (64). Further evidence to support this is forthcoming as we now turn to the problem at hand.
According to the Wikipedia article on Greater London, “Greater London is the top-level administrative subdivision covering London, England. The administrative area was created in 1965 and covers the City of London and 32 London boroughs” [of which, Islington is one]. According to the Wikipedia article on London boroughs, “The London boroughs are administered by 'London Borough Councils' ....” Thus, Greater London and the Borough of Islington are both administrative units.
As stated above, a key clue towards disambiguating these two properties comes from the scope note for P89: "It addresses spatial containment only, and no ‘whole-part’ relationship between the two places is implied" (64). Although they are not directly ruling out the possibility of a hierarchical whole-part relationship (I believe that they should be!), they are attempting to differentiate this from P88, which is (and I believe must be) a proper hierarchical whole-part relationship.
Unfortunately, the example used for P89 is not altogether clear, at least not as a single example. Perhaps more examples would help clarify the difference. “[T]he area covered by the World Heritage Site of Stonehenge (E53) falls within the area of Salisbury Plain (E53)” (64). To provide some needed context, Wikipedia tells us a World Heritage Site “is a specific site (such as a forest, mountain, lake, desert, monument, building, complex, or city) that has been nominated and confirmed for inclusion on the list maintained by the international World Heritage Programme administered by the UNESCO World Heritage Committee, …” and that “[t]he programme aims to catalogue, name, and conserve sites of outstanding cultural or natural importance to the common heritage of humankind.”
The Columbia Gazetteer of the World informs us that Salisbury Plain is an “undulating chalk plateau (area: c.300 sq mi/777 sq km), Wiltshire, S England, between the Hampshire Basin and the Vale of Pewsey.” Thus, Salisbury Plain is a typical geographical feature which, while certainly given a name via a culturally determined appellation process, exists regardless of whether it is named or not. But the same cannot be said for any World Heritage Site, which exists simply because it has been stipulated as such. Thus, Salisbury Plain and “the World Heritage Site of Stonehenge” are, in fact, in different fundamental categories.
Making use of the context provided above, we find that the examples provided by the CIDOC CRM for P88 and P89 are appropriate. The example used in "P88 consists of (forms part of)" is appropriate because Greater London and the Borough of Islington are both administrative units (same kind of entities within the same fundamental category) of which the Borough of Islington is by definition a constituent part, and which is both spatially and contextually contained by Greater London. The example in "P89 falls within (contains)" is also appropriate, even if not as clear as it could be. A World Heritage Site is simply not of the same fundamental category as a plain (contextual entity versus physical entity), nor even if it were of the same fundamental category is it of the same entity type.
As a possible useful second example for distinguishing P89, our in class example of Lesotho and South Africa also works. According to Wikipedia, Lesotho, “officially the Kingdom of Lesotho, is a land-locked country, entirely surrounded by the Republic of South Africa." While they are both of the same fundamental categories and the same kind of entities, nonetheless, Lesotho in no way is a constituent part of South Africa. It only has the (historically) contingent geographic property of being surrounded by South Africa. In that respect, Lesotho falls within South Africa and, reciprocally, South Africa contains Lesotho. But Lesotho neither forms part of South Africa (spatially and contextually), nor does South Africa consist of Lesotho (and other entities) in the manner in which P88 is stipulated.
Thus, using the concept of hierarchical whole-part relationships contained within three standards for thesaurus construction, and one of the leading texts on the subject, we are able to fully disambiguate the confusion contained in the CIDOC CRM’s less than clear scope note and examples for the two properties of P88 consists of (forms part of) and P89 falls within (contains).
Can we make this a bit more relevant to our concerns within the bibliographic universe? Looking at FRBROO, the only Property listed for F12 Place is R63 incorporates (is incorporated in), although it is not listed in the FRBR Property Hierarchy, nor in the notes for R63, which is a relation between F20 Self-Contained Expression and F2 Expression. Thus, I believe it is mistakenly listed as a property of F12 in the one place where it is.
I find it interesting that there are no Place - Place relations in FRBROO. Perhaps they are not needed, but I am supposing that these Properties (relationships) just might be required in, say, rare books cataloging. Thus, I'd like to take a preliminary stab at what P88 and P89 might translate into in the realm of F12 in FRBROO.
An example of F12 Place could be the title page of "the 1st edition of the novel titled ‘Da Vinci Code’, as it presents itself in the copy that was used to create a bibliographic record for that edition in the Library of Congress’s catalogue." Another instantiation of F12 is the publisher's statement on said title page of the same Item. A third instantiation of F12 place might be the LoC's received date stamp on said title page.
I envision the second example (publisher's statement) to be in a P88 consists of (forms part of) relationship with the first example (title page), and the third example (date received stamp) to be in a P89 falls within (contains) relationship with the title page.
The second example is a spatial and contextually contained part of the title page; at least as it applies to our cataloging codes and general publisher practice. The third is in a spatial relationship, but is not in a contextual containment relationship with the title page. It is only contingently the case that LoC puts the date-received stamp on the title page. It could just as easily be stamped anywhere.
One possible objection to my analysis is that the publisher's statement is also contingently placed on the title page and could be anywhere else. Granted. But, nonetheless, it is there and it is a contextually contained part of the package of information that a title page is designed to impart. The date-received stamp is not part of that same contextual package. Maybe something a little less mundane?
As I assume you are all aware, illuminated manuscripts were generally made in two stages, at least regarding content. Once the writing surface was prepared, scribes inscribed the text. When the text was complete, the illustrator added the illuminations, i.e., decorated initials, borders and miniature illustrations ("Illuminated manuscript", Wikipedia).
The Lichfield Gospels, "is an eighth century Gospel Book housed in Lichfield Cathedral. There are 236 surviving folios, eight of which are illuminated. Another four contain framed text. The manuscript is also important because it includes, as marginalia, some of the earliest known examples of written Welsh" ("Lichfield Gospels", Wikipedia).
It is my contention that the illuminations adhere to the P88 consists of (forms part of) relationship, while the marginalia adhere to the P89 falls within (contains) relationships.
As overly broad as the F12 Place entity is in FRBRoo, I have no doubt that other important Place - Place relationships will need to be elucidated. I can certainly find examples of the two symmetrical Place - Place properties in CIDOC CRM, P121 overlaps with P122 borders with. What other currently non-specified Place - Place properties might we need?
Aitchison, Jean; Alan Gilchrist and David Bawden. Thesaurus construction and use: a practical manual, 4th ed. London: Aslib, 2000.
ANSI/NISO. Guidelines for the Construction, Format, and Management of Controlled Vocabularies, ANSI/NISO Z39.19-2005. Bethesda, MD: NISO Press, 2005.
British Standards Institution. BS 5723 : 1987 British Standard guide to establishment and development of monolingual thesauri. London, BSI, 1987 [This document is also the same as ISO 2788-1986 Documentation – Guidelines for the establishment and development of monolingual thesauri.]
British Standards Institution. (DRAFT) BS 8723-1:200X Structured vocabularies for information retrieval – Guide – Part 1: Definitions, symbols and abbreviations. London, BSi, 2004.
Crofts, Nick; et. al. Definition of the CIDOC Conceptual Reference Model, version 4.2, June 2005.
Doerr, Martin and Patrick Le Bouef. FRBR: object-oriented definition and mapping to the FRBRER (version 0.7.1), February 2007. http://cidoc.ics.forth.gr/frbr_drafts.html
“Greater London.” Wikipedia article. Accessed 8 March 2007. http://en.wikipedia.org/wiki/Greater_London
"Illuminated manuscript." Wikipedia article. Accessed 15 April 2007. http://en.wikipedia.org/wiki/Illuminated_manuscripts
International Organization for Standardization (ISO). ISO 2788-1986 Documentation – Guidelines for the establishment and development of monolingual thesauri. Switzerland: ISO, 1986.
“Lesotho.” Wikipedia article. Accessed 8 March 2007. http://en.wikipedia.org/wiki/Lesotho
"Lichfield Gospels." Wikipedia article. Accessed 15 April 2007. http://en.wikipedia.org/wiki/Lichfield_Gospels
“London boroughs.” Wikipedia article. Accessed 8 March 2007. http://en.wikipedia.org/wiki/London_borough
“Salisbury Plain.” Columbia Gazetteer of the World available via UIUC Online Reference Collection, under Countries, States, Cities. Accessed 8 March 2007.
Smith, Barry and Achille C. Varzi. (2000) “Fiat and bona fide boundaries.” Philosophy and Phenomenological Research 60 (2), March 2000: 401-420.
Smith, Barry and David M. Mark. "Geographical categories: an ontological investigation." International Journal of Geographic Information Science 15 (7), 2001: 591-612.
“World Heritage Sites.” Wikipedia article. Accessed 8 March 2007. http://en.wikipedia.org/wiki/World_Heritage_Site
This paper licensed under a Creative Commons Attribution-NonCommercial-ShareAlike license by Mark R. Lindner