2 Background and Related Work
2.2 Central Concepts
2.3.1 Conceptual analysis method
2.3.1.3 Differences and similarities in conceptual analysis method ontologies
There are various differences and similarities between the ontologies which represent the conceptual analysis method. The editors of the A Thesaurus of Old English, in the
introduction to the work (Roberts et al., 1995, p. xxv), suggest that "Schemes of classification have no inherent truth, but represent the best attempts of the compilers to present their
materials within a coherent and illuminating framework." Fischer (2004, p. 49, 54–55) postulates that a truly universalist scheme cannot even exist, firstly, because there is no
general consensus about what is a "natural" or "logical" order of things, and secondly, because any scheme will be coloured by the culture from which it originates and by the language in which it is written. Moreover, Fischer points out that classifications also differ for the reason that categorization is a multidimensional operation in that human beings will see most concepts as belonging to several categories. Kay et al. (2009, pp. xix) propose very similar thinking in their introduction to the HTOED. They remark that no semantic category is likely to be wholly clear-cut and cite the example of the categories of "Music" or "Religion". Their content is typically well-defined, but this brings into question as to how religious music
should be categorized. A corresponding issue arose in section 2.2.3.3: Wilson and Thomas (1997, pp. 58–59) point out that words do not always fall conveniently into predefined semantic fields and cite as an example the word "sportswear" which could be classified both in the semantic field of clothing and in the semantic field of sports. This view is also shared by Hüllen (2006, pp. 14–15; 2009, p. 60) and Schmidt (1986, p. 788). Schmidt compares the chase after the purely "objective" system to the chase after the same illusion as Kant’s "pure object" and states:
The true test for any pre-established conceptual system can only be to what extent it is acceptable to as many human minds as possible beyond the boundaries of individual languages and cultures. That means the higher the degree of abstraction the greater is the likelihood of universal acceptance. As long as we cannot reach general
understanding at this higher level of abstraction we cannot possibly find it on the level of specific meaning. This does not mean that there should be only one system, it just means that the basic ingredients of each system should be the same, while there could be many different degrees of differentiation as well as differences in hierarchical order. (Schmidt 1986, p. 788–789)
In fact, Archer, Rayson, Piao, & McEnery (2004, p. 817) have observed that even though many semantic category systems are different in terms of their structure and granularity, they often agree to a greater or lesser extent on the basic major categories they contain. Schmidt (1986, p. 788) even suggests that a simple conversion program could rearrange any
conceptual dictionary from one conceptual system to another.
The similarity is clearly evident in the conceptual systems presented above, despite the fact that they are arranged differently. They vary a great deal as to the depth of the hierarchy and the number of categories they include, but they comprise the same "basic ingredients". In
addition to the differences discussed earlier, there are some other disparities as well.
McArthur’s categorization appears the most practical, including several categories for
concrete entities, whereas, for example, the majority of Roget's and Dornseiff's categories are
more abstract by nature. Interestingly, McArthur’s 20 top level categories also include "Entertainment, Sports, and Events". A category covering those topics does not exist on such a high level in any of the other deep hierarchies. In the HTOED, there is a second level category "Leisure" which subdivides on the third level into the categories "Entertainment", "Social", "The Arts", "Sport", and "Dancing". In Roget’s plan of classification, the category "Leisure" is placed on the third level. In Hallig and von Wartburg's system, there is a fifth level category "Celebrations, Games, Amusements" which further expands into the subcategories of "Festivals/Festivities", "Games/Diversions", "Sport", and
"Traditions/Customs".
Another interesting difference is the treatment of religious and supernatural issues. With regard to the deep hierarchies, Roget, McArthur, and Hallig and von Wartburg place them in the same top level category, whereas in the HTOED, they are placed in separate top level categories. Within the HTOED, there is a second level category called "Faith" under the top level category "The Social World", whereas the second level category "Supernatural" can be found under the top level category "The External World". With regard to the shallow
hierarchies, Laffal’s one level system contains the categories "HOLY" (referring to religious figures, activities, and objects) and "MYTH" (referring to the supernatural, the mythical and the magical), and, similarly, Louw and Nida's system contains the top level category
"Supernatural Beings and Powers" as well as the top level category "Religious Activities". In addition, Louw and Nida's system has a second level category "Be a Believer, Christian Faith" under the top level category "Hold a View, Believe, Trust". In comparison, Dornseiff's
includes topics related to both religion and the supernatural in the same top level category. Most of the category systems discussed here represent a modern view of the world and may not be altogether applicable to historical texts14, and the division between religion and the supernatural is a good example of such a case. In the modern world, they are considered two different issues, while, in earlier times, they were not necessarily separate from each other.
Looking back at the features 3–5 which Wilson and Thomas (1997, pp. 55–57; see section 2.2.3.3) suggest to be taken into consideration when choosing which system to use or when developing a new system:
3) The system should be adaptable to possible amendments which are necessary for treating a different period, language, register, or textbase.
4) Related to point 3, the system should operate at an appropriate level of granularity. This means that the annotation system should contain conceptually related words at varying levels of generality. There is no absolute in terms of granularity, but the correct level depends at least partly on the aims of the end user.
5) Related to point 4, a hierarchical structure would be an advantage for being able to adjust the granularity to the aims of the end user. If a system had a hierarchical structure based on increasingly general levels of related words, it would be possible to identify all these different levels without having to try to decide which is the level the end user wishes to employ. Indeed, it would be easy for the end user to look at all the different levels by simply moving up or down to the next level of granularity in the hierarchy.
On the basis of these guidelines, systems which are built as deep hierarchies would be a more practical choice than systems which are built as shallow hierarchies. Indeed, a deep hierarchy
14 An exception to this is the
is more flexible, since the end user can determine the appropriate level of granularity
depending on the task at hand. Furthermore, it would be easier to amend a deep hierarchy if a particular task requires.
Even though most of the work discussed here has concentrated on the English language, there are many conceptual systems for other languages as well, in addition to the systems mentioned above. One such is Paul Fortier’s (1989) ontology for his computer-aided analysis system which he used for the analysis of French prose fiction. Moreover, Julio Casares (1942) compiled the Diccionario ideológico which is the only existing large thesaurus for Spanish and was later made available in electronic format (Valderrábanos, Dïaz, & Pérez, 1994). To date, no thesauri have been compiled for Finnish. There are two synonym dictionaries, the Synonyymisanakirja ("Synonym Dictionary") (Jäppinen, 1989) and the Synonyymisanasto ("Synonym Lexicon") (Leino & Leino, 1990), but these are simply synonym finders which do not utilize any categorization but only list total and partial synonyms in their entries.