8. Approach for the management of quasi-equivalent concepts
8.1 The decision process
In order to define the term of quasi-equivalent concepts it was decided to use the definition given by E. Hovy, T. Mitamura, F. Verdejo, J. Araki, and A. Philpot in the article "
Events are Not Simple: Identity, Non-Identity, and Quasi-Identity". "Mention1 is quasi- (partially) identical to mention2 if they refer to the ‘same’ DE but one mention includes information that is not contained in the other, not counting information understood/inherited from the context." [1]. From a semantical point of view, they are not fully identical, but the core part is. They can be replaced by each other, but some information might be changed, added, or lost. The importance of this information for the respective domain is the main motivation whether to keep concepts separate or to merge them.
To analyze the information that can be changed, added, or lost in case of merging concepts and to support the decision whether this additional information is important and quasi-equivalent concepts must be kept separate the ontology engineer can use additional resources.
The resources can be used in combination to build a stronger justification for the decision.
The most common and reliable recourse to consult is the expert of the respective domain or domains in case the ontology is cross-domains. A recent trend in ontology engineering research aims at encouraging the active participation of domain experts in ontology creation. [59] It has multiple benefits as it minimizes the dependency on knowledge engineers at every step of ontology development, the most common case is to ask domain experts to provide relevant knowledge sources. A domain expert can judge how important is information,
65 that can be changed, added, or lost in case of merging and decide if it is relevant and beneficial for the ontology to keep the quasi-equivalent concepts separate.
Another resource that is easy to use is thesauri. An ontology engineer can compare the terms of the concepts in thesauri to check if synonym sets in thesauri, possibly containing the labels of both concepts as synonymous terms. If yes, it is a strong justification to merge the concepts. However, in some cases, even synonyms can have slightly different meanings. This difference might be beneficial for the ontology domain or the research. To prove the decision, it is possible to use dictionary or encyclopaedic (incl. wiki) definitions of the concepts. The definition helps to clarify in what cases the meanings of the quasi-equivalent concepts are overlapping and what exactly is the difference. The empirical contexts of the use of both terms in textual corpora also provide required information about the difference of quasi-equivalent concepts.
Not only the meaning of concepts can be checked but also sub-classes and instances of the concepts. The overlap of their sets of relevant sub-classes (i.e., the labels of subclasses proper, not of their instances) can show the degree of equivalence. It is important to take into account if concept 1 has, for example, four classes and concept 2 has the same four sub-classes and one extra. If concept 1 and concept 2 are merged, the information, that concept 1 did not have an extra sub-class and concept 2 has will be lost. The same can happen with the overlap of existing data instances (of the pre-cursor/s of the concepts).
Other resources based on expert experiences can also be used to find differences and support decisions. Various ontologies can also be checked to see how the same or similar case was solved by another ontology engineer.
After additional resources are reviewed an ontology engineer should define the motivation of the decision. The difference between quasi-equivalent concepts that have been discovered with additional resources assistance can motivate to separate concepts.
Additionally, there are stakeholders behind each of the two concepts who preferred (or might have preferred) to see it as separate. Moreover, few or no plausible axioms could be formulated for the merged concept, while the separate concepts could be axiomatized more richly. Any other options of motivation to keep concepts separate are possible based on the expert experience or an evaluated case.
66 When the difference between quasi-equivalent concepts that have been discovered with additional resources assistance is not significant, an ontology engineer might choose to merge quasi-equivalent concepts. The motivation can be to keep ontology small, for manageability or comprehensibility concerns. The problem of overlapping knowledge becomes critical with the growing usage of ontologies. Merging the concept helps reduce duplicated information. It allows keeping all respective data instances under the same type, making the management of data easier. Any other options of motivation to merge concepts are possible based on the expert experience or an evaluated case.
After a decision is made and justified by certain motivation, an ontology engineer should choose what concept name to use in case of merged concepts or what link to establish between separate concepts.
If an engineer decides to merge the concepts, the name of the concept should be chosen.
It is important to compare sub-classes and instances to keep the structure of an ontology.
If an engineer decides to keep concepts separate, the following links can be used:
owl:equivalentClass, owl:sameAs, rdfs:subClassOf, skos:exactMatch, skos:closeMatch. Any other option of a link is possible based on the evaluated quasi-equivalent concepts. It is recommended to establish the link between quasi-equivalent concepts to not lose information, that two concepts are close, but merging is not recommended. Since the decision is taken based on certain motivation, it is valuable to highlight it via linkage of concepts.
According to LOV analysis, the predicate that is used the most is owl:equivalentClass, which appeared almost two times more often than owl:sameAs. The predicates of SKOS (skos:exactMatch and skos:closeMatch) are rare in use to define the level of identity or similarity within LOV ontologies. The predicate skos:closeMatch has a lower level of similarity of the linked subject/object than skos:exactMatch. Nevertheless, the difference between the usage of close match and the usage of exact match is difficult to observe.
In case quasi-equivalent concepts have a parent-child relationship, it is possible to link them with rdfs:subClassOf. It is not recommended to link separated concepts with rdfs:seeAlso.
It states that one concept may provide additional information about another concept. Hence the information about quasi-equivalent relations is lost.
67 Flowchart 1. The decision process for the management of quasi-equivalent concepts
68 To summarize the decision process, to choose to merge concepts or keep them separate below five steps should be applied:
1. Choose 2 concepts
2. Consult additional resources
3. Find the different information that can be lost in case merging 4. Define the downside of each decision
5. Decide if the information is relevant for the ontology purpose
▪ In case of merging - choose the concept name
▪ In case of separating - choose the link to connect concepts 8.2 Test of the decision process
Some concepts from the LOV analysis have been chosen to test the decision process.
Since in the LOV analysis there are only separated concepts, it is possible to find a difference that can be lost in the case of merging and define the downside of merging and separation. For each pair, the current decision to keep concepts separate can be confirmed or rejected based on the outcome of the decision process. The analysis will be done based on 1 or 2 examples for each of the most common links from the 9th question of the questionnaire:
owl:equivalentClass, owl:sameAs, rdfs:subClassOf, skos:exactMatch, skos:closeMatch.
Dictionary, encyclopedic definitions, and synonyms thesauri are used during the analysis of additional resources as domain expert consultation is not always available. The Oxford English Dictionary [64] and Cambridge Dictionary [65] have been chosen to provide the definitions of the terms. Additionally, each concept has a definition or explanation within the respective ontology.
8.2.1 owl:sameAs
The first example linked by the owl:sameAs is the pair Address (http://sparql.cwrc.ca/ontologies/cwrc#Address) and Postal Address (http://schema.org/PostalAddress). For the term Address Oxford Dictionary provides multiple definitions:
1. Guidance.
2. An act or instance of addressing someone.
69 3. The action of directing or dispatching.
4. An introductory inscription dedicating a literary work to a specified person, god, etc.; a dedication.
5. A manner or style of speaking or writing; the manner a person employs in conversation or writing; language.
6. The particulars of the place where a person lives or an organization is situated, typically consisting of a number, street name, the name of a town or district, and often a postal code; these particulars considered as a location where a person or organization can be contacted by post. [64]
The Ontology description of the concepts states that Address is a mailing or street address, which relates to the 6th Oxford definition. That means that the rest of the meanings are not taken into account by the ontology. Cambridge Dictionary gives only two definitions:
one in the category of home details - the number of the house, name of the road, and name of the town where a person lives or works, and where letters can be sent; a place where someone lives. And the second definition belongs to the category of computes - a series of letters and symbols that tell you where to find something on the internet or show where an email is sent to the place where a piece of information is stored in a computer's memory. [65] The Ontology definition covers the first definition of the Cambridge Dictionary.
The second concept from the pair is Postal Address. In Oxford Dictionary the definition of the term Postal Address is equal to the 6th Oxford definition of the term address. Cambridge Dictionary does not provide any definition for the term Postal Address. The Ontology definition states that Postal Address is the mailing address. Additionally, it also includes the definition for the term Address, which is a physical and mailing address of the item.
The difference between the two concepts is the information that postal address is only mailing address and address includes both physical and mail. In case physical address and mailing address should be kept separated, the concepts also should be kept separated.
The second example linked by the owl:sameAs is pair Bell (http://w3id.org/charta77/jup/Bell) and Doorbell (http://dbpedia.org/resource/Doorbell). The Oxford English Dictionary gives below three definitions of the term 'Bell':
70 1. A hollow body of cast metal, formed to ring, or emit a clear musical sound, by the sonorous vibration of its entire circumference, when struck by a clapper, hammer, or other appliance. The typical form, found in all large bells (and indicated by the expression bell-shaped), is that of an inverted deep cup with a recurving brim, which is struck by a ‘clapper’ or ‘tongue,’ usually suspended from the centre of the interior.
2. A bell rung to tell the hours; the bell of a clock; whence the obsolete phrases of the bell, on the bell, at the bell = o'clock.
3. The larger kinds are used for giving signals of various import (time, danger, etc.) to the inhabitants of a town or district, and especially in connection with public worship (cf. chime n.1); the smaller kinds are used for similar purposes in a house (e.g. doorbell, dinner-bell, electric-bell).[64]
Cambridge Dictionary provides only two definitions of the term 'Bell': the first one is an electrical device that makes a ringing sound when you press a button (also doorbell).
The second one is a hollow metal object shaped like a cup that makes a ringing sound when hit by something hard, especially a clapper.[65]
The Ontology description includes all descriptions that it is possible to find in both dictionaries: A bell is a directly struck idiophone percussion instrument. Most bells have the shape of a hollow cup that when struck vibrates in a single strong strike tone, with its sides forming an efficient resonator. The strike may be made by an internal "clapper" or
"uvula", an external hammer, or—in small bells—by a small loose sphere enclosed within the body of the bell (jingle bell). Bells are usually cast from bell metal (a type of bronze) for its resonant properties but can also be made from other hard materials. This depends on the function. Some small bells such as ornamental bells or cowbells can be made from cast or pressed metal, glass or ceramic, but large bells such as a church, clock and tower bells are normally cast from bell metal. Bells intended to be heard over a wide area can range from a single bell hung in a turret or bell-gable, to a musical ensemble such as an English ring of bells, a carillon or a Russian zvon which are tuned to a common scale and installed in a bell tower. Many public or institutional buildings house bells, most commonly as clock bells to sound the hours and quarters. Historically, bells have been associated with religious rites, and are still used to call communities together for religious services.
71 Later, bells were made to commemorate important events or people and have been associated with the concepts of peace and freedom.
The second concept from the pair is Doorbell. In Oxford Dictionary the definition of the term Doorbell relates to the term Bell in ecclesiastical use, as bearing-bell, houseling-bell, lich-houseling-bell, sacring-houseling-bell, sanctus-houseling-bell, or saunce-bell. This meaning corresponds with the third Oxford definition of the term Bell. Cambridge also provides one definition - a bell, operated by a button on or next to the door of a house, that you push to tell the people inside that you are there. [65] This meaning corresponds with the second Cambridge definition of the term Bell.
The Ontology description for the concept 'doorbell' correlates with the definitions from Cambridge and Oxford dictionaries - A doorbell is a signaling device typically placed near a door to a building's entrance. When a visitor presses a button the bell rings inside the building, alerting the occupant to the presence of the visitor. Although the first doorbells were mechanical, activated by pulling a cord connected to a bell, modern doorbells are electric, operated by a pushbutton switch. Modern doorbells often incorporate intercoms and miniature video cameras to increase security.
The difference between the two concepts is large the Doorbell is only one type of Bell.
The term Bell is wider and includes more meanings than the term Doorbell. In case of merging and using the concept Bell, the information about the exact type of the Bell - Doorbell will be lost.
The third example linked by the owl:sameAs is pair Tourist Guide (http://w3id.org/charta77/jup/TouristGuide) and Tour Guide (http://dbpedia.org/resource/Tour_guide). The Oxford English Dictionary terms Tour guide and Tourist Guide are linked as synonyms and have the same definition: tourist guide = tour guide at tour; a guidebook which introduces tourists to a region or locality. [64] Cambridge Dictionary does not have a definition for Tourist Guide, but it provides three definitions for the tour guide. First one - guide as a book, that gives you the most important information about a particular subject. The second one - guide as help. Something that helps you form an opinion or make a decision about something else. The third one - guide as a person, whose job is showing a place or a particular route to visitors.[65]
72 In Ontology the description is a tour guide (U.S.) or a tourist guide (European) is a person who provides assistance, information on cultural, historical and contemporary heritage to people on organized sightseeing and individual clients at educational establishments, religious and historical sites such as; museums, and at various venues of tourist attraction resorts. In the Ontology description, it is mentioned that tour guide is used mainly in the U.S., and tourist guide is used mainly in Europe. However, there is no confirmation neither in Oxford Dictionary nor Cambridge.
Based on additional resources analysis there is no information that could be lost in case of merging the concepts of tour guide and tourist guide.
8.2.2 skos:exactMatch
The first example linked by the skos:exactMatch is pair Lift (https://w3id.org/timebank#Lift) and Elevator (http://dbpedia.org/resource/Elevator) For the term Elevator Oxford Dictionary provides below three definitions:
1. A machine used for raising corn or flour to an upper storey.
2. chiefly North American. A lift, hoist, ascending chamber.
3. One who or that which elevates. [64]
The Oxford definition for the term lift shows equivalence with the term elevator - An apparatus for raising or lowering persons or things from one floor or level to another; an ascending chamber or compartment; = elevator n. 3d. Also, the well or vertical opening in which the apparatus works [64].
Cambridge Dictionary links both terms and provides only one definition for both. Lift noun (CARRYING DEVICE) (US elevator) - a device like a box that moves up and down, carrying people or goods from one floor of a building to another or taking people up and down underground in a mine. [65]
The Ontology description links both terms into one definition. An elevator (North American English) or lift (Commonwealth English) is a type of cable-assisted, hydraulic cylinder-assisted, or roller-track assisted machine that vertically transports people or freight between floors, levels, or decks of a building, vessel, or other structure. They are typically powered by electric motors that drive traction cables and counterweight systems such as a
73 hoist, although some pump hydraulic fluid to raise a cylindrical piston like a jack. In agriculture and manufacturing, an elevator is any type of conveyor device used to lift materials in a continuous stream into bins or silos. Several types exist, such as the chain and bucket elevator, grain auger screw conveyor using the principle of Archimedes' screw, or the chain and paddles or forks of hay elevators. Languages other than English, such as Japanese, may refer to elevators by loanwords based on either elevator or lift. Due to wheelchair access laws, elevators are often a legal requirement in new multistory buildings, especially where wheelchair ramps are not possible. Some elevators can also travel horizontally in addition to the usual vertical motion.
The difference between the two concepts is the territory of usage. An elevator is used in North American English and a lift is used in Commonwealth English. If the territory of usage is not relevant for the ontology, terms can be merged into one. The information to be lost is the territory of usage.
8.2.3 owl:equivalentClass
The pair Purpose (http://softeng.polito.it/rsctx#Purpose) and Goal (http://www.ontologydesignpatterns.org/ont/dul/DUL.owl#Goal) is the first example linked by the owl:equivalentClass. For the term Purpose Oxford Dictionary provides below three definitions:
1. That which a person sets out to do or attain; an object in view; a determined intention or aim.
2. The reason for which something is done or made, or for which it exists; the result or effect intended or sought; the end to which an object or action is directed; aim.
3. That which forms or ought to form the subject of a discourse; the matter in hand; the point at issue. [64]
Cambridge Dictionary as well gives three definitions for the term Purpose, which are the same as Oxford definitions:
1. purpose noun - why you do something or why something exists
2. purpose noun (REASON) - an intention or aim; a reason for doing something or for allowing something to happen