Yutaka Matsuo*1, Masahiro Hamasaki*2, Junichiro Mori*3, Hideaki Takeda*2, Koiti Hasida*1
*1 National Institute of Advanced Industrial Science and Technology
*2 National Institute of Informatics
*3 University of Tokyo
In FOAF vocabulary, many kinds of relationships between people are deliberately simplified as "knows" relation. So far, there are several attempts to describe relationships between people. However, to strengthen the expressive power and at the same time to avoid confusion of many relationships, we have to make an ontological consideration in the first place. This paper makes some considerations on relationship between people, and based on that discussion we describe some proposals toward a new relationship vocabulary.
Recently FOAF [FOAF] is getting popular with two technologies: Semantic Web and social networking. Social networking becomes popular also in Japan. Currently two big social networking sites in Japan are Orkut and Gree, the latter is a Japanese original networking site. Most of social netwoking sites are based on a centralized database architecture, and FOAF can provide the standard for such systems to import/export the data depending on the user demand for control of "their" data.
One of the very interesting points about FOAF is that it can describe the relation to other people by using the property "knows." Against the background of heightened interests about social networking, a vocabulary to describe the relationship in more precise manner is needed so that people can describe their relationship as they want. Because a big attraction of social networking lies in the feeling that "we can connect to others," the relationship vocabulary is very important also for the excitement of FOAF.
In the older version of FOAF, there are a lot of variants about human relationship. However, the current version employs only "knows" relation because the main purpose of that relation is to enable describe the connected persons and jump to the person’s FOAF files by "seeAlso" property. There are several attempts to expand the "know" relation. However, if we propose more and more kinds of relations, it might be very confusing. We have to strengthen the expressive power and at the same time avoid confusion of many relationships. But currently no overall policy is proposed so as to build relationship vocabulary between people (as far as we know by published materials). In our opinion, ontological considerations are necessary for building good vocabularies.
This paper tries to consider what is a relationship between people and promote the debate on constructing and refining the current and future relationship vocabularies.
The rest of the paper is organized as follows. In section 2, we introduce some of the relationship vocabularies. In section 3, we describe that human relation is sometimes made up by, what we call, event coparticipation relationship. In section 4, we show another type of relationship called common property relationship. Based on these kinds of relationship, the communication/behavior relationships can be observed. Section 6 addresses related issues. We make proposals toward future relationship vocabulary in Section 7 and conclude this paper.
Current version of FOAF vocabulary has only "knows" property to describe relationship to others. Some feels it is too simple, but the main role of that relation is not to specify the characteristic of the relation but to specify the connected person's email and FOAF files by mbox and seeAlso property and enable linking FOAF files.
The old version of FOAF has 11 properties of relationship as shown in Table 1. However, to make the relation to be relatively harmless and inoffensive, only knows relation is survived. For example, knowsWell might cause a problem because A knowsWell B but B just knows A [Brickley03].
Table
1. Old FOAF relationship vocabulary. |
|
Property |
Description |
metMbox |
Met person with personal mailbox |
knows |
Knows person |
knowsWell |
Knows person well |
livesWith |
Lives with |
sigOther |
Significant other |
wife |
Wife |
husband |
Husband |
sister |
Sister |
brother |
Brother |
father |
Father |
mother |
Mother |
Recently, Ian Davis and Eric Vitiello Jr. propose a vocabulary to describe relationships between people called RELATIONSHIP [Ian]. It includes 33 properties as shown in Table. 1. All of them are defined as subproperty of "knows" in FOAF. Its domain and range is foaf:person. There are some debates about imperfection or redundancy, definition of each properties and so on. The vocabulary is currently under development.
Table
2. Vocabulary Properties of RELATIONSHIP. |
|
Vocabulary Property |
Description |
friendOf |
A person who shares mutual friendship with this person. |
acquaintanceOf |
A person having more than slight or superficial knowledge of this person but short of friendship. |
parentOf |
A person who has given birth to or nurtured and raised this person. |
siblingOf |
A person having one or both parents in common with this person. |
childOf |
A person who was given birth to or nurtured and raised by this person. |
grandchildOf |
A person who is a child of any of this person's children. |
spouseOf |
A person who is married to this person. |
enemyOf |
A person towards whom this person feels hatred, intends injury to, or opposes the interests of. |
antagonistOf |
A person who opposes and contends against this person. |
ambivalentOf |
A person towards whom this person has mixed feelings or emotions. |
lostContactWith |
A person who was once know by this person but has subsequently become uncontactable. |
knowsOf |
A person who has come to be know to this person through their actions or position. |
wouldLikeToKnow |
A person whom this person would desire to know more closely. |
knowsInPassing |
A person whom this person has slight or superficial knowledge of. |
knowsByReputation |
A person known by this person primarily for a particular action, position or field of endeavour. |
closeFriendOf |
A person who shares a close mutual friendship with this person. |
hasMet |
A person who has met this person whether in passing or longer. |
worksWith |
A person who works for the same employer as this person. |
colleagueOf |
A person who is a member of the same profession as this person. |
collaboratesWith |
A person who works towards a common goal with this person. |
employerOf |
A person who engages the services of this person. |
employedBy |
A person for whom this person's services have been engaged. |
mentorOf |
A person who serves as a trusted counselor or teacher to this person. |
apprenticeTo |
A person to whom this person serves as a trusted counselor or teacher. |
livesWith |
A person who shares a residence with this person. |
neighborOf |
A person who lives in the same locality as this person. |
grandparentOf |
A person who is the parent of any of this person's parents. |
lifePartnerOf |
A person who has made a long-term commitment to this person's. |
engagedTo |
A person to whom this person is betrothed. |
ancestorOf |
A person who is a descendant of this person. |
descendantOf |
A person from whom this person is descended. |
participantIn |
--- |
participant |
--- |
Joel De Gan et al. proposes MeNowDocument [Joel], which is a FOAF extension for defining often changing variables. It includes properties of "thinking", "listeningTo", "located" and so on. Especially, for relationships between people, it has isWith, linking, chatting, dating, crush, isMadAt and isHappyWith. Some of them are in testing or proposal status.
As a related issue to FOAF, Matsuo et al. propose an automatic detection algorithm of human relation from the Web [Matsuo03]. Currently it concentrates on researchers relation because the information of researchers are abundant on the Web compared to business people or students. It classifies human relation into four categories: Coauthorship, Lab (the same laboratory), Project (the same project), and Conf (participation to the same conference).
Orkut classifies the relation into "haven't met", "acquintance", "friend", "good friend", and "best friend." In summary, there are many relationships used in many systems and vocabularies.
We assume that a basic structure of human relationship is that persons participate in an event as shown in Figure 1. An event is classified as perdurant in DOLCE ontology, which is one of the most popular ontologies [Masolo03]. DOLCE ontology is based on a fundamental distinction between enduring and perduring entities, i.e., between what philosophers usually call continuants and occurrents. Similarly as DOLCE ontology, we assume that objects (e.g., persons) cannot be parts of occurrences, but rather they participate in them. In DOLCE ontology, event occurrences are called achievements if they are atomic, otherwise they are accomplishments, but we do not distinguish them.
Fig. 1: Participation to an event. |
For example, an event is a meeting, a conference, a baseball game, a walk, and so on. Sometimes, a person takes an indispensable role for an event, and sometimes not.
A person has properties. In DOLCE ontology, we say qualities. (Though there are some differences between properties and qualities, we do not consider the distinction because this paper aims to make a rough sketch of person relationship.) Properties include a name, an address, an e-mail, nationality, character, ability, interests and so on and so forth. It also includes past history of the person, for example, past affiliations, a graduating school, awards and so on.
From the viewpoint of the structure in Figure 1, properties of a person can be categorized as follows
The value of a property sometimes changes. For example, affiliation and position change. Interest and publication also changes over time.
Assume that person A and person B participate in Event X. In this situation, we call that A and B are in event coparticipation relationship under event X. Especially, if A and B take the same role to Event X, A and B are in same role relationship under event X. If A cannot take over B's role or B cannot take over A's role toward X, A and B are in the role-sharing relationship.
There are many distinction how a person contributes to an event. For example, in a baseball game, the degree and types of contribution differ for a pitcher, a catcher, a manager and an umpire. In functional ontology, there are such distinction of how each component contributes to the whole device function [Keuneke91, Sasajima95]. But for simplicity, in this paper we only distinguish whether the roles are the same or not.
Based on the basic structure of an event coparticipation relationship, we can consider many variations.
In RELATIONSHIP vocabulary, collaboratesWith property can be considered as event coparticipation relationship (while colleagueOf relation is rather common property relationship as described below). The mentorOf and apprenticeTo relation is the event coparticipation relationship to event "teach". The livesWith is also this kind of relationship but it focuses on superclass of events. The employerOf and employedBy is also event coparticipation relationship that focuses on control of participation. Coauthorship is the event coparticipation relation to events such as research, survey, discuss and so on.
There are relationships that are not based on event coparticipation. For example, person A and B are graduates of the same university. Even if A and B has not met before, when they are introduced and know that they are graduated from the same university, they feel some familiarity each other. The same nationality people sometimes make a community in another country. This is also an example of the relationship that is not based on the event coparticipation. We say this kind of relationship as common property relationship.
Fig. 2: Common property relationship. |
A common property relationship between people is generated by sharing of the same property value. For example, person A and person B have a common working place, common interests, and common experiences, then they are in common property relationship with regard to working place, common interests, and common experiences. The colleagueOf property in RELATIONSHIP vocabulary is the relationship with common working organization (sometimes in the same building and in the same group.) Lab and Project relationship are also of this type.
The common properties sometimes imply several other things: For example, two persons who share the same working place have common topics about the workplace and business, knows the same product or the same person of the organization, Mac users (compared to Windows users) share the topic of Mac and its software, and they sometimes tend to have preferences to stylish goods. Common property relationship generate common topics and empathy, and enables dealing and consultation based on the same role relationship to some events.
If person A and B are in either event coparticipation relationship or common property relationship, they often make communication. The communication media can be diverse: face-to-face conversation, telephone call, e-mail, chat, posting to Weblogs and so on.
Participation to event X sometimes results in observable information. For example, attending the same meeting might result in a meeting minutes. Participation to a baseball game results in players names in a score board and a score book. Collaboration of research makes coauthoring papers.
If we want to discover the relationship by observation, we have to estimate what kinds of relationships they are in from what we can observe. Actually, what we can observe is only superficial results. Therefore, we have to analyze and estimate the relationship underlying the observables. For example, assume two persons are talking much and we do not know the reason. If we have the profiles of both persons, and find that they are from the same university, then we can estimate that they are frineds of the university, i.e., the common property relationship. Thus, automatic discovery of relationship has always this kind of estimation: we have only observe the superficial information of communication/behavior level and should guess the underlying event coparticipation/common property relationships.
On the other hand, if a user declares what relationship s/he is in with others as in social networking services, we can predict the communication/behavior level vice versa. Therefore, we can recommend a person who is in the same birthplace to another person, expecting that those two will make much communication.
Fig. 3: Communication/behavior observation. |
Sometimes relationships are nominal and/or logical. For example, boss-follower relationship in a company is determined by the company's regulations. Also there is a difference between nominal marriage relationship and actual marriage relationship. Usually nominal relationship is intended to corresponds to the actual relationship, but sometimes there are differences.
Relationships including both event coparticipation relationship and common property relationship are explicitly or implicitly evaluated in comparison with other relations. For example, if we want to use "know" relation, it has to represent that we know someone relatively better than others do. We are in the event coparticipation relationship with many people as for event "talk," but we don't usually give special meaning to such event because it is very common. On the other hand, if one travels abroad with another, those two think they are in the event coparticipation relationship to the travel event because going to abroad is usually rare.
The same is true for common property relationship. We do not think that people who has nationality of Japan are in the common property relationship in Japan, but do so in foreign countries. We do not think that we all are Terran, but if alien attack the earth, we think that we are in the common property relationship of being Terran. Therefore, such relationships are not absolute but relative.
Assuming that there are two kinds of relationship, i.e., event coparticipation relationship and common property relationship, what can we utilize that knowledge in building a relationship vocabulary?
One thing that is very important is: if we want to describe an event coparticipation relationship, we have to identify the event, or at least we have some way to identify the event. For example, when we write mentorOf relation, there are events behind this relation. Therefore, sometimes we want to write from when to when and what subject the person teaches the another. When we write livesWith relation, we might want to describe where we live from when to when. Thus, event coparticipation relationship requires an event that stands between two persons.
Figure 4 is a directed label graph of event coparticipation relationship. Though it is sometimes difficult to assign URI to each event, some devices might makes this representation possible.
Fig. 4: Directed label graph of event coparticipation relationship. |
On the other hand, as for common property relationship, we can write the direct relationship between the two. However, it is not so a clever approach to write down all the common property relationship. Rather, we should write the properties for individuals and then calculate the relationship. For example, if all of one's colleagues will join the FOAF community, s/he has to write colleagueOf relation to many people. We prefer describing our office and then a system can automatically compute the colleague relationship. When we calculate the relationship, we have to consider the relative strength of the property depending on a context. As described in section 6.2, some common properties are important in some context.
Furthermore, there are many options to event coparticipation relationship as describe in 3.2. For example, like/dislike relationship can hopefully be written as "I like to going to movie with this person" or "I like to doing anything with this guy", that is, "I like this person." Therefore, it is an adjunct of the feeling to the event X participated by A and B. Gain/loss, control, and other options exist that are adjunct to the event. We have to clarify such independent factors, and try not to define the relation with many instantiated options, because that will make the vocabulary less expressive and confusing.
In this paper, we discuss the relationships between people from ontological point of view. Though the discussion is in a very preliminary stage, we hope that our paper potentially contributes to building the relationship vocabulary for FOAF by promoting discussions. Relationship is one of the most interesting parts of FOAF. We believe that more elaborate consideration on relationship will lead to more easy-to-write and expressive relationship vocabulary, and more excitement to FOAF in the future.
[Brickley03] Why FOAF has foaf:knows rather than foaf:friend..., http://rdfweb.org/mt/foaflog/archives/2003/08/04/01.23.34/
[FOAF] Dan Brickley and Libby Miller: FOAF Vocabulary Specification, http://www.foaf-project.org/
[Ian] Ian Davis and Eric Vitiello Jr: RELATIONSHIP: A vocabulary for describing relationships between people, http://vocab.org/relationship/
[Joel] Joel De Gan, MeNowDocument: A FOAF extension for defining often changing variables in FOAF, http://schema.peoplesdns.com/menow/
[Keuneke91] Anne M. Keuneke: Device representation: the significance of functional knowledge, IEEE Expert, 24:22-25, 1991
[Masolo03] Claudio Masolo, Stefano Borgo, Aldo Gangemi, Nicola Guarino, and Alessandro Oltramari: WonderWeb Deliverable D18, http://wonderweb.semanticweb.org/deliverables/D18.shtml
[Matsuo03] Yutaka Matsuo, Hironori Tomobe, Koiti Hasida, Mitsuru Ishizuka: Mining Social Network of Conference Participants from the Web. Proc. Web Intelligence 2003, pp. 190-193, 2003
[Sasajima 95] Munehiko Sasajima, Yoshinobu Kitamura, Mitsuru Ikeda, and Riichiro Mizoguchi: FBRL:A Function and Behavior Representation Language, Proc, IJCAI-95, p. 1830-1836, 1995