USING CULTURAL DIFFERENCES TO JOIN PEOPLE WITH

COMMON INTERESTS OR PROBLEMS IN ENTERPRISE

ENVIRONMENT

Gilberto Astolﬁ, Vanessa M. A. de Magalh

aes, Marcos A. R. Silva and Junia C. Anacleto

Computing Department, Federal University of S

ao Carlos, Washington Luis Road, S

ao Carlos, Brazil

Keywords:

Culture, Enterprise environment, Social networking service, Collaborative work.

Abstract:

This paper describes a methodology to identify people with common subjects but different cultures. Each

person has a culture and it inﬂuences in his way to express himself or to write, etc. Because of that, nowadays,

on the Web, it is very difﬁcult a person who writes ’validating enterprise system’ ﬁnds another who writes

’attesting B2B solution’ because there are different words. On the other hand, there is a common subject that

search engines can not identify, because it is necessary taking into consideration the cultural knowledge and

experience from each person. In this context, the methodology presented here intends to consider this cultural

information in order to identify when two or more people are talking about the same subject.

1 INTRODUCTION

Nowadays, the most of the people are concerned

about having more experience e knowledge in order to

improve their abilities in work, study and life. There

are many ways to achieve these objectives, such as:

studying in a university, attending a speciﬁc course,

working in some companies, among others.

On the other hand, there is a particular way, called

social networking service (SNSs) (Boyd and Ellison,

2007), which many people are using to know others

people, to talk about many subjects, to talk with their

friend, coworkers, etc. Because of this, this service

can be a useful tool to help people to know others, in

the some or in different areas, to talk about any situa-

tion, technology, problem, etc., i.e., people can learn

and improve their experience with others through this

service.

According to Hovard (Howard, 2008), the use

of this type of service is very common between

people, for instance, Orkut (www.orkut.com),

LinkedIn (www.linkedin.com), Facebook

(www.facebook.com), Hi5 (www.hi5.com), among

others, have been billions of accesses. There are two

of them with a bigger number of accesses than others

in the world: Facebook and Orkut, mainly in Brazil.

It is a well-known fact that most of people whom

surface on the web have a good chance of partic-

ipating in an on-line social network. Therefore,

these social networks are improving their services

to connect more and more people. For example,

Orkut recommends people to connect with based on

a ”friend-of-a-friend” approach. Facebook and Son-

ico (www.sonico.com) added import wizards, which

allow importing contacts from email and instant mes-

saging clients.

Through this new world that is connecting people,

it is also possible to use it in order to allow people

to learn with each other, i.e., allow people with com-

mon subjects or problems, etc., meet themselves on

the web and talk about these themes.

Some users usually use some services, such as: fo-

rums or chats, to have any contact with others, but

there is a problem. It important to describe that there

is not problem in these services; the problem is in how

people ﬁnd these forums or chats. For example, a per-

son who wants to know more about Enterprise System

can type these words in a search ﬁeld to ﬁnd many

people, forums and chats that like to talk about it, but

it will more difﬁcult to ﬁnd any people that like also

it, but they write as B2B Solution.

In this case, it is necessary to observe that people

with common interests can know or write the some

thing in different ways, i.e., Enterprise System and

B2B Solution are common themes, but the social net-

works do not consider this different to do the search,

in other words, they do not consider people culture.

Depends on culture, country, state, region, among

192

Astolﬁ G., M. A. de Magalhães V., A. R. Silva M. and C. Anacleto J. (2010).

USING CULTURAL DIFFERENCES TO JOIN PEOPLE WITH COMMON INTERESTS OR PROBLEMS IN ENTERPRISE ENVIRONMENT.

In Proceedings of the 12th International Conference on Enterprise Information Systems - Human-Computer Interaction, pages 192-197

DOI: 10.5220/0002975701920197

 SciTePress

others, the people can express themselves in differ-

ent ways, but with the same objective. A person can

search about evaluation and just ﬁnd people that talk

about it, but it is also interesting to ﬁnd people that

talk about test.

In this context, we propose, in this paper, a

methodology that allows people to search and to ﬁnd

people with common interests in different culture, be-

cause this difference can be a good way to allow peo-

ple exchange experience, knowledge, etc. i.e., the cul-

tural difference can be used to join people not to dis-

tance them.

There are some projects which intend to connect

people but they do not consider the cultural differ-

ence, i.e., if a person writes in a different way than

other, they do not know to each other easily. Some

projects are described bellow.

2 RELATED WORKS

There are many research works in literature whose

goals are to identify group of people and recommend

them to social networks formation. These studies

have diverse way, for example, to propose social net-

works to people, to suggest communities to people in

social networks, etc.

Some studies are focused on applications of algo-

rithms designed to identify and join people related to a

speciﬁc professional community, such as researchers.

In (Tang et al., 2008) is proposed the formation of so-

cial networks to participate in a scientiﬁc congress,

considering that the participants have some common

research interests. The authors use data mining tech-

niques to search information about people on the web

and relate to each other, after which participants in-

terfere with this network to reﬁne it excluding people

they do not know.

Kautz et. al. (Kautz et al., 1997b) use data mining

techniques that consider the names of individuals co-

occurring in web pages, in the publication of articles,

on network news, and charts (university departments)

are probably potential training social networks. In ad-

dition, Kautz et. al. (Kautz et al., 1997a) taking into

consideration the same data source to propose a sys-

tem to enhance searching for people by combining

collaborative ﬁltering and social networks. Through

this same strategy, related to data mining, Matsuo et.

al. (Matsuo et al., 2006) developed a system called

POLYPHONET, which to measure the co-occurrence

of names on web sites and, the social networking is

built with the names more co-occurrences. In Tang

et. al. (Tang et al., 2008) also use the web to search

data that can identify the expertise of people to relate

them according to this information.

The authors in (Spertus et al., 2005) also experi-

mented algorithms in order to propose communities

to users on the Orkut. Communities are suggested

from a communities base that certain user is a mem-

ber. The similarity between communities is calculated

overlapping them, e.g., based on the number of users

in each community and the number of common users

between them, so if two communities have a large

number of common users they are considered simi-

lar. Using the same reasoning (Chen et al., 2009b)

evolved two types of algorithms to recommend cus-

tomized communities for users in Orkut.

The work presented in (Chen et al., 2009a) adopts

techniques that have proved effective to recommend

people for social networking. The authors improved

four types of algorithms that were applied on a social

networking site. The most important from our per-

spective is the one that stress the possibility of two

people to post content related to the same subject and,

if it happens, those people have great chance to be

similar; this is measured by comparing a set of words

posted by each user, extracted from their proﬁle’s in-

formation, comments on photos, shared lists, etc.

As can be observed, the studies mentioned here

do not take into account the cultural aspects of people

in searching for similarity to recommend social net-

working. The most is done is to consider some words

or specialty that people share among them. Our work

differs in that point, because we believe that taking

into consideration the culture we can ﬁnd more people

with same interests in different cultures, i.e., country

region, among others.

3 POTENTIAL OF CULTURAL

Our methodology, described here, considers the use

the semantic context to verify similarities among peo-

ple taking into consideration a subject and context.

We assume that if two people have the same consen-

sus related to a certain subject and context, they have

good chances to share the same interests and conse-

quently be similar to each other in that context.

In any country in the world there is a diversity

of vocabularies. Thus, following the reasoning set,

we hypothesize that the user uses in social interaction

(mainly in SNSs) their natural vocabularies when they

need to write a text. For example, in order to exchange

messages. Therefore, taking into consideration this

reality, we believe that there are many users, in SNSs,

with same consensus or interests related to a subject

and a context, but they can express themselves in a

different way. For example, see this both sentences

USING CULTURAL DIFFERENCES TO JOIN PEOPLE WITH COMMON INTERESTS OR PROBLEMS IN

ENTERPRISE ENVIRONMENT

193

from two different users: ”Usability Test validates

Enterprise Systems” and ”Evaluation Usability attests

B2B Solutions”. When we read these both sentence,

we can identify that both represent a consensus re-

lated to a subject and context, i.e., they say something

in different way, but the computer, just trough these

words, can not process these sentences and identity

that they are similar using comparison words (Chen

et al., 2009a). In this context, this paper presents a

methodology that allows computer identify and pro-

cess cultural information in order to identify familiar

sentences wrote in different way by people because

their culture, knowledge, etc. This methodology tak-

ing into consideration a cultural knowledge base.

3.1 Cultural Knowledge Base

The cultural knowledge base is obtained from a

project called Open Mind Common Sense Project

Brazil (OMCS-Br) (Silva and Anacleto, 2009).

OMCS-Br project has been collected culture of

a general public through a web site (through

http://www.sensocomum.ufscar.br). After entering,

the person can register and have access to various ac-

tivities and themes available in this site. Most of the

activities and themes are templates as shown in Fig-

ure 1. For instance, template: Enterprise System can

be called as B2B Solutions.

Figure 1: Template example.

Templates are simple grammatical structures.

They have ﬁx and dynamic parts. Dynamic parts

change when they are presented to users. They are

ﬁlling out with data from other users’ contribution

already registered on the site. Therefore this base

uses the stored knowledge to collect new one. Tem-

plates also have a ﬁeld to be ﬁlled by users consid-

ering their everyday experiences, knowledge and cul-

ture. Finally, the ﬁxed part is strategically deﬁned by

the project coordinators to consider the theory deﬁned

by Marvin Minsky (Minsky, 1988) and it keeps rich

semantic relation (Liu and Singh, 2004) that repre-

sent the common sense of different people. For in-

stance, the template relation ”Enterprise System can

be called as B2B Solutions” (Figure 1) is DeﬁnedAs.

Because the user typed that ”Enterprise System can

be deﬁned as B2B Solutions”. This template is stored

(DeﬁnedAs ’Enterprise System’ ’B2B Solutions’), see

Figure 2.

Figure 2: Simple example of semantic network.

There are many templates to collect cultural infor-

mation; another example is ”Enterprise System IsA

Enterprise software”. The template relation in this

case is IsA. (see Figure 2). Others relations are also

possible, such as: PropertyOf, MotivationOf, Used-

For, CapableOf, etc (Liu and Singh, 2004). These

relations are used to connect the whole information

in the cultural knowledge base. It is important to ob-

serve that there is the cultural knowledge in this base

because people from different cultures, regions, etc.,

type what they know how about a speciﬁc subject,

in these examples about Enterprise System. Finally,

the whole cultural information stored as semantic net-

work we called as Cultural Knowledge Base (see Fig-

ure 2).

Our methodology taking into consideration rela-

tions: DeﬁnedAs and IsA, which represents synony-

mous and specialization respectively:

• IsA(concept1, concept2), where concept1 is a spe-

cialization of concept2;

• DeﬁnedAs(concept1, concept2), where concept1

and concept2 are synonymous.

According to these data we can deﬁned that:

• If concept1 is a specialization of concept2, then

feature (concept1) contains feature (concept2);

• If concept1 is synonymous of concept2, then fea-

ture (concept1) is equal feature (concept2).

In other words, the concepts connect with one of

these relations are similar, i.e., strongly connect.

4 HOW THE METHODOLOGY

WORKS?

First the methodology needs a sentence typed by user.

The user can be typed this sentence in forums, chats,

search ﬁeld, etc. For example: “Usability Test vali-

dates Enterprise Systems” The computer “read” this

sentence and through of a parser, called PALAVRAS

a syntactic parser for Portuguese (Bick, 2000), iden-

tiﬁes the grammar structure, i.e., if it has a subject,

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

194

verb and object. It is possible to observe that in this

example, Figure 3, has a subject (Usability Test), verb

(Validates) and object (Enterprise Systems).

The next step is Normalization (Figure 3) of the

concepts (words), because nouns and adjectives of the

sentence need are in singular and verbs in inﬁnitive

form. Normalization is necessary to increase the po-

tential search in cultural knowledge base, because the

whole concepts in it are normalized.

Figure 3: Normalization by PALAVRAS.

The next step is to create a semantic network with

this sentence. There are in this semantic network the

subject and object connected by verb. For instance,

trough the sentence: subject (Usability Test), verb

(Validate) and object (Enterprise System) was cre-

ated: validate(Usability Test, Enterprise System) (see

Figure 4).

Figure 4: Semantic Network with user’s sentence.

It is important to write that the Normalization al-

lows creating the same semantic network with differ-

ent sentences typed by many users, as Table 1. This

table shows four different sentences typed by four dif-

ferent users and just one semantic network as result.

Other projects as Chen (Chen et al., 2009a), de-

scribed in related works, do not do this process, be-

cause of this, the sentences (Usability Test, validates,

Enterprise System) and (Usability Test, validated, En-

terprise System) will be considered as different sen-

tences.

Our methodology through the Semantic Network

does some searches in the Cultural Knowledge Base

in order to expand it, as Figure 5.

Figure 5: Searching in the Cultural Knowledge Base.

Table 1: Sentences from different users and their Semantic

Network.

Users Sentences Semantic Network

user 1 Usability Test

validates Enter-

prise System

validate(Usability

Test, Enterprise

System)

user 2 Usability Test

validated Enter-

prise System

validate(Usability

Test, Enterprise

System)

user 3 Usability Test

validating En-

terprise System

validate(Usability

Test, Enterprise

System)

user 4 Usability Test

has validated

Enterprise

System

validate(Usability

Test, Enterprise

System)

The main objective of doing this search, in this

base, is to identify others cultural concepts related

with concepts in Semantic Network. For instance, it

is possible to identify what concepts are connected

with Enterprise System, taking into consideration two

Minskys relation: DeﬁnedAs and IsA. In this case, the

results were: B2B Solution, Enterprise software, etc.

Therefore, there are other people that know Enterprise

System, but they call it with different names, such as:

B2B Solution, Enterprise software. This difference

can be inﬂuenced by culture, experience, knowledge,

etc., but, there is this information stored in Cultural

Knowledge Base. There is the some process to iden-

tify cultural concepts related to each concept in Se-

mantic Network, because of that, the next steps are to

search with the concepts: Usability test and Validate,

as show in Figure 6.

The Figure 6 shows the user’s Semantic Network.

It is possible to search in any forums, chats, etc., who

are people talk about the same issue typed by user in

the sentence “Usability Test validates Enterprise Sys-

tems”. Because of that, different users that talk some-

thing with different ways can be identiﬁed. For exam-

ple, trough this methodology, it is possible to observe

that a user that write a sentence with “B2B Solution”,

“attest” and “evaluation usability” can be interested in

knowing other user that write a sentence with “usabil-

ity test”, “validate” and “enterprise system”, because

there are different words but there is a common issue,

as show in Figure 7.

5 EXPERIMENT

We conducted a survey by email in February 22th,

2010 to a group user Twitter (www.twitter.com) in or-

der to observe the use of methodology described in

USING CULTURAL DIFFERENCES TO JOIN PEOPLE WITH COMMON INTERESTS OR PROBLEMS IN

ENTERPRISE ENVIRONMENT

195

Figure 6: Semantic Network with Cultural concepts.

Figure 7: People with common interests but different cul-

tures.

this paper. There were three steps in this experiment:

First, it was sent an email requesting a sentence to

each user, which showed their interest about a certain

issue and context. We guide each one, because they

needed to provide and to type a sentence according

the grammatical structure required by our methodol-

ogy, i.e., subject, verb and object. Second, using the

sentence of the each user and API from Twitter it was

applied our methodology in order to ﬁnd users who

have some interests taking into consideration a certain

issue and context. Third, we sent to each user a new

email with the recommendation and we also asked

their opinion about the results through this email. The

questions were:

• Do you already ”follow” this person? [yes/no];

• Is this a good recommendation? [yes/no].

5.1 Results

There were more thousand “tweets” (user’s post in

the Twitter) recovered of different users of the Twitter

taking into consideration the sentence typed by partic-

ipants of the experiment. Through this amount only

3.36% of users were chosen to be recommended to

participants of the experiment. This is possible be-

cause we ﬁrstly seek “tweets” with the subject and

object of each one (see Figure 6) and after we analyze

the semantic relation between them in order to verify

whether the “tweet” is within the issue and context

desired by the user, i.e., if they have the same inter-

ests.

The Figure 8 shows the relation among the

“tweets” recovered in the search. The part in yel-

low (96.64%) shows the “tweets” discarded, because

words were not related semantically. The red part

(2.87%) shows the “tweets” were found with the se-

mantic relation generated from the sentences pro-

vided by the participants of the experiment. The green

part (0.49%) shows the “tweets” were found with the

aid of the cultural expansion.

Figure 8: Resulted web mining in the Twitter.

These results show that there was low number

of users selected for recommendation. On the other

hand, there were good chances that they are in-

serted into the same issue and context. If we rec-

ommender 100% of users selected without consider-

ing the semantics involved between the words used

in the search, we would do certainly mistaken recom-

mendation and we would not satisfy the user’s inten-

tion, which is to ﬁnd people with similar interests. In

addition, 0.49% of the “tweets” found from the cul-

tural expansion justify the use of our methodology to

ﬁnd people with same interests but different cultural

knowledge, i.e., to search and ﬁnd others people who

express themselves differently. As an example, a user,

of the experiment, typed the sentence “Amazon ﬁlters

the oxygen of the world” and from this sentence was

found another user typed the sentence “Amazon forest

strainers O2 of the world, because of that it is very im-

portant to consider human life and the cultural knowl-

edge... ”, i.e., different in expression, but inserted on

the same issue and context.

Overall, our users rated 88% of the good recom-

mendations; followed by 12% of bad. Considering

these results the methodology proved to be useful in

order to ﬁnd people who express themselves differ-

ently in the same issue and context. Although he has

been experiment found few people with common in-

terest, our methodology was efﬁcient to ﬁnd people

considering the same words, semantically connected,

that people typed, as well as, others concepts obtained

through cultural knowledge, i.e., just the same inter-

ests.

6 CONCLUSIONS

This paper described a methodology in order to iden-

tify users with some interests but different cultures.

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

196

Through this methodology, the users can identify who

are people talking about the some subject with differ-

ent ways, then, they can talk about it and exchange

experience, knowledge, solutions, among others. Be-

cause of that, employees in an enterprise environment

have another opportunity to use social networking ser-

vice to ﬁnd people, with or without the same culture,

to talk, to ask help, etc.

This methodology can be used in many systems,

such as: social networking service, enterprise sys-

tem or any tool that needs to improve the search

mechanism because in this case, the methodology was

used to identify people but the same process can used

to identify educational materials, reports which each

user deﬁned a different name taking into considera-

tion his culture, among others. These results suggest

that if the Social Match Systems considers semantics,

issue, context and culture to search similar people,

they could do recommendation more robust.

ACKNOWLEDGEMENTS

We thank CNPq, FAPESP, EMBRAPA and CAPES

for partial ﬁnancial support to this research. We also

thank all the collaborators of the Open Mind Com-

mon Sense in Brazil Project who have been building

the common sense knowledge base considered in this

research.

REFERENCES

Bick, E. (2000). The Parsing System ”PALAVRAS”: Auto-

matic Grammatical Analysis of Portuguese in a Con-

straint Grammar Framework. PhD thesis, Department

of Linguistics - University of Arhus.

Boyd, D. M. and Ellison, N. (2007). Social network

sites: Deﬁnition, history, and scholarship. Journal of

Computer-Mediated Communication, 13(1).

Chen, J., Geyer, W., Dugan, C., Muller, M., and Guy, I.

(2009a). Make new friends, but keep the old: recom-

mending people on social networking sites. In CHI

’09: Proceedings of the 27th international conference

on Human factors in computing systems, pages 201–

210, New York, NY, USA. ACM.

Chen, W.-Y., Chu, J.-C., Luan, J., Bai, H., Wang, Y., and

Chang, E. Y. (2009b). Collaborative ﬁltering for orkut

communities: discovery of user latent behavior. In

WWW ’09: Proceedings of the 18th international con-

ference on World wide web, pages 681–690, New

York, NY, USA. ACM.

Howard, B. (2008). Analyzing online social networks.

Commun. ACM, 51(11):14–16.

Kautz, H., Selman, B., and Shah, M. (1997a). Referral web:

combining social networks and collaborative ﬁltering.

Commun. ACM, 40(3):63–65.

Kautz, H., Selman, B., and Shah, M. A. (1997b). The hid-

den web. AI Magazine, 18:27–36.

Liu, H. and Singh, P. (2004). Conceptnet — a practical com-

monsense reasoning tool-kit. BT Technology Journal,

22(4):211–226.

Matsuo, Y., Hamasaki, M., Nakamura, Y., Nishimura, T.,

Hasida, K., Takeda, H., Mori, J., Bollegala, D., and

Ishizuka, M. (2006). Spinning multiple social net-

works for semantic web. In AAAI’06: proceedings of

the 21st national conference on Artiﬁcial intelligence,

pages 1381–1386. AAAI Press.

Minsky, M. (1988). The Society of Mind. Simon and Schus-

ter, New York, NY, USA, 2nd edition.

Silva, M. A. R. and Anacleto, J. C. (2009). Enter-

prise Information Systems, volume 24 of Lecture

Notes in Business Information Processing, chapter

Promoting Collaboration through a Culturally Con-

textualized Narrative Game, pages 870–881. Springer

Berlin/Heidelberg, Berlin, Germany.

Spertus, E., Sahami, M., and Buyukkokten, O. (2005).

Evaluating similarity measures: a large-scale study in

the orkut social network. In KDD ’05: Proceedings of

the eleventh ACM SIGKDD international conference

on Knowledge discovery in data mining, pages 678–

684, New York, NY, USA. ACM.

Tang, J., Zhang, J., Yao, L., and Li, J. (2008). Extrac-

tion and mining of an academic social network. In

WWW ’08: Proceeding of the 17th international con-

ference on World Wide Web, pages 1193–1194, New

York, NY, USA. ACM.

USING CULTURAL DIFFERENCES TO JOIN PEOPLE WITH COMMON INTERESTS OR PROBLEMS IN

ENTERPRISE ENVIRONMENT

197