THE DESIGN OF A SOCIAL SEMANTIC SEARCH ENGINE
Preserving Archived Collaborative Engineering Knowledge
with Ontology Matching
Jörg Brunsmann
Faculty of Mathematics and Computer Science, Distance University of Hagen, D-58097 Hagen, Germany
Keywords: Semantic Web, Social Search, Semantic Search, Natural Language Processing, Ontology Matching,
Onotoloy Alignment, Linked Data, Provenance, Digital Archives, Knowledge Management, Resource
Description Framework, Ontologies.
Abstract: Private and business related knowledge acquisition is either performed via learning by doing or via human
dialogue that includes transmission of social or collaborative questions and answers. Unfortunately it can be
a time consuming task to find a trusted friend on the web for private recommendations or to find a qualified
expert colleague in a (virtual) organisation for work-related questions or to find a suitable company contact
person as a customer. Recently, such social question and answering is conducted with internet based
technologies like social search engines which route a question to a appropriate human selected from a social
or expert network. However, even if social search engines are involved, it is unlikely that existing social
search approaches exploit machine-readable lightweight ontologies that enable classifying, publishing and
sharing questions and answers to support subsequent semantic search without human involvement. This
paper proposes the combination of semantic web and social search technologies in order to publish and
archive social and collaborative generated knowledge for future reuse. Since knowledge classifying
vocabularies evolve over time the paper also describes why archived knowledge may become obsolete and
how ontology matching methods are used to migrate knowledge to conform to contemporary vocabularies.
1 INTRODUCTION
Probably without being aware of it, at some point
everyone has been in touch with social search
knowledge. Posting or answering a question in an
internet forum, asking a colleague via phone in daily
job activities, searching for responses to technical
problems using a web search engine, asking a
company agent as a customer for contract related
help, writing a product review for a e-commerce
web site or asking a friend about his private opinion
via microblogging web sites, mobile phone or
instant messenger are all valid examples of social
search. In all these examples new explicit
knowledge is created because one person asks or
searches for knowledge from another person.
Because processing of natural language
questions is not yet fully supported by traditional
search engines, social search (Narasimhan, 2010)
enables users to write down questions in natural
language and let other users in their social or expert
network answer the question. Selecting the user who
is most competent to answer a question is based on
social rank which reflects reputation and
connectivity and other metrics (Hangal, 2010).
While page rank selects a document based on
authority, social rank selects a person based on
intimacy (Horowitz, 2010) and trust (Morris, 2010).
During the search workflow (Evans, 2009), users
try first to search on their own and use their social
network only if the intial search was not successfull.
To support this workflow, previosuly conducted
questions and answers must be annotated with
vocabularies based on the Ressource Description
Framework (RDF). Such annotated documents are
then published and archived for future reuse.
If metadata that represents contextualized social
search knowledge is archived, it is necessary to
maintain the annotated knowledge for future search
and access. However, RDF based ontologies evolve
and archived annotated knowledge must be migrated
by processing ontology alignments.
The remainder of the paper is structured as
follows. The next section provides a characterization
of engineering knowledge. Section 3 describes
200
Brunsmann J..
THE DESIGN OF A SOCIAL SEMANTIC SEARCH ENGINE - Preserving Archived Collaborative Engineering Knowledge with Ontology Matching .
DOI: 10.5220/0003070402000205
In Proceedings of the International Conference on Knowledge Engineering and Ontology Development (KEOD-2010), pages 200-205
ISBN: 978-989-8425-29-4
Copyright
c
 2010 SCITEPRESS (Science and Technology Publications, Lda.)

semantic web knowledge representation
technologies. Section 4 elaborates on classifications
of social search and section 5 proposes the
combination of social and semantic search
technologies. The last section concludes with a
description of future work.
2 KNOWLEDGE
Knowledge that is based on question and answer is
articulated into a language and then transmitted and
communicated to others. Knowledge as concept is
formalized in the Data-Information-Knowledge-
Wisdom (DIKW) model (Fricke, 2009). In the
DIKW model, the data layer consists of raw
elements whereas information provides declarative
answers to who, what, where and when questions.
Finally, knowledge provides answers to how
(procedural) and why (causal) questions.
Knowledge acquisition processes by individuals
and groups in enterprises are described by the
famous knowledge spiral (Nonaka, 1995) as a
conceptual foundation for enterprise knowledge
management. Explicit knowledge is capable of being
stored in machines whereas tacit knowledge is in
person’s heads and is very difficult to be represented
in machines. Existing knowledge is internalized to
create new tacit knowledge which is socialised
afterwards, then externalised and so on. This process
builds the knowledge spiral.
Mapping this spiral to social search activities, we
find that during socialisation knowledge acquisition
is done by verbal questions and answers. Annotating
and publishing these questions and answers pairs is
externalisation. Knowledge acquisition by searching
published questions and answers is internalisation.
Finally and especially, knowledge combination is
performed during enterprise collaborations in the
engineering industry.
2.1 Engineering Industry Knowledge
The SHAMAN digital preservation project
(SHAMAN, 2009) investigates the knowledge
preservation of different domains including the
industrial design and engineering industry
(Brunsmann, 2009). This industry use tools
organized by product life cycle management (PLM)
systems (SHAMAN, 2008) and strongly depend on
heterogeneous knowledge resources like employees,
processes, documents, databases (Kamara, 2002).
Use cases for social engineering search knowledge
include:
During collaborative innovation processes an
idea is converted it into a sellable product by
performing collaborative brainstorming sessions or
by interviewing customers. The idea needs to be
educated to colleagues, business partners and
customers so that the partners can contribute their
own ideas. Therefore it is necessary, that during
brainstorming sessions and subsequent collaboration
sessions questions and answers are recorded.
During domain and enterprise collaboration, co-
operations between different enterprises (virtual
organisations) and different engineering domains are
formed. Such cross-enterprise and cross-domain
collaborations exploit the specific knowledge area of
each cooperation partner.
These two use-cases show that tacit “know-how”
and “know-why” engineering knowledge is
exploited during social search activities. For future
enterprise benefit it can be made explicit, if it is
expressed with machine-readable semantic web
technologies which enable archiving and reuse of
knowledge.
3 SEMANTIC SEARCH
The semantic web allows to reason over knowledge
which is modelled as sets of assertions. In recent
years the focus of the semantic web switched to
publishing, integrating and retrieving linked data
following the principle that web resources are
identified with interlinked resolvable HTTP URIs.
Linked data is modelled with lightweight
vocabularies (ontologies) like SKOS, SIOC, FOAF,
Dublin Core, vCard (YAHOO, 2010) expressed by
the Resource Description Framework (RDF).
RDF triples can also be integrated into existing
HTML pages by using RDFa, eRDF or Microformat.
Search engine crawlers extract and store relevant
RDF triples. However, semantic content that has
been crawled is useless if it cannot be searched and
accessed. Current search engines are keyword based
and do not reduce the communication gap between a
human and a computer.
Semantic web technologies promise that search
engines will be able to answer natural language user
queries. Current approaches either apply natural
language processing to unstructured text or they
assume the existence of structured statements over
which they can reason. For example, (Lopez, 2007)
describes an ontology-driven question answering
system that takes an ontology and a natural language
query as input and returns answers from a triple
store.
THE DESIGN OF A SOCIAL SEMANTIC SEARCH ENGINE - Preserving Archived Collaborative Engineering
Knowledge with Ontology Matching
201

Not only on the internet but also in enterprises,
ontologies make tacit knowledge explicit. A shared
and common meaning is modelled with ontology
classes and properties which were formerly tacit in
the head of employees so that they may be
understood by other employees and partners.
Unfortunately, knowledge described with
ontologies face the threat of syntactic and semantic
heterogenity: the conceptualization of a domain
varies from different author viewpoints, has different
terminology (e.g. synonymy), overlap with other
ontologies, cover different portions of a domain and
can be represented by different ontology languages
(e.g. RDF or Topic Maps).
In addition to such syntactic and semantic
heterogeneitiy, it is very common that a real world
domain is continually changing so that the ontology
evolves as well. In order to keep the triples
interpretable, ontology alignments can be used
which are produced by ontology matching.
Ontology matching is the process of relating two
ontologies sharing one domain or two versions of
one ontology. These ontology alignments can then
be used to migrate ontology instances and to
integrate different ontologies of the same domain.
4 SOCIAL SEARCH
While semantic search uses the contextual meaning
of keywords to improve search results, social search
describes the process of incorporating content
generated by individuals and the individuals
themselves into the generation of search results
(Smyth, 2009). In contrast to social search, the term
people search denotes the process of finding
information about individuals across internet
documents. The following sections give an overview
of existing social search classification and lay out a
use cased based social search classification.
4.1 Social Search Classification
Social search is the process of finding information
online with the assistance of social resources. While
searching for knowledge via keywords in a search
engine, also results from the social network are
presented, because it is likely that friends are more
trusted than documents from the internet (Morris,
2010). Collaborative search is a social search where
one or more individuals share an information need
and work together to fulfill that need.
According to (Narasimhan, 2010) social search
processes improve search results by
 Machine-based passive search using social
media (user generated content). The social
graph is used to filter trusted content and
contextualize the response.
 Human-based active search using social graph
(user interactions). The social graph is used to
socialize the query and to classify user
expertise which enables to route to incentivize
participation.
Figure 1: Social search according to (Narasimhan, 2010).
Figure 1 shows a social search system according
to (Narasimhan, 2010) which can be regarded as:
 As sensor: the crowd as collective and real-time
intelligence collector.
 As filter: trusted social relationships are used to
extend search results.
 As router: social relationships are used to
forward queries to a person with relevant
expertise.
(Evans, 2009) defines three different types of
social search.
 Collective social search is capturing real-time
network trends or group wisdom.
 Friend-filtered social search is using the social
network data exclusive or alongside traditional
search results.
 Collaborative search (question answering) is
when two or more users work together to find
the answer to a problem.
4.2 Use Case based Classification
All existing classifications that were described
above regard the term social as a network of friends.
However, the full potential of social does not end at
the list of friend for private reasons. It also includes
KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development
202

colleagues, experts, agents and other actors. That
means, social relationships not only exists in the
private realm but also within enterprises or virtual
organisations between colleagues. And social
relationships exist in customer relationship
environments (e.g. consumer to company agent).
This will get evident if one takes a closer look at
the types and topics of questions and the motivations
for questioning and answering that were described in
(Morris, 2010). Table 1 maps question type and
topic to use case.
Table 1: Type, topic and use case of questions.
Type Topic Use case
Recommendation Technology,
Restaurant
Private
Opinion Ethics Private
Invitation Family Private
Offer Shopping Business
Factual knowledge Technology,
Contract
Personal
Factual knowledge Professional Enterprise
Based on table 1, one can identify four different
use cases:
Private social search often asks for
recommendation and opinions. The motivation for
the questioner is trust or failed search whereas the
motivation for the answerer includes altruism, free
time, to connect socially or ego motivations. The
interaction can be regarded as “Friend2Friend”.
Personal social search include acquisition of
factual knowledge for problems solving. A customer
might have a question regarding the products that
are distributed by a company. The motivation for the
questioner is answer speed and quality and the
motivation for the answerer might be customer
satisfaction. The type of interaction is
“Customer2CompanyAgent”.
Business social search enables a marketing unit
of a company to contact a customer. E.g., the
customer has extended his contract and the company
wants the customer to recommend the product to his
social network. The marketing unit can use the
customers’ social network. The motivations for the
questioner are business opportunities and the
motivations for the answerer might incentives. The
interaction type is “Company2Customer”.
Enterprise social search means to search for a
work related answer which includes contacting a
expert colleague. The enterprise collaboration and
innovation process are examples of enterprise social
search which might involve more than two persons
and thus can be regarded as collaborative search.
The motivation for the questioner is answer speed
and quality and the motivation for the answerer
might be expertise, ego and incentives. The
interaction type is “Colleague2Colleague”.
Making the social and collaborative search
knowledge explicit for others by attaching semantics
is important for future knowledge reuse.
5 SOCIAL SEMANTIC SEARCH
(Evans et al, 2009) investigated people's search
processes and preferences and found that they want
to attempt to search on their own first or do not wish
to interrupt their colleagues before they have tried to
search on their own independently. Later, if the
searcher did not find a satisfactory answer to a
problem, they often turn to a colleague for help.
Therefore, early social support should be passive.
Figure 2: Social and semantic search.
Social search aims to find a human to answer a
question whereas semantic search tries to find
relevant documents that conforms to keywords or
natural language queries. Combining these two
complementary search strategies will provide a
growing and real-time collective social semantic
knowledge system that enables human/human
interaction and human/machine interaction. Figure 2
shows a schematic view of such a system.
The system has access to semantically annotated
documents and the user’s social network. The social
semantic knowledge system does not force user to
generate documents, it rather captures question-
answering interaction, contextualize the social
search results, publishes and archives the previously
conducted questions and answers pairs.
THE DESIGN OF A SOCIAL SEMANTIC SEARCH ENGINE - Preserving Archived Collaborative Engineering
Knowledge with Ontology Matching
203

Answering a question by verbal communication
can be regarded as socialisation of knowledge. If the
answering process is performed by social search
processes it involves written words and thus can be
seen as externalisation because it made available as
document annotated with metadata. Externalized
knowledge needs to be contextualized and attached
with metadata and must be published. Such
published documents can be found via semantic
search and internalised by individuals.
5.1 Social Semantic Search Discussion
This section discusses requirements, advantages and
disadvantages for social semantic search.
5.1.1 Requirements
The network of individuals has to be big enough in
order to make social search effective. If only few
individuals exists in the social network, it is likely
that some questions remain unanswered which
lowers the acceptance of the social search engine.
Experts must be proven and competent in their
domain to avoid false answers. After receiving the
answer the questioner must rate the answer in order
to document the answer and answerer quality. In an
enterprise, the answerer and questioner must be kept
anonym to enable dispassionate ratings and prevent
unmotivated answers.
Whereas the questioner has an immediate and
present need or interest, the participation of the
answerer has to be incentivized, so that also the
answerer has a satisfactory experience.
Social search must be symmetric, e.g. a company
must be able to submit an offer to a customer and a
customer must be able to pose a question to a
company.
The social semantic search system should
identify if a user has difficulties in searching without
human help. For example, if a user already searches
for 10 minutes, an appropriate domain-specific
expert could be suggested to chat with.
From a list of (anonym) individuals one should
be able to select the answerer based on some criteria.
In other situations the routing method should select
the answerer on its own. Therefore, the social search
system should support three different
communication methods: route the question and
answer synchron (e.g. select a user in instant
messenger), semi-synchron (routing algorithm
selects an appropriate answerer) or asynchron (like
Yahoo answers).
5.1.2 Advantages
The social semantic search system increases the
answer speed and reduces spam search results as it
helps to generate answers from a trusted network of
friends or experts so that answers have more
relevance to the questioner.
Social search enables interaction via dialogue
which helps the process of understanding and fosters
the generation of implicit and explicit knowledge.
Finding the right answer is faster compared to
traditional search. In addition, the enterprise
knowledge base gets better and is kept up to date as
more people participate.
Semantic search capabilities increase the
probability to find satisfactory results and thus
reduce the probability that human involvement is
needed. Finally, a company can improve the
customer satisfaction by providing a real-time
knowledge base to the customer.
5.1.3 Disadvantages
The questioner needs to trust the social ranking
algorithm as probably non-experts will answer
questions. In addition, blind trust can be misleading,
since the answer of a close friend can still be wrong.
Additionally, since the system is based on human
contributions it is dependent on the input of the users
and if the user base is small, it may not reach full
acceptance.
Since the world is changing fast, experts need to
keep up to data with knowledge explosion which is
neither an easy nor cheap task. In addition, the
answerer is interrupted in his normal work activities
and receives incentives. All these aspects have to be
evaluated by comparing costs and benefits of a
social semantic search engine.
In enterprise social search the participation will
decline if incentives are low. And finally, on the
internet sooner or later spam will reach social
search, which definitely will reduce the reputation of
the social search engine.
6 CONCLUSIONS AND
OUTLOOK
This paper described how knowledge is acquired by
social and semantic questions and answers on the
web, in the enterprise and in customer relationship
management affairs. It also proposes to support the
intuitive search workflow that first includes
searching via machine and then involves a human
KEOD 2010 - International Conference on Knowledge Engineering and Ontology Development
204

from the social network. The initial non-human
search is improved by semantic web technologies.
The paper described social and collaborative use
case scenarios in the engineering industry and
elaborated on how to archive semantically annotated
question and answers pairs for future reuse. Such
archived engineering knowledge is exposed to
threats like syntactic and semantic heterogeneity
which could result in semantic obsolescence.
Fortunately, ontology mappings help to overcome
such issues. The contributions of this paper include:
 Combination of semantic web and social search
technologies.
 Extension of social search for customer
relationship management purposes and
enterprise collaborations.
 Publishing and archiving of RDF annotated
questions and answers pairs.
 Usage of ontology matching in archiving of
RDF based question and answer knowledge.
Further investigations include a wide variety of
research topics:
 Further evaluation of existing social search
approaches and systems.
 Types of communication and dialogue
workflow in private and business scenarios.
 User interface design for different usage
scenarios (private, business).
 Evaluation of collaboration patterns (Pattberg,
2007) for usage in social search.
 Explore other social search use case scenarios
(e.g. collaborative ontology engineering).
 Exploiting social network analysis metrics for
social rank calculation.
 Detailed capturing of search workflow.
 Exploration of objective rating methods.
 Exploring incentive possibilities (real and
virtual currencies).
 Description of multilingual problems of social
semantic search.
 Evaluation of question analyzing methods.
 Appropriate ontology matchings technologies
for evolving RDF vocabularies.
 Evaluation of enterprise social search costs and
benefits.
ACKNOWLEDGEMENTS
This paper is supported by the European Union in
the 7th Framework within the IP SHAMAN.
REFERENCES
Brunsmann, J., Wilkes W., 2009. Enabling product design
reuse by long-term preservation of engineering
knowledge. International Journal of Digital Curation,
Vol 4, No 3, 17–28.
Evans B. M., 2009. 3 Flavors of Social Search: What to
Expect.http://www.readwriteweb.com/archives/3_flav
ors_of_social_search_what_to_expect.php
Evans B. M., Kairam S., Pirolli P., 2009. Do Your Friends
Make You Smarter? An Analysis of Social Strategies
In Online Information Seeking. Information
Processing and Management.
Fricke, M., 2009. The knowledge pyramid: A critique of
the dikw hierarchy. Journal of Information Science,
35(2), April 2009, pp.131–142.
Hangal S., MacLean D., Lam, M. S., Heer J., 2010. All
Friends are not Equal: Using Weights in Social Graphs
to Improve Search. Proceedings of the SIGKDD 4th
Workshop on Social Network Mining and Analysis.
Horowitz D., Kamvar, S. D., 2010. The Anatomy of a
Large Scale Social Search Engine, WWW 2010
conference.
Kamara J. M., Augenbroe G., Anumba C. J. and Carrillo
P. M., 2002. Knowledge management in the
architecture, engineering and construction industry,
Journal of Construction Innovation, Vol. 2, 53-67.
Lopez, V., Motta, E., Uren, V. and Pasin, M., 2007.
AquaLog: An ontology-driven Question Answering
System for organizational Semantic intranets, Journal
of Web Semantics, 5, 2, pp. 72-105, Elsevier.
Morris, M. R., Teevan, J. Panovich, K., 2010. What do
people ask their social networks, and why? A survey
study of status message Q&A behavior. In
Proceedings of CHI.
Narasimhan, N., 2010. The Evolution of Social Search,
http://nityan.wordpress.com/2010/03/30/the-evolution-
of-social-search/
Nonaka, I. and Takeuchi, H., 1995. The Knowledge-
Creating Company, Oxford University Press.
Pattberg, J.; Fluegge, M., 2007. Towards an Ontology of
Collaboration Patterns. In (Adam Pawlak et. al. Eds.):
Proceedings of 5th International Workshop on
Challenges in Collaborative Engineering (CCE’07),
pp.85-95, GI-Edition – Lecture Notes in Informatics.
SHAMAN Project, 2009. SHAMAN Homepage.
http://shaman-ip.eu/
Smyth, B., Briggs, P., Coyle, M., O’Mahony, M., 2009.
Google Shared. A Case-Study in Social Search.
Proceedings of UMAP.
YAHOO, 2010. SearchMonkey documentation, http://
developer.yahoo.com/searchmonkey/smguide/profile_
vocab.html
THE DESIGN OF A SOCIAL SEMANTIC SEARCH ENGINE - Preserving Archived Collaborative Engineering
Knowledge with Ontology Matching
205