CENTRIC 2014 : The Seventh International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services
User Models and Domain Ontologies for Generating Personalized Questions and
Answers
Marta Gatius
Computer Science Department
Technical University of Catalonia
Barcelona, Spain
e-mail: gatius@lsi.upc.edu
Abstract—The incorporation of user models has become more incorporate diverse information from the user. Most models
popular in recent years. This paper describes how user models, use one of the following features about the user: background
domain ontologies and natural language generation techniques (i.e., profession), knowledge, interest, goals and individual
can be integrated to generate personalized questions and traits (personality and cognitive factors). Information on the
answers related to a particular scenario. For this purpose, a
context of the user's work can also be modeled. Additionally,
general organization of different types of knowledge involved is
proposed, that facilitates its adaptation to different domains, more than one of these features can also be combined.
user models and languages. In this proposal, the domain The use of a specific user's feature is usually related to
conceptual knowledge is represented by a domain ontology, the type of the application that incorporates the model.
that provides a framework for representing the user’s Thus, user's background is used most frequently for content
knowledge and interests. Ontologies provide a powerful adaptation (as in the work on medical domain described in
formalism because they allow the representation of detailed [3]), user's interests are usually incorporated in information
descriptions of concepts, attributes and relations, thus retrieval and recommended systems (such as the dialogue
facilitating richer inferences about user knowledge and system described in [4]), user’s knowledge is useful in
interests. Additionally, in the approach presented, the domain
educational systems, user's goals are incorporated in
concepts in the ontology are also related to a general syntactic-
semantic taxonomy that facilitates the generation of more intelligent interfaces, while individual traits are considered
appropriate questions and answers for each user. The in several commercial systems (such as the system
proposed organization has been applied for generating described in [5]).
personalized questions and answers, that help the user when This paper describes the use of information about the
searching information about university courses. user's backgrounds to generate personalized questions (and
their answers) in a specific domain. Information about the
Keywords-user models; personalized content generation; user profession or experience is considered when classifying
domain ontologies, multilingual language generation. the users in different groups or stereotypes. Specific
questions and answers for each group are generated semi-
I. INTRODUCTION automatically from the domain ontology.
The integration of user models and domain ontologies is Stereotypes have been used for more than 25 years
a relatively new direction of research to personalize because they are simple and can be powerful for several
different types of contents and applications. Although user domains. Stereotypes have been used in domains where the
models have been used since the 80's, during the last years distinction of two or three user groups is easy and useful,
the interest in adapting the contents to the different types of such as the medical domain, where users can be classified as
users is increasing in many areas (as described in [1] and professionals (and subclassified as doctors, nurses and
[2]). The huge amount of contents and systems in the web, students) and patients (and subclassified considering
where user population is diverse, have also increased the different knowledge levels). Information necessary to
interest on the incorporation of users models in different classify the user is provided explicitly, usually by the user.
types of systems, such as web communication systems, There are already several research works that integrate
intelligent web systems, information extraction, e-learning user models and domain ontologies [6][7][8]. Ontologies
and e-health. provide a framework for representing the user knowledge (or
User's models can be used for personalized content interest) about different domain concept, or about the
generation and can also be incorporated in adaptive systems, attributes describing the concept. Ontologies provides a more
that behave different for different users. User's models can powerful formalism than other forms of semantic
Copyright (c) IARIA, 2014. ISBN: 978-1-61208-369-8 39
, CENTRIC 2014 : The Seventh International Conference on Advances in Human-oriented and Personalized Mechanisms, Technologies, and Services
representation, used in previous work on user adaptation different domains. Furthermore, they are especially
(such as concept networks), because they allow the appropriate for representing conceptual knowledge in
representation of detailed descriptions of concepts, attributes multilingual systems, because they provide a common
and relations, thus facilitating richer inferences about user semantic representation for the several languages supported.
knowledge and interests. Recently, domain ontologies have also been used for
The approach described in this paper differs from other personalizing presentation in several communication
works, such as the above mentioned, in that it is focused on systems (such as [12] and [13]). Ontologies favor the
the optimization of the generation of questions and answers generation of personalized content because they provide a
in several languages. For this purpose, it uses a general framework for representing user’s domain knowledge (or
syntactic-semantic taxonomy (described in [9]), that interest). In more simple semantic representations the user’s
establishes general relations between the ontology concepts knowledge about a domain concept is represented as a
and the linguistic structures supporting questions and binary value (known-not known) or as a weighted value.
answers about them. Ontologies allow the representation of more detailed
The integration of user models and language generation information on the user’s knowledge about a specific
techniques has also focused several research works, such as concept, such as that related to the different attributes
the relevant information-systems Ilexand and PEBA-II describing the concept and the relations with other concepts.
(described in [10]), that dynamically generates pages Additionally, they also support inference and other
according to the user profiles. However, those works used reasoning mechanisms.
other language generation techniques, not based on This section describes how domain ontologies are used
ontologies. to generate personalized questions and answers for several
The organization of the conceptual and linguistic user’s stereotypes. In a particular domain, groups of users or
knowledge that is proposed in this paper was already stereotypes are distinguished because they have different
incorporated in an adaptive dialogue system (described in interests and/or knowledge. This information is represented
[11]), where information on how well the communication is as an overlay of the domain concepts in the ontology.
doing is used by the system to select the most appropriate The approach presented has been applied to the scenario
respond to a user's intervention: a directed (explicit) question in which the user searches for specific information about a
or a more open message. The approach presented in this specific university course. In this particular scenario, the
paper incorporates a user model that uses the domain
representation of the domain concepts and the interests of
ontology to represent the user background related to the
domain concepts. the different type of users is simple, as shown in Figure 1.
In the approach presented, the process to generate the There are only two domain concepts involved and two
personalized questions and answers consists of four steps: different types of users: teachers and students. The domain
The study of the different types of users, the representation concepts are described by a set of attributes, and user
of the domain concepts, the relation of the attributes domain interests are represented as a binary value associated
describing concepts to the syntactic-semantic taxonomy and, with each of the conceptual attributes. When generating
finally, the generation of questions and answers. personalized questions, only the conceptual attributes
The work described is being done in the framework of interesting for each specific user type are considered.
This approach uses of a syntactic-semantic taxonomy that
the Spanish project SKATER, for Scenario Knowledge
defines general relations between conceptual attributes and
Acquisition by Textual Reading. The personalized questions
linguistic knowledge to facilitate the generation of the
and answers are generated from the ontology describing a questions for different domains and languages.
particular scenario, previously built from text. The process of generating personalized questions and
The next section describes how the proposed approach answers for a particular scenario consists of the four steps:
has been applied for a particular scenario, in which the user
is seeking specific information about a particular university Study of the needs of the different types of users.
course. Then, the last section describes the conclusion and
Description of the domain concepts involved in
future work.
the scenario. Each concept is described by a set
II. GENERATING PERSONALIZED QUESTIONS AND of attributes and those attributes are associated
ANSWERS with the specific types of users interested on
their values.
As mentioned in the introduction, several systems Relating conceptual and linguistic knowledge.
incorporate user models and ontologies to adapt contents Generation of the questions and answers from
and presentations to different types of users. For example, in the conceptual attributes for each type of user.
several scenarios, the generation of personalized questions,
dialogues and summaries can help the adaptation of the web The next subsections describe the adaptation of these
contents and systems to the user’s needs. four general steps to generate the personalized questions and
Domain ontologies have been incorporated in many answers about university courses.
different types of applications because they provide a great
flexibility to represent knowledge and are easy to adapt to
Copyright (c) IARIA, 2014. ISBN: 978-1-61208-369-8 40
