Semantic Integration between Context-awareness and Domain Data

to Bring Personalized Queries to Legacy Relational Databases

Vinícius Maran

1,4

, Alencar Machado

, Iara Augustin

and José Palazzo M. de Oliveira

Coordination Office, Federal University of Santa Maria, Av. Presidente Vargas, 1958, Cachoeira do Sul, Brazil

Polytechnic School, Federal University of Santa Maria, Santa Maria, Brazil

Center of Technology, Federal University of Santa Maria, Santa Maria, Brazil

Institute of Informatics, Federal University of Rio Grande do Sul, Porto Alegre, Brazil

Keywords: Database, Context-awareness, Ubiquitous Computing, Query, Semantic Web, Ontology.

Abstract: Context-awareness is a key feature in ubiquitous middleware. Mainly, it is applied to adapt services and

interfaces of applications that use ubiquitous features. The application of context information to personalize

data queries is a recent topic in computing and still presents a large number of challenges. One of the main

gaps evidenced by this research field is the lack of integration between context information, which is designed

and used by ubiquitous middleware, and domain data, which are frequently persisted in relational databases.

This integration is necessary because context can be used as a filter for content query. This position paper

presents a motivational scenario that clarifies the necessity of the integration between context, used by

ubiquitous middleware, and relational data, a comparison between the state of the art of the field, a list of

research opportunities in the field, and a proposal of a framework that uses ontologies to integrate context and

domain data, modeled and stored in relational databases.

1 INTRODUCTION

Mark Weiser (1991) presented a series of scenarios

where computing acts to help users in their daily

tasks, without even they are able to notice the use of

computers in these tasks. These scenarios originated

the ubiquitous computing research field. Recently,

ubiquitous computing area involves a set of

technologies and methodologies of implementation.

One of the key concepts used in ubiquitous

computing is called context-awareness. In recent

definitions (Makris et al., 2013) (Perera et al., 2014),

context information is defined as a measured and

inferred data about current state of entities present in

the environment. This information can be used in

systems for adaptation in query of content and

execution of services. Context is modeled and used in

many forms by ubiquitous systems. Recent surveys

(Bettini et al., 2010) (Strang; Popien, 2004) show that

the representation of context based on ontologies

presents a series of benefits over other forms of

representation, like the existence of patterns to define

ontologies and high expressiveness of them.

Ontology can be defined as a formal and explicit

specification of a shared conceptualization (Borst,

1997) and it can be represented in several languages.

These languages are classified as: (i) based on logic, or

(ii) serialized in XML languages. The second group is

the most currently used because there are standards for

representation, managed by W3C (2016).

Context information can also be applied in many

forms in ubiquitous systems. The most frequent usage

of context are (Dey et al., 2001): (i) dynamic

adaptation of services, (ii) personalization of user

interfaces, (iii) search of resources, and (iv) content

and data querying. Content and data, related to the

domain of the application, must be queried in a

personalized form in ubiquitous systems, mainly

using context information to filter it. This feature

differs from the process of querying data in traditional

systems in two ways. First, ubiquitous systems must

query heterogeneous sources of data, including

legacy relational databases. Second, the filtering

information that is used in the query is not informed

by the user explicitly, because the context

information is automatically collected and inferred by

ubiquitous middleware and then it is used in the query

(Maran et al., 2015a). So for these domain data,

persisted in relational databases, to be recovered in a

contextualized manner, it is necessary to create forms

of interconnection between context information,

which are often represented in ontologies and domain

238

Maran, V., Machado, A., Augustin, I. and Oliveira, J.

Semantic Integration between Context-awareness and Domain Data to Bring Personalized Queries to Legacy Relational Databases.

In Proceedings of the 18th International Conference on Enterprise Information Systems (ICEIS 2016) - Volume 1, pages 238-243

ISBN: 978-989-758-187-8

information, persisted in relational databases

(Bolchini et al., 2013). This paper presents a

motivational scenario of the data access based on

context, an overview of the state of the art in the data

access based on context research area, a list of

research opportunities in the area, and a proposal of a

framework to link legacy relational databases and

context models, used in ubiquitous middleware.

The paper is organized as follows: In Section 2 a

motivating scenario and the main concepts related

areas of context-awareness, ontologies, and

contextualized data querying are presented. In

Section 3 a qualitative comparison between recent

research and a list of research opportunities are

presented. In Section 4 a framework to integrate

legacy relational databases and context models based

on ontologies is presented. In Section 5 the

conclusions and future work are presented.

2 WHY UBIQUITOUS

COMPUTING NEEDS TO GET

CLOSER TO RDBMS?

Relational Database Management Systems

(RDBMSs) traditionally support SQL queries. For a

system make a query in the database, it is necessary

that this system and the user inform in an explicit

form the terms and conditions that will filter the

results over the relations. This works well with

traditional system and database design process,

mainly because the database schema is designed to fit

the application and domain information. But with

recent advances in systems and the large increase of

the information size in databases, some well-known

problems became relevant about queries in RDBMSs

field: (i) To perform a query is required to system and

user to inform filtering criteria (the selection and

projection criteria). A known problem is that often a

few users report, or do not report filtering criteria. So

the amount of information retrieved is large and users

need to filter the information manually (Bolchini et

al., 2013); (ii) The same instance of the database must

be accessed by multiple systems, modelled in

different manner, although of being designed for the

same domain. This fact induces a known problem

about semantics of the data, frequently supported by

the use of ontologies (Dey et al., 2001).

Ubiquitous middleware frequently use ontologies

to describe context models (Bettini et al., 2010). This

way, applications that use ubiquitous features must

use these context models to perform actions in a

personalized form. These context information,

previously measured and inferred by ubiquitous

middleware, can be informed to the application or the

database at the moment of the query (Perera et al.,

2014). However, context information is not informed

all in the same way every time. For example, in a

specific moment, a body sensor can send some

important information, but in other moment, this

sensor cannot be available, and the context associated

to this sensor cannot be used (Makris et al., 2013).

This problem is related to the two previously

mentioned problems in RDBMS queries (i and ii),

mainly because the context information is not

informed in a complete and in the same form in every

query, and because the context is frequently modelled

in ontologies. Furthermore, the design process of a

relational database and context modeling occur in

isolation from each other.

2.1 Motivational Scenario

Ubiquitous middleware have been designed and

applied in several fields. Figure 1 presents an

overview of the motivational scenario, described

below. Recent researches (Stavropoulos et al., 2013)

(Machado et al., 2014) describe smart university

environments, which have a wide variety of

educational resources that are managed by a

ubiquitous middleware. Some of these resources can

be recommended to students, which study in these

universities (Stavropoulos et al., 2013).

Figure 1: Motivational Scenario.

Let us imagine that a university campus (a) uses

an university administration system (b), and a content

management system (c), which allows teachers to

provide supplementary materials, exercises and

activities for students, which in turn access the system

to get the materials. These systems store and retrieve

information from a relational database (d), using SQL

Semantic Integration between Context-awareness and Domain Data to Bring Personalized Queries to Legacy Relational Databases

239

queries. In a determined moment, the university starts

to use a ubiquitous middleware (e) with an adaptive

Virtual Learning System (VLE) (Maran et al., 2015)

(f) to show content and resources of the university to

students. To adapt its execution and content to the

user, the middleware manages the context

information, captured from the environment (g). To

represent the context, the middleware uses an

ontology, which represents the context after it being

measured and inferred, and stores and retrieves this

context from a database (h).

Peter is a student (i) in an Electrical Engineering

course in this smart campus. Currently, Peter is on the

2nd semester of the course and is attending the

AL101EE discipline, called Algorithms and Data

Structures 1. In a particular class, the context

management middleware informs the context

information about the user (1), about the educational

context (2), information about location of the user (3),

information about the device used by the user, and

temporal and activity context information (5) to the

adaptive VLE:

Context_User = {Peter, Student, 2ndSem, EEngineering}

(1)

Context_Educational = {Peter,AL101EE,Class4,CBranch}

(2)

Context_Location = {Country, City, SmartUniv}

(3)

Context_Device = {AndroidBasedPhone}

(4)

Context_Activity = {in_class, Init_time, End_time}

(5)

This context information is modelled in the

ontology, and it is used by the ubiquitous middleware.

Currently, Massive Open Online Courses (MOOC)

are offered in VLE (f) to an audience of the university

as an opportunity to expand their knowledge. In a

class of Algorithms and Data Structures 1, the teacher

introduced the concept of Conditional Branch. At the

end of the lesson, the recommender system showed to

Peter in his smartphone the recommendation of the

MOOC about algorithms. Peter is interested in more

information about this topic, as recorded in MOOC.

By the time, the student enrols in MOOC, the

ubiquitous middleware transfers context information

(with the permission of the student) to the VLE

system, and this information is used by the MOOC.

Thus, profile and devices information associated with

the student can be used by MOOC to filter course

information to the student. The MOOC about

Algorithms that Peter signed up presents algorithms

and questions to assess student understanding on the

basic algorithm structures applied to the student’s

course. Currently, the completion rate MOOCs varies

between 5 and 10% (Pretz, 2014). Recent work

(Pretz, 2014) (Quinn et al., 2014) attributed as a cause

for this low completion rate the fact that the

information related to courses are presented in the

same way to all students, regardless of context

information and information related to the student's

profile. Originally, the MOOC platform was designed

to use a relational database, with a pre-defined

structure (EDX, 2015). As the context model is

represented in ontologies and is managed by the

ubiquitous middleware, it can not be informed

explicitly at query time. Thus, there is a need to use

up a framework (j) that implements a model that

allows that context information used by the

ubiquitous middleware to be used in information

filtering. This scenario presents two relevant features

regarding to recent research related to context:

• There is an exchange of context information

between the middleware that manages context

information on university campus and an external

system (represented by the VLE). This is a research

problem addressed in recent work (Makris et alk.,

2013) (Perera et al., 2014);

• There is a gap for binding between context

information and field data regarding persistence and

recovery of data using context information in the

query process. This gap is directly related to two

characteristics: (i) information relating to the scope

are usually modeled relationally and persisted in

relational databases. This is evidenced by the wide

adoption of RDBMS (DBEngines, 2016) and (ii)

information related to context used by ubiquitous

systems are often modeled on ontologies (Bettini et

al., 2010). In recent work related to ubiquitous

middleware, application independent ontologies were

used as the basis for context modeling.

2.2 Context-awareness and Ontologies

Context is a broad term and has a set of definitions,

according to specific fields. Context can be defined as

"any information that might be used to characterize

the situation of entities that are considered relevant

to the interaction between a user and an application"

(Dey et al., 2001). Context can be modeled in various

ways. Some of the most common forms to represent

context are: key-value pairs, object orientation model,

logic based model, ontologies, and mark-up

languages. According to Bettini et al. (2010),

ontological models have greater capacity for

representation and inference. Spatial models are more

efficient if compared to ontologies and object

oriented modeling, but they do not have as much

representation capacity compared to the ontological

models. This way, ontologies are the most used form

to represent context information in ubiquitous

architectures. Context-Aware Data Query states that

the data retrieval and filtering operations should be

based on context information reported to the system

ICEIS 2016 - 18th International Conference on Enterprise Information Systems

240

time of information query. Thus, stored information,

whether in structured or unstructured form, can be

adapted according to the given context. As the context

involves large sets of information, defined in fields, it

can also be considered a document in a collection.

Thus, contexts should be available for query in the

same way that a common document related to the

domain. Therefore, the context can be used in two

ways in information retrieval (Bolchini et al., 2013):

(a) To derive a query that returns the documents that

best fit the required context; (b) To treat context as a

document, i.e., the context becomes the source of

information being queried.

Ontology is defined as a formal and explicit

specification of a group of concepts in a shared form

(Borst, 1997). As context model is a type of

knowledge about the environment and the entities that

composes the environment of the user and system, it

can be represented in ontologies. Research have

modeled context ontologies. These ontologies vary in

multiple aspects, for example the number of defined

concepts, domain of application, validation methods

and language of representation. Rodríguez et al.

(2014) conducted a comparison between ontologies

that represent context information and activities. It

was found that PiVOn Ontology (Hervás et al., 2010)

was the ontology that better attends the comparison

criteria. This ontology has been used in a set of works

related to ubiquitous computing (Rodriguez et al.,

2014). Context information are essential for

ubiquitous systems because the treatment of this

information and its use allows that ubiquitous system

be able to adapt itself to the needs of users and other

systems. These adjustments must be made in real time

and can result in both application behaviour change

and in information retrieval.

3 STATE OF THE ART

Recent research propose models and extensions of

existing tools and models to integrate context models

and domain data querying. HyConSC (Anderson et

al., 2006) is a framework that allows context-based

consultations to be integrated to applications that use

relational databases. To realize the extension of

queries, the framework uses its own context model,

represented in graphical model. The context

information are persisted as notes in documents. The

Context-Relational Algebra was proposed as an

extension of relational algebra, which supports a

logical model that allows to integrate contextual

information to relational databases (Martinenghi et

al., 2009). A photo recommendation based on context

tool was proposed by Viana et al. (2011). This tool

uses semantic context information to recommend

photos based on similarity calculation.

The CARVE methodology (Context-Aware

Automatic View Definition over Relational

Databases) was defined as a proposal for integration

between relational databases and context information

in the form of a process of automatic generation of

views based on context (Bolchini et al., 2013). The

implementation of the methodology is performed in a

number of phases, some of which must be manually

set. The context information is modelled in Context

Dimension Trees. HARE (Time-Aware Location-

Aware and Health-Aware Recommender) is a content

recommendation application that uses time, location

and health data of the patient to perform

recommendations (López-Nores et al., 2013). To

perform the content recommendation, ontologies

were previously used to describe the metadata about

that content. An architecture was proposed by Hahm

et al. (2014) to perform the custom recovery of

engineering documents based on analysis of user

profile. To conduct a qualitative analysis on the state

of the art in data query based on context, it was made

a list of important features for analysis:

Context Model: The way that context

information are modeled. It can be modeled in

relational form (BDR), object-oriented (OO), based

on logic (BL), graphical models (G) or ontologies

(Onto); Domain Model: The way the domain

information are modeled and integrated with context

information. It can be modeled in relational form

(BDR), as documents with semantic annotations

(DAS), or as domain ontologies (Onto); Integration

Mode: The way the integration of context

information and domain data is made. It can be based

on ontologies alignment (Alin), relational-algebra

expressions (RA), or integration by algorithms by

property analysis (Alg); Query Language: Data

recovery can be defined by relational algebra

expressions (AR), defined in SQL, SPARQL or

SQWRL; Database Model: Some researches do not

specify (NE) using DBMS, others use models like

Triple Stores (TS) or Relational databases (BDR).

Some researches do not use databases (NU). Based on

the list set up to carry out analysis of related work.

The result of the features analysis is shown in Table

1. As can be seen, none of the research models context

based on a generic context ontology. This contributes

to the context share issue (Makris et al., 2013) (Perera

et al., 2014) between ubiquitous systems. Generic

ontologies are often used in ubiquitous systems

generally to adapt the execution of services and to

infer contexts. However, there is the existence of a

Semantic Integration between Context-awareness and Domain Data to Bring Personalized Queries to Legacy Relational Databases

241

Table 1: Qualitative analysis of state-of-the-art.

Features / Work (Anderson

et al., 2006)

(Martinenghi

et al., 2009)

(Viana et

al., 2011)

(Bolchini et

al., 2013)

(López-Nores

et al., 2013)

(Hahm et

al., 2014)

Context Modeling

BDR BDR Onto G Onto Onto

Context Modeling based on a generic ontology

- - - - - -

Domain Modeling

BDR BDR Onto / DAS BDR Onto Onto / DAS

Integration Mode

Alg AR Alin AR Alin Alin

Query Language

SQL AR SPARQL AR SPARQL SPARQL

Specific for a Domain

No No Yes No Yes Yes

Database Model

NE BDR NE BDR TS NU

gap for binding contexts and ubiquitous systems

domain data (Perera et al., 2014).

Another important feature that was observed in

relation to the related work is that none of the tools

modeled context information in ontologies - to be the

most complete and extensible, and integrated it with

relational databases - the most widely used format for

domain data, and persisted only in a database format.

Thus, it is proposed in this paper a framework of

integration and recovery of domain information based

on context, modeled in ontologies, and relational

databases, which represent domain information.

4 A FRAMEWORK TO

INTEGRATE CONTEXT AND

RDBMS

This paper presents a framework for domain

information query. This information is modeled and

persisted in relational databases, and context

information are modeled on ontologies. This way, the

model uses an ontology-based model to perform

information filtering based on context information.

So even without changing the structure of the

database that represents the application domain

information, systems can use context information

modeled with high expressiveness for querying this

information. Figure 2 presents a high-level view of

the proposed model to integrate context modeling

with relational data querying. The framework is

divided in five main levels, named:

(a) External Entities: Ubiquitous middleware

perform management of context and environmental

resources. Even in these environments, external

systems can use resources from middleware. As

ubiquitous middleware manages context information,

they may be required to inform the current context of

the environment when performing a query. External

systems in turn hold queries to data using the model,

which returns a data set that comply with the filters

informed by external systems, and context informed

by ubiquitous middleware;

(b) Interfaces: To communicate with the

framework, ubiquitous middleware use REST

interfaces, informing the context through

representations in JSON-LD language. The process of

serialization of context ontologies in JSON-LD was

previously presented in (Maran et al., 2015). External

systems in turn carry the information query using

SQL language;

Figure 2: Overview of the framework.

systems through SQL format is extended through a

process. This process performs the query in association

with rules that define links between context and domain

data and checks for relationships that can be used to

query. In this process, the definitions of ontologies are

used to describe context, domain database schema, and

the alignment of definitions of concepts;

(d) Conceptual Layer: This layer represent

conceptual schemas and ontologies used by the

model. In this paper, conceptual model is a modeling

on a specific area, where this model was the basis for

the logic model for a database that stores data about

the domain. Some authors classify this type of

conceptual modeling as a lightweight ontology

(Maran et al., 2015a). To perform context-based

information query, the model uses three distinct sets

of settings: (i) an ontology that describes context

based on PiVOn (Hervás et al., 2010), a context

ontology independent of application, (ii) A conceptual

ICEIS 2016 - 18th International Conference on Enterprise Information Systems

242

model that describes the domain database., and (iii) An

ontology defining generic concepts and relationships to

allow the alignment of context and domain

information. The alignment of the ontologies was

previously presented in (Maran et al., 2015);

(e) Persistence: Instances of conceptual models

and ontologies used by the proposed model are

persisted in a relational database. An initial

implementation of the serialization of the definitions

was presented in (Maran et al., 2015a).

5 CONCLUSIONS

Ontologies have been used by ubiquitous

architectures for representing context information.

Furthermore, inference rules have been used for

making inferences about the context, which according

to current definitions is measured and inferred

knowledge about the status of entities.

Relational databases are used in most applications.

As shown by motivating scenario, the context of use in

the structured information retrieval in relational

databases is relevant. In this work, an overview of the

field was presented, as an study about the state of the

art and the proposal of a model of integration between

context, modeled on ontologies and domain

information, modeled in relational databases. The

framework is in implementation phase. As the future

work, we pretend to evaluate it in a scenario based in

the motivational scenario presented in this work.

REFERENCES

Anderson, K. M.; Hansen, F. A.; Bouvin, N. O., Templates

and queries in contextual hypermedia. In: Proceedings

of the seventeenth conference on Hypertext and

hypermedia. ACM, 2006. p. 99-110.

Bettini, C., Brdiczka, O., Henricksen, K., Indulska, J.,

Nicklas, D., Ranganathan, A., Riboni, D. A survey of

context modelling and reasoning techniques. Pervasive

and Mobile Computing, v.6, n.2, p.161-180, 2010.

Bolchini, C; Quintarelli, E; Tanca, L. CARVE: Context-

aware automatic view definition over relational

databases.Information Systems,v.38,n.1,p.45-67, 2013.

Borst, W., Construction of engineering ontologies for

knowledge sharing and reuse. Universiteit Twente, 1997.

DBEngines. Knowledge Base of Relational and NoSQL

Database Management Systems. Website. Available at:

http://db-engines.com/en/ranking.

Dey et al., A conceptual framework and a toolkit for supporting

the rapid prototyping of context-aware applications.

Human-computer interaction, v.16, n.2, p.97-166, 2001.

Edx Website. Available at: code.edx.org/.

Hahm, G. J. et al. A personalized query expansion approach

for engineering document retrieval. Advanced

Engineering Informatics, v.28, n.4, p.344-359, 2014.

Hervás, R.; Bravo, J.; Fontecha, J., A Context Model based

on Ontological Languages: a Proposal for Information

Visualization. JUCS, v. 16, n. 12, p. 1539-1555, 2010.

López-Nores, M. et al. Context-Aware Recommender

Systems Influenced by the Users’ Health-Related Data.

In: User Modeling and Adaptation for Daily Routines.

Springer London, 2013. p. 153-173.

Makris, P.; Skoutas, D. N.; Skianis, C. A Survey on

Context-Aware Mobile and Wireless Networking: On

Networking and Computing Environments' Integration.

Communications Surveys & Tutorials, IEEE, v. 15, n.

1, p. 362-386, 2013.

Maran, V., Palazzo M. de Oliveira, J., Pietrobon, R.,

Augustin, I. Ontology Network Definition for

Motivational Interviewing Learning Driven by

Semantic Context-Awareness. In Computer-Based

Medical Systems, 2015 IEEE 28th I.S. on. p. 264-269.

Maran, V., Machado, A., Augustin, I., Wives, L. K., & de

Oliveira, J. P. M. (2015a). Proactive Domain Data

Querying based on Context Information in Ambient

Assisted Living Environments. In: 17th International

Conference on Enterprise Information Systems.

Martinenghi, D.; Torlone, R., Querying context-aware

databases. In: Flexible Query Answering Systems.

Springer Berlin Heidelberg, 2009. p. 76-87.

Machado, G. M.; Palazzo, J., Context-aware adaptive

recommendation of resources for mobile users in a

university campus. In: Wireless and Mobile

Computing, Networking and Communications, 2014

IEEE 10th International Conference on. p. 427-433.

Perera, C., S. Member, A. Zaslavsky, e P. Christen. Context

aware computing for the internet of things: A survey.

Communications Surveys & Tutorials, IEEE, v. 16, n.

1, p. 414-454, 2014.

Pretz, K. “Low Completion for MOOCs”. IEEE Roundup.

Available at: http://theinstitute.ieee.org/ieee-roundup/o

pinions/ieeeroundup/low-completion-rates-for-moocs.

2014.

Quinn, S., Bond, R., & Nugent, C. D. “An Ontology Based

Approach to the Provision of Personalized Patient

Education”. In: Ambient Assisted Living and Daily

Activities. p. 67-74. Springer. 2014.

Rodriguez, N. D. et al., A survey on ontologies for human

behavior recognition. ACM Computing Surveys, v. 46,

n. 4, p. 43, 2014.

Stavropoulos, T. G. et al. aWESoME: A web service

middleware for ambient intelligence. Expert Systems

with Applications, v. 40, n. 11, p. 4380-4392, 2013.

Strang, T.; Linnhoff-Popien, C. A context modeling survey.

In: Workshop Proceedings. 2004.

Viana, W. et al. Towards the semantic and context-aware

management of mobile multimedia. Multimedia Tools

and Applications, v. 53, n. 2, p. 391-429, 2011.

W3C Website. Available at: www.w3.org/.

Weiser, M. The computer for the 21st century. Scientific

american, v. 265, n. 3, p. 94-104, 1991. Available at:

http://doi.acm.org/10.1145/329124.329126.

Semantic Integration between Context-awareness and Domain Data to Bring Personalized Queries to Legacy Relational Databases

243