An Ontology-Based Approach for Detecting Drug Abuse Epidemiology

Abstract

Drug addiction is a kind of brain disease that is associated with developing its habit which in turn affects human behavior. The impact of the addiction can be far more obliterating. Drug abuse affects human being as in Cardiovascular syndrome, stroke, melanoma, HIV/AIDS, hepatitis, kidney, lung and liver damage, neurological and hormonal effects and hence mortality. The ontology based system that has been developed on drug abuse is unfortunately only one that is named as PREDOSE. The main goal of PREDOSE was to extract the knowledge that is relevant to the illicit drug usage from websites and blogs. PREDOSE lacks the brand names and slang terms of drugs along with the risk factors associated with the illicit drugs. It does not provide the information of banned drugs in different countries. The extended version that is the main aim of this research is E-DAO that overcomes all the lacks of PREDOSE and thus provides efficient results. It facilitates the end users in many ways by providing the useful information related to drug abuse and its banned information. Using this new approach for extended ontology, users can query the knowledge base for information like getting the street names or slang terms for the drugs. Users can also get to know about different company names for a single chemical formula drug alongwith their side effects. Scarce information is updated and extended in E-DAO as numbers of classes were increased from 43 to 114. Moreover, the model becomes more time efficient by defining the semantic rules.

Full text

!
An Ontology-Based Approach for Detecting Drug
Abuse Epidemiology!
Saleha Asad1, Mansoor Ahmed2, Abid Khan3, Adeel Anjum4, Munim Ali Shah5, Qaisar Javaid9 , Nadeem Javaid10 ,
Umar Manzoor6, Bassam Zafar7, Zeeshan Ahmed8
salehash1@yahoo.com1, {mansoor, abidkhan, adeel.anjum, mshah}@comsats.edu.pk, uelahi@kau.edu.sa6,
bzafar@kau.edu.sa7, zeeshan.ahmed@indus.edu.pk8, qaisar@iiu.edu.pk9, nadeemjavaidqau@gmail.com10
COMSATS1,2,3,4,5,10 Institution of Information Technology,
Islamabad, Pakistan
8Department of Computing, FEST
Indus University Karachi
6,7King Abdulaziz University, Jeddah, KSA
9Department of Computer Science & Software Engineering,
International Islamic University, Islamabad, Pakistan
Abstract!
Drug addiction is a kind of brain disease that is associated with developing its habit which in turn
affects human behavior. The impact of the addiction can be far more obliterating. Drug abuse
affects human being as in Cardiovascular syndrome, stroke, melanoma, HIV/AIDS, hepatitis,
kidney, lung and liver damage, neurological and hormonal effects and hence mortality. The
ontology based system that has been developed on drug abuse is unfortunately only one that is
named as PREDOSE.The main goal of PREDOSE was to extract the knowledge that is relevant to
the illicit drug usage from websites and blogs. PREDOSE lacks the brand names and slang terms
of drugs along with the risk factors associated with the illicit drugs. It does not provide the
information of banned drugs in different countries. The extended version that is the main aim of
this research is E-DAO that overcomes all the lacks of PREDOSE and thus provides efficient
results. It facilitates the end users in many ways by providing the useful information related to
drug abuse and its banned information. Using this new approach for extended ontology, users can
query the knowledge base for information like getting the street names or slang terms for the
drugs. Users can also get to know about different company names for a single chemical formula
drug along with their side effects. Scarce information is updated and extended in E-DAO as
numbers of classes were increased from 43 to 114. Moreover, the model becomes more time
efficient by defining the semantic rules.!
Keywords–Drug Abuse; Epidemiology; Ontology; Vocabulary; Knowledge Base; Opioid!
!
!
!

!
1. INTRODUCTION
The eternally increasing intensification in the number of illicit use of drugs due to the esteem of
blogging and social media podium discussion is creating a foremost public health problem. An
estimated 183,000 illicit drug usage related deaths were reported in 2012[1]. There are an
increased number of deaths using illicit opioid. A poisoning drug accelerating accidental death
rates because of overdose death rates[2]. The maturity of effectual prevention and strategy
processes needs timely and unswerving information on up-to-the-minute and evolving drug
trends. There are some present epidemiological structures, such as “National Survey on Drug
Use and Health (NSDUH)”[3], the “Community Epidemiology Work Group (CEWG)”[4]and
the “Drug abuse Warning Network (DAWN)” provides an essential data for drug abuse
tendencies. The “National Survey on Drug Use and Health (NSDUH)” attain information on
nine diverse classifications of illegal drug use that consist of: heroin, marijuana, , hallucinogens,
cocaine, inhalants, and nonmedical usage of prescription-type pain relievers, narcotics,
stimulating substance, and tranquilizers [5]. !
!
Drug addiction is a kind of brain disease that is associated with developing its habit which in
turn affects human behaviour. The impact of the addiction can be far more obliterating. Drug
abuse affects human being as in Cardiovascular syndrome, stroke, melanoma, HIV/AIDS,
hepatitis, kidney, lung and liver damage, neurological and hormonal effects and hence mortality.!
!
With the developing World Wide Web, there is a mammoth increase in sharing knowledge of
every class. People dig up the platform to thrash out their issues and get elucidation. Sometimes
this can go wrong. There are many blogs and social networking websites where people explore
addict certification of the illegal usage of particular pharmaceutical opioid[6]. To study the
variety of areas, Web-based data can be used. Previous work on web-based data lacks
methodology and they relied on a limited scope. Whereas ontology-based system used meant for,
drug abuse epidemiology study.!
!
A semantic web application was created named as PREDOSE [7], which is capable of extracting
semantic information from social media and provide levels of content analysis that support the
more effective epidemiologic description of prescription drug abuse. The main goal of
PREDOSE was to help researchers to gain the knowledge of attitudes and deeds of drug abusers
related to unlawful use of opioid through automatic extraction of relevant contents. Sometimes
there are some few reasons that show to be deficient in structure or designation. To extend the
usability of previous work we have added some additional features that enhance the working of
the system along with the efficiency of the framework. DAO attains some of the additional
requirements like it lacks the ability to accommodate street or slang terms of drugs along with
the information of banned drugs in different countries and risk factors associated with them
whereas brand names of such drugs were not being dealt by existing framework: to overcome
those flaws are the main contribution of this research.!
!

!
A. Ontology vs. DB approach
“Ontology refers to the explicit specification of the conceptualization of a domain which
captures its context”[8]. Mostly non-ontological approaches are based on an invaluable abet for
distilling and interpreting drug abuse hence resulting in different analysis, they suffer from few
limitations. Most notably, there is need for data integration and consistency, need to understand,
identify and analyze higher granular level. !
!
PREDOSE a novel semantic web approach used to extract semantic information about drug
abuse epidemiology from social networks and media. It also provides modules for data analysis
and understanding that support emerging pattern analysis, content examination; and tendency
detection. Ontological model is flexible and is easy. Ontological models can be merged, shared
and reused amid the entities. Table 1 defines are major differences between ontology and non-
ontology approaches.!
!
Ontology Based
DB Based
Focus on meaning
Focus on Data
In ontology axioms are used to specify
meaning or integrity
In DB, constraints are used that may
hind for meaning
ISA hierarchy is backbone
No ISA hierarchy
Infer new information and ensure
consistency
User can just query and see the views
Instances are optional
Instances are central points
Comments as a part of ontology
Data dictionary as separate artifact
Table 1: Ontology Vs Non Ontology Based approach!
B. Contribution
The ontology used in PREDOSE known as Drug Abuse Ontology (DAO) is a formal depiction
of concepts and relationships connecting them for the recommendation drug abuse sphere[7].
There were some limitations of DAO, like providing information about banned drugs and their
risk factors along with the information of street or slang terms of illicit drugs that are
incorporated in the proposed extended ontology. The proposed extended ontology named as
Extended- Drug Abuse Ontology (E-DAO) is a novel approach for employing the use of
semantics. It has the following key contributions:!
1. E-DAO removes scarcity in the previous version of the ontology. DAO has 43 classes and
20 properties whereas E-DAO has 114 classes, 5 data property and 43 object property.
2. Slang or street named for illicit drugs are not mentioned completely in DAO whereas they
are precisely mentioned in E-DAO.

!
3. In DAO, there was no information regarding drugs that are banned in specific countries or
states, whereas E-DAO deals with this flaw as well.
4. Only chemical names of drugs are mentioned or queried in DAO, the brand names were not
conferred. E-DAO mentioned chemical as well as company or brand names of drugs.
5. Ontology is evaluated on following criteria: It includes, (1) Formal correctness or Accuracy
(2) Structure, (3) Completeness and Conciseness (domain coverage), (5) Vocabulary,
Concept, Data (6) Expandability and Reusability, (7) Clarity, (8) Computational complexity,
Integrity and Efficiency, (9) Preciseness and Quality, (10) Ontology expressiveness.
!
2. RELATED WORK
In the past few years, scientific neighbourhood, especially from medical domain, has involved
many efforts and resources in the development of e-technologies. The Semantic Web is
considered as a school-of-the-art of the World Wide Web where the data at hand can be
integrated and evaluated effortlessly. Review of the literature on the use of ontological and non-
ontological models in drug abuse epidemiology is mentioned in Appendix A, it also states about
the related work, our paper discussion is on following lines [7], [9] [10] [11].!
A. PREDOSE:
“PREscription Drug abuse Online Surveillance and Epidemiology”[7] is the semantic web
platform. The knowledge base that is developed for this framework is DAO (Drug Abuse
Ontology). Semantic information is extracted from User Generated Content (UGC) by web
crawlers and other semantic techniques. PREDOSE is publically available as web application
along with the DAO is also freely available. Limitations of DAO include the lack of
demographic indicators along with the unavailability of geographical information; all the slang
terms are not included in the current DAO. In current DAO only chemical names of drugs were
being handled whereas the company names were being omitted in it. !
B. SimARC:
Alcohol [15] is the most extensively battered psychoactive drug. Too much drinking habit can
harm your health. 88,000 deaths happen due to excessive alcohol intake and it shortens the life
of those who die by almost 30 years [16]. To investigate the intricacy of alcohol use and abuse,
we need to include the already said risk factors along with the progression, for this our agent-
based model SimARC[9] (Simulation of Alcohol-Related Consequences) is designed aim to
summarized levels of originality. In SimARC, there are three levels for analysis; A Micro-Level
- it contains individual, his/her neurologic, physiologic and social distinctive, A Meso-Level –
includes groups, peer sway and momentous others, A Macro-Level – includes open policies,
metropolitan topography and communal responses. !
C. Twitter data study on marijuana concentrates through the U.S:
Marijuana is getting famously familiar. They are typically inhaled or vaporised via a vaporizer or
electronic cigarettes. Statistics on the epidemiology of marijuana use remains inadequate. For
analysing the twitter data on marijuana concentrates, tweets were collected using twitter’s
streaming API and then filtered using eDrug Trends[17] platform. !

!
!
D. A Web-Based Study of Self-Treatment of Opioid Withdrawal Symptoms with
Loperamide:
There is a huge number of websites that provides a medium for individual to discuss illicit drugs.
Such website was selected and analysed by retrieving the web-forum posts using Web Crawlers
and kept in the local text corpus. Those text corpuses were queried to extract post having
Loperamide with its brand names and slang or street terms. Loperamide is a drug used to reduce
the incidence of diarrhoea and every so often used for short bowel, syndrome gastroenteritis and
inflammatory disease. But this drug is not suggested for those people who have blood in the stool
[18]. From UGC, 1200 posts were identified with Loperamide and then the reason of using,
dosage and effects of Loperamide use were also extracted. Since 2005, there has been a rise in
increasing illicit use of Loperamide particularly in 2010 and 2011.!
!
E. Study of Social Web Data on Buprenorphine Abuse Using Semantic Web Technology
Using the PREDOSE platform, researchers’ uses paradigmatic approach combining with
Semantic Web technology, NLP and machine learning to achieve the specific aims of this study.
This study depicts knowledge, attitude and behaviour of drug users discussed on web forums
towards the non-medical usage of Suboxone and Subutex along with its identification and
description of patterns[12].!
!
!
F. Semantics enabled role based sentiment analysis for drug abuse on social media: A
framework
The main aim of this study is to speed up the epidemiological drug abuse using social media, a
comprehensive semantic web based framework is introduced that incorporates with role based
ontology model. The structural design of this framework not only develops the trends of drug
abuses but also has a tendency to suggest stakeholders that how abusers are inclined towards
restraining them the illegitimate drug usage. Role-based ontology hierarchy calculates the
influence factor of people surrounding the abusers and suggests the suitable personality who
could have a positive influence for the abuse to withhold them from substance abuse[13].!
!
G. Social Media Mining for Toxicovigilance: Automatic Monitoring of Prescription
Medication Abuse from Twitter
The role of social media behavior in drug abuse is increasing these days. The goal of this study
is to authenticate the information regarding the prescription drug abuse. For this, the researchers
prepare an annotation recommendation to explain the Twitter records and use reasonable tasters
of annotations to examine if programmed categorization methodologies used to identify the
information related to drug abuse from twitter records automatically. The system is proficient
enough to classify abusive and non-abusive tweets accurately whereas its performance can also
be increased using the larger amount of annotated data[14]. !

!
H. Google Play Store Applications:
There are some of the applications that are related to drug abuse epidemiology in the Google
play store. Some of them are listed below along with their features and efficiencies:!
i. Prescription Drug Abuse:
This Google application provides complete information about prescription drug abuse disease. It
covers the symptoms and treatment of prescription drug abuse. This application has been
developed for everyone. The main features of this application are as follows:!
● How to prevent from drug abuse!
● Definition, description of disease and symptoms sign!
● Test and diagnosis!
● Internal medical disorders!
● Drug and medication for prescription drug abuse!
● Treatments and prevention!
!
ii. Drug Addiction:
This application is for drug addicts, the family of addicts or friends or anyone who wants to
help or break the addiction. A number of emotional and psychological programs introduce into
taking drugs. With the passage of time, drug addiction increasing among people. For this, a
12-step program used as a recovery method. Alcoholics anonymous and Narcotics anonymous
are prominent examples used to break addiction hence there are habilitation centres for it. A
good treatment and care can take along about the recovery from drug addiction.!
!
iii. Drugs.com
● This Google application is based on a website by providing all the information about
the drugs in an application. The main key features of this app are listed below:!
● My Drug List – one can increase the prescriptions to instantaneously accumulate
relevant medical information in a modest and easy to read format.!
● Complete Drugs A to Z listing – It provides fast searching and accurate suggestion.!
● Pill Identifier – As its database can be updated on daily basis, It Identifies the
medicines by simply entering its name, colour or shape.!
● Interaction Checker – It provides a list of relations among different drugs in use at
the same time.!
!
iv. Pharmapedia Pakistan
Pharmapedia Pakistan provides comprehensive information on drugs, brands, prices and much
more. It gives the access to information on generic drugs, medicine brands in Pakistan with
prices and available forms, alternate brands and much more. It has following features:!
● Drugs overview, dosage, indications, side effects, contraindications and high-risk
groups.!

!
● Available brands or every generic drug formula, including a combination of generic
drugs.!
● Prices and available forms.!
● Alternate brands with rice and company along with the form of the drug.!
!
3. ONTOLOGY ENGINEERING METHONTOLOGY:
“Ontology is an explicit specification which uses a formal data structure with the common
depiction of domain concepts”[8]. To design Ontology-based approach is an iterative process. It is
used to define different concepts of ontologies and its scope. E-DAO has been developed with
following given objectives that provide the key contributions to our extended version of ontology:!
1. Detailed description of all of drugs that are used as unlawful act
2. Ability to accommodate Street and slang terms of illicit drugs
3. Provides generic risk factors associated with the illicit drugs
4. Including chemical along with the brand or company names to identify the illicit drugs
5. Providing the information of banned drugs in different states
6. Facilitate the reusability of existing ontology over time according to the framework logic
!
3.1. Methontology (ontology engineering framework)[19]:
Based on IEEE standards[20] for software development, Methontology is basically an ontology
engineering framework for ontology development. Broadly this concentrates on various stages of
ontology engineering and smoothes the progress of ontology structure in a methodical way. The
base for this framework is developing prototypes; along with procedure and techniques to carry
out each movement. Lifecycle permits the maturity of the ontology that allows untimely
authentication and tuning. The life cycle activities of Methontology are composed of the
specification, conceptualization, formalisation, and implementation and maintenance activities.!
!
4. DEVELOPMENT OF ONTOLOGICAL MODELS FOR DRUG ABUSE:
This scaffold facilitates manufacture of ontology in a methodical way. Ontology substantiation
and justification has been carried out using OntoClean[21]. This methodology helps ensure the
steadiness, comprehensiveness and succinctness of the ontology. The bottom-most layer
represents the conceptualization in which concepts of our domain in question i.e. drug abuse via
the Protégé tool's API[22]. The ontology layer facilitates the expression of classes in logical
form. The inference layer infers new knowledge and adds to the knowledge base for future
extractions. The later layer representing Rules uses Semantic Web Rule Language (SWRL)[23]
for parsing, reasoning and rule generation. SWRL also allows Query construction for retrieving
information via the simple protocol and resource query language (SPARQL)[24].Ontological
Knowledge Base provides the finest acquaintance demonstration, reduces redundancy and
challenge among the concepts and facilitates the information analysis.!
!
!

!
4.1. Formal representation of ontology model:
Ontologies are not a just combination of terminologies or vocabularies rather they can conquer a
semantic level that can incarcerate the milieu of a specific domain. The basic endeavour of the
proposed ontology is to articulate the multifarious awareness of drug abuse domain in a way that
it can be computationally distinguishable, machine processable and composition of information
facilitates communiqué among software agents[25]. The ceremonial designing of the ontological
model can be characterised as:!
!
O = Ontology system = (C, A, I, P)!
!
C is the conceptualization for instances and concepts; P stands for Properties and relationships;
A: Axioms and rules and I is for interpretation of model. They can additional be intricate as
follows[26]: !
!
!
!
!
!
Where C is a set of all classes with type t, I is a set of all instances with type i and P is a set of all
properties of type p, including data type and object type. The symbol Rel represents the set of all
relationships of type e. Similarly, A is a set of all axioms and R stand for rules.!
!
Properties and relationships can be formally represented as:!
!
P = (D: data type; O: object; T: transitive; F: functional) !
!
Rel = (Ξ: equivalence, C: subsumed, Π: Disjoint)!
!
The system acquaintance base consists of all concepts, instances, assertions, properties and
relationships in ontology representation. Formally represented as below:!
!
!
!
!
Domain of property: !
!
!
!
Range of Property: !
!
!

!
!
!
!
!
!
The interpretation of ontology O is known as the model of O. Every concept is allied with some
object. There are many relationships possessed by classes and properties like equivalence,
subsuming and disjoint relationship respectively. ≡ is the equivalence; ∩ is the conceptually
disjoint and is symmetric, reflexive and transitive. is known as conceptually subsume and this
relation is irreflexive, transitive and asymmetric.
!
Some examples of these relationships in the drug abuse ontology model are:!
!
Place≡Region!
DrugUse∩DrugEffects!
DrugDose Unit!
Antihistamines Drug!
Drug Cause!
!
4.2. Extended Drug abuse ontology model
The ontological layer defines restrictions upon classes by using OWL(Ontology Web
Language) and facilitates logic. Rule layer can be used to define some rules on ontology by
using SWRL (Semantic Web Rule Language) The top-most layer also called inference layer
using Inference engine that will check the consistency of classes and perform reasoning on
defined ontology and get conclusions. Knowledge contains the information of all possible kinds
of drugs, its consequences and drug effects. Knowledge engineer populates the Knowledge Base
and involved technically and scientifically in building, maintaining and using KBS. The end
user (Researcher, hospital or patient) can interact with Knowledge-base by using query
interface; in results. The model includes the semantic annotations of Drug, Drug effects, etc.
The moreover model represents specific details for each drug and its effect along with the
detailed properties. Our Ontology Layer Model has been shown in figure 1 and it has following
steps:!
!

!
!
Figure2: Architecture of Drug Abuse Ontology!
!
Step 1: Populate Knowledge Base!
!
Knowledge engineer populates the Knowledge base and involved technically and scientifically in
building, maintaining and using KBS (knowledge base systems). Knowledge Engineering plays an
important role and first appeared in 1990 in the field of AI. Knowledge engineer can be involved in
the development of several technologies such as Knowledge base systems, data mining, decision
support systems, AI (artificial intelligence), expert systems and in neural networks (NN). The main
job of knowledge engineer is to work with a client who wants a knowledge base or an expert
system, created for them and involve in verification and validation.Validation means they ensure
that the data they bring together is correct and conform within accepted standards. It is necessary
for a knowledge engineer to collect detailed domain knowledge by interacting with a domain
expert. Domain expert (or Admin) is who is an expert in their particular domain knowledge. During
the development process, firstly knowledge retrieves from the domain expert and makes discussion
with users. After retrieving domain knowledge design a prototype model for reasoning and planning
of a particular problem. Then last but not the least performs the evaluation. In an evaluation,
validation and verification of the knowledge acquired by the experts can be done.!
!
Step 2: Knowledge Base Modeling!
!
To solve complex problems the knowledge Base (KB) consists of the collection of related
domain knowledge. Knowledge contains the information of all possible kinds of drug abuse that is
being done by humankind. It describes the knowledge base modelling and all the rules, drugs &
side effects that will be included in ontology including their slang and street codes.!
!
Step 3: Reasoning!
!
The Knowledge Base sends extracted knowledge to Inference Engine and Adding some rules to the
reasoning logic. Inference engine will perform reasoning on the extracted information and get
conclusions. As reasoner operated on knowledge Base, the important aspect is that the resulted

!
information is stored in the knowledge base for effective and efficient use of knowledge in the
future. Rule-based reasoning can be used for solving problems. Figure 6 describes the overall
reasoning by queries. Knowledge Base sends extracted knowledge to reasoner. The reasoner
performs reasoning on the extracted information. Results are stored in knowledge base for future
information extraction.!
!
4.3. Results of E-DAO
SPARQL query language is used to query the E-DAO ontology. Accurate results show the
accuracy and correctness of Extended- Drug Abuse Ontology (E-DAO). Furthermore, reasoner is
used to checking the consistency of our proposed ontology and it also supports inference. There are
several types of reasoners that support protégé such as HermiT 1.3.7, Fact++, Pellet and TrOWL
but E-DAO used Fact++ to reason the extended ontology.!
4.3.1. Hierarchical representations of classes and subclasses
Taxonomic contractor rdfs:subClassOf can be used to define the classes and subclasses, These
classes are more accurate, for that we can say If Drug is a parent class of Alcohol, then every
instance of Alcohol is an instance of Drug. In this model, drugs can be represented as the top level
concept; the subclasses of drug effects are acute effects and medical conditions. In extended drug
abuse ontology model, attributes are interring linked concepts. This extended ontology model
covers top 8 types of drugs. Ontology inconsistent means that the defined individual is an instance
of two disjoint classes.!
!
Figure 3: Class hierarchy of E-DAO!
4.3.2. Object & Data Properties of E-DAO:
The E-DAO defines object properties[27] as causes, effects, banned_in, not_banned_in,
Has_amount, has_brand_Name, has_molecular_weight, has_risk _of and many more as shown in
fig.12. To provide the descriptive metadata annotation properties can be used[28]. But during

!
reasoning process annotation properties[29] are not involved. Also object property defines the
restrictions and condition on predicate among two objects.!
!
!
Figure 4: Object & data Properties of E-DAO!
!
4.3.3. Instance of E-DAO
After constructing the classes and properties, now it is easy to describe their members. About
individual specific property can be asserted. To define the particular instance (also called
individual), first select the required class for instance and then create its instances as in figure 5!
!
!
Figure 5: Instances by type of E-DAO!
!
!

!
5. IMPLEMENTATION & EVALUATION
The best advantages of ontology-driven system are the provision of presumption competence that
reason the knowledge base. The new knowledge is also created after reasoning as a result of
understood derivations for future extractions. The inferred facts are derived from ontological
concepts and their relationships. Rich and expressive constraints on concepts and relationship can
boost the reasoning capability. The purpose of reasoning is to substantiate the knowledge base. New
facts are inferred from existing knowledge base during the reasoning process. There are two basic
types of reasoning: Inductive and Deductive.Another way or reasoning classification is forward and
backwards chaining.!
5.1. Cardinality:
Within the class environment, OWL offers three concepts for checking the cardinality of properties
locally. OWL Lite can be used to check all these three types of cardinalities, but it can only be used
with values "0" and "1"[30]. !
5.1.1. OWL:MinCardinality[31]
Maximum cardinality defines, as a class that contains all individuals having at least N semantically
miscellaneous values, where N is the value of the cardinality constraint. The following example
describes a class of individuals using minimum cardinality:!
!
!
5.1.2. OWL:MaxCardinality[32]
Owl: maxCardinality describes a class that contain all individuals having at most N semantically
diverse values, where N is a cardinality constraint.!
!
!
!
5.2. Ontology Evaluation Methodology – OntoClean[26]
The evaluation process provides substantiation for whether or not the yield of project swerves from
a frame of reference. Verification, assessment and validation are included in the evaluation
process[33]. The model should perform desired functionality in the real world; the process of
ontology model verification should authenticate it. Ontology can be evaluated against different

!
criteria like consistency, clarity and many more. Consistency defines as all specifications,
comments, axioms and logical descriptions are driving to no contradictions and that the knowledge
can be inferred from model correctly whereas completeness test on ontological model checks the
entire concepts and relationships of the model i-e no incomplete concept exists in the model. !
5.3. Ontology Evaluation
The standard ontology evaluation is based on below mentioned a criterion that helps in the
effectiveness of the ontology.!
● Correctness!
● Consistency !
● Orientation of task!
● Completeness & conciseness!
● Expandability & reusability!
● Clarity!
!
!
!
Figure 6: The core concepts of ontology of extended drug abuse epidemiology!

!
The extended ontology E-DAO has some additional drug and its types as shown in figure 6. It
shows detail types of drug type named as Stimulant. !
!
Figure 7: Extension to the previous ontology (part a)!
Figure 7 shows the new hallucinogen drug type subclass named as Deliriants, Dissociative and
Psychedelics and further their subtypes that were not mentioned before.!
!
Figure 8: Extension to the previous ontology (part b)!
Figure 8 is the extended version of drug type Cannabinoids. The complete detail of subtypes of this
drug type is mentioned in the extended version of ontology.!
!
Figure 9: Extension to the previous ontology (part c)!
!
5.3.1. Correctness
This criterion states that all the information in ontology model like definitions, classes, properties,
individuals, axioms etc should be accurate according to some standards in domain of drug abuse
epidemiology. All the inferred statements should be valid according to our knowledge-base [34].
!

!
5.3.2. Consistency
Model is unswerving; all stipulations, annotations, axioms and commonsensical descriptions are
associated and motivating no contradictions. Moreover, all the equivalence (≡), subsume (7) and
disjoint (⊓) relations in the model were found to abide by with its characteristics. In the model, all
equivalence relation are reflexive, symmetric and transitive. All conceptually subsume (7) relations
are irreflexive, transitive, and asymmetric. Similarly, no conceptually disjoint (⊓) relations infringe
its properties of reflexivity, symmetry and transitivity. Every consistent model always pursue these
properties and will be free from any cycles and every acyclic ontology model is always sound[35].
5.3.3. Orientation of Task
Task Orientation is based on the criteria that the developed ontology is fulfilling the requirements
necessary to develop it. In present scenario, the aim of the ontology is to detect the outcome and
effects of illicit drug usage effectively and efficiently. !
5.3.4. Completeness and Conciseness
This evaluation tells that either the drug cruelty domain coverage is done properly. The proposed
ontological model covers all the well-known drug types and provides the required information
needed to express the illegal drug use effects. Moreover, the ontology consists of essential axioms
with negligible ontological commitments. It is free from the semantics of superfluous terms and
irrelevant axioms.!
5.3.5. Expandability and Reusability
E-DAO is expendable and it can be reused. The effort is required to add some new drugs with brand
names that are updated in coming years. New drugs can be defined without modification of existing
ontology model. The model can easily be reused for drug abuse epidemiology.!
!
5.3.6. Clarity
Our Ontology model can be easily tacit, analysed, manipulated, and reused. Each concept,
properties, relationship and axioms are undoubtedly stated, and documented using natural language.
The OntoClean methodology applies to class Drug of Extended Drug Abuse Ontology is shown in
Figure 10. +R shows that all the instances of class Drug will always be the instances of this concept
and +I on class Drug depicts that all the instances of this class carry unique identification criteria. -
D on class Drug shows that there is no dependency of this class on external concept whereas +U
shows the instances of this class are available as “Whole”.!

!
!
Figure 10: the core concepts of Extended Drug abuse ontology validation through OntoClean!
Figure 11 is another example of OntoClean evaluation of a part of E-DAO showing the evaluation
of class “Drug Effects”. +R in class “Drug effects” means that it is rigid and hence classes like
“acute effect” and “medical conditions” are also rigid in nature. +I show the unique identification of
all the classes. All are unique in nature and have their own uniqueness. –D in classes like “acute
effect” and “medical condition” shows that they are not dependent on the class drug effect and they
are independent as they have their own different subclasses from each other. !
!
Figure 11: Drug Effects in ontology evaluation through OntoClean!
5.4. Results of E-DAO
Following are the snapshots of results that were queried via SPARQL through the interface to check
the consistency of the proposed ontology:
!

!
1. Members or Slang terms
!
!
Figure 12: Members or slang terms of drugs!
2. Drugs banned in country
!
!
!
Figure 13: Query for drugs banned in specific country!
3. Brand names of a drug
!
!
Figure 14: Brand names of drugs!

!
4. Risk Factors of a drug!
!
!
Figure 15: Risk factors of drug!
5.5. Comparison of DAO & E-DAO
The ontology used in PREDOSE known as Drug Abuse Ontology (DAO) is a “formal depiction of
concepts and relationships connecting them for the recommendation drug abuse sphere”[7]. There
were some limitations of DAO that are incorporated in the proposed extended ontology The
proposed extended ontology named as Extended- Drug Abuse Ontology (E-DAO) is a novel
approach for employing the use of semantics. Comparison of E-DAO and DAO is measured in
below mentioned figure 16:!
!
Figure 16: Comparison of DAO and E-DAO!
6. CONCLUSION:
In the medical area very little work has been done on drug abuse domain in semantic knowledge,
whereas a lot of work has been done in this domain using other knowledge as mentioned in related
work section of this paper. Ontologies are widely used along with the thesaurus and vocabularies to
describe biomedical data. As the amount of data is increasing day by day, there should be an
increase in automation of data analysis. The medical domain is dynamic because new information is

!
constantly being added as the number of users increases consequently, network traffic also
increases.!
!
Easy access and process to the information by machines can only be done by a general framework
provided by Semantic Web technology. The heterogeneity of healthcare data makes it suitable for
Semantic Web Application. It endeavour to inspect the data by the judge it against the other similar
ontology and then connect and interlink the data providing easy querying by the end user[36].
Ensuring the drug safety is the key importance in public health. Semantic web technologies can be
used to integrate the data across all drug discovery and development units by providing the most
compassionate milieu for the premature revealing of safety related issues.[37]. Our framework
contains the extended version that is the main aim of this research is E-DAO that overcomes all the
lacks of PREDOSE and thus provides efficient results. It facilitates the end users in many ways by
providing the useful information related to drug abuse and its banned information. Using this new
approach for extended ontology, users can query the knowledge base for information like getting
the street names or slang terms for the drugs. Users can also get to know about different company
names for a single chemical formula drug along with their side effects. Scarce information is
updated and extended in E-DAO as numbers of classes were increased from 43 to 114. Moreover,
the model becomes more time efficient by defining the semantic rules.!
!
Future work includes the increment inconsistency along with the new information in drug abuse
domain. Whereas this ontology can also be mapped with other ontologies to give better results to
anti-narcotics domain to overcome the immense challenge for the health management that is the
issue of the non-medical exercise of prescription and illicit use of drugs.
!
7. REFERENCES!
[1] V. United Nations office on drugs and crime, World drug report. 2010 Jun 1.!
[2] M. N. Ilhan, E. Durukan, S. O. Ilhan, F. N. Aksakal, S. Ozkan, and M. A. Bumin, “Self-medication with antibiotics:
questionnaire survey among primary care center attendants.,” Pharmacoepidemiol. Drug Saf., vol. 18, no. 12, pp.
1150–1157, 2009.!
[3] S. A. and M. H. S. Administration, “Results from the 2013 National Survey on Drug Use and Health: Summary of
National Findings,” NSDUH Ser. H-48, HHS Publ. No. 14-4863. Rockville, MD Subst. Abus. Ment. Heal. Serv.
Adm., pp. 1–143, 2014.!
[4] Nih, “E pidemiologic Trends in drug Abuse,” Natl. Inst. Drug Abus. - Cewg, no. January, 2013.!
[5] A. Hughes, “State estimates of substance use from the 2005-2006 National Surveys on Drug Use and Health,” 2008.!
[6] S. F. Butler, S. W. Venuti, C. Benoit, R. L. Beaulaurier, B. Houle, and N. Katz, “Internet surveillance: content
analysis and monitoring of product-specific internet prescription opioid abuse-related postings.,” Clin. J. Pain, vol.
23, no. 7, pp. 619–28, Sep. 2007.!
[7] D. Cameron, G. a. Smith, R. Daniulaityte, A. P. Sheth, D. Dave, L. Chen, G. Anand, R. Carlson, K. Z. Watkins, and
R. Falck, “PREDOSE: A semantic web platform for drug abuse epidemiology using social media,” J. Biomed.
Inform., vol. 46, no. 6, pp. 985–997, 2013.!
[8] G. Antoniou and F. van Harmelen, A Semantic Web Primer, vol. 24, no. 04. 2009.!
[9] F. Lamy, P. Perez, A. Ritter, and M. Livingston, “SimARC : An Ontology-driven Behavioural Model of Alcohol
Abuse,” Alcohol, pp. 128–133, 2011.!

!
[10] R. et al. Daniulaityte, “Time for dabs: Analyzing Twitter data on marijuana concentrates across the U.S.,” Drug
alcohol Depend. 155 307-311., 2015.!
[11] Raminta Daniulaityte, Robert Carlson, Russel Falck, D. Cameron, S. Perera, L. Chen, and A. Sheth, “A Web-Based
Study of Self-Treatment of Opioid Withdrawal Symptoms with Loperamide,” Coll. Probl. Drug Depend., 2012.!
[12] R. Daniulaityte, R. Carlson, G. Brigham, D. Cameron, and A. Sheth, “A Study of Social Web Data on
Buprenorphine Abuse Using Semantic Web Technology,” 2016.!
[13] B. S. Raja, A. Ali, M. Ahmed, A. Khan, and A. P. Malik, “Semantics enabled role based sentiment analysis for drug
abuse on social media: A framework,” in 2016 IEEE Symposium on Computer Applications & Industrial Electronics
(ISCAIE), 2016, pp. 206–211.!
[14] A. Sarker, K. O’Connor, R. Ginn, M. Scotch, K. Smith, D. Malone, and G. Gonzalez, “Social Media Mining for
Toxicovigilance: Automatic Monitoring of Prescription Medication Abuse from Twitter,” Drug Saf., vol. 39, no. 3,
pp. 231–240, Mar. 2016.!
[15] P. Alcohol and U. States, “Alcohol Profile.”!
[16] C. D. C. National, C. Disease, and H. Promotion, “Alcohol and Your Health,” 2006.!
[17] D. Raminta, A. P. Sheth, R. Carlson, N. Ramzi, and P. T.K., “EDrugTrends - Knoesis wiki.” [Online]. Available:
http://wiki.knoesis.org/index.php/EDrugTrends. [Accessed: 22-Oct-2015].!
[18] T. Litovitz, C. Clancy, B. Korberly, A. R. Temple, and K. V. Mann, “Surveillance of Loperamide Ingestions: An
Analysis of 216 Poison Center Reports,” J. Toxicol. Clin. Toxicol., vol. 35, no. 1, pp. 11–19, Sep. 2008.!
[19] M. Fernández-López, a Gómez-Pérez, and N. Juristo, “Methontology: from ontological art towards ontological
engineering,” Assessment, vol. SS-97–06, pp. 33–40, 1997.!
[20] S. Engineering and S. Committee, IEEE Standard for Developing Software Life Cycle Processes, vol. 1997. 1997.!
[21] N. Guarino and C. Welty, “Evaluating ontological decisions with OntoClean,” Commun. ACM, vol. 45, no. 2, pp.
61–65, Feb. 2002.!
[22] S. A. McIlraith, D. Plexousakis, and F. van Harmelen, Eds., The Semantic Web – ISWC 2004, vol. 3298. Berlin,
Heidelberg: Springer Berlin Heidelberg, 2004.!
[23] M. O’Connor, H. Knublauch, S. Tu, B. Grosof, M. Dean, W. Grosso, and M. Musen, “Supporting rule system
interoperability on the semantic Web with SWRL,” Lect. Notes Comput. Sci. (including Subser. Lect. Notes Artif.
Intell. Lect. Notes Bioinformatics), vol. 3729 LNCS, pp. 974–986, 2005.!
[24] J. Pérez, M. Arenas, and C. Gutierrez, “Semantics and Complexity of SPARQL,” Semant. Web - ISWC 2006, vol.
4273, pp. 30–43, 2006.!
[25] T. R. Gruber, “Toward principles for the design of ontologies used for knowledge sharing?,” Int. J. Hum. Comput.
Stud., vol. 43, no. 5–6, pp. 907–928, Nov. 1995.!
[26] A. Razzaq, Z. Anwar, H. F. Ahmad, and K. Latif, “ScienceDirect Ontology for attack detection : An intelligent
approach to web application security,” Comput. Secur., vol. 45, pp. 124–146, 2014.!
[27] S. Melnik, P. Mitra, and S. Decker, “Framework for the semantic Web: an RDF tutorial,” IEEE Internet Comput.,
vol. 4, no. 6, pp. 68–73, 2000.!
[28] N. Bawany and N. Nouman, “A Step towards Better Understanding and Development of University Ontology in
Education Domain,” Res. J. Recent Sci., vol. 2, no. 10, pp. 57–60, 2013.!
[29] G. Antoniou, E. Franconi, and F. Van Harmelen, “Introduction to Semantic Web Ontology Languages,” Reason.
Web, pp. 1–21, 2005.!
[30] “OWL Web Ontology Language Reference.” [Online]. Available: http://www.w3.org/TR/owl-
ref/#CardinalityRestriction. [Accessed: 21-Dec-2015].!
[31] “Class.” [Online]. Available: http://www.infowebml.ws/rdf-owl/minCardinality.htm. [Accessed: 21-Dec-2015].!
[32] “Class.” [Online]. Available: http://www.infowebml.ws/rdf-owl/maxCardinality.htm. [Accessed: 21-Dec-2015].!
[33] M. Poveda-Villalón, M. C. Suárez-Figueroa, and A. Gómez-Pérez, “Validating Ontologies with OOPS ! State of the
Art,” Knowl. Eng. Knowl. Manag., pp. 267–281, 2012.!

!
[34] P. Ladkin, “Ontological Analysis,” Saf. Syst., vol. 14, no. 3, pp. 1–7, 2005.!
[35] Y. Tzitzikas, N. Spyratos, P. Constantopoulos, and A. Analyti, “Extended Faceted Ontologies,” Jornal, vol. volume,
no. issue, pp. 2–5, 2006.!
[36] P. Rajbhandari, R. Gosai, R. Shah, and K. Pramod, “Semantic Web in Medical Information Systems.,” Int. J.
Adv. …, vol. 5, no. 1, pp. 536–543, 2012.!
[37] S. Stephens, A. Morales, and M. Quinlan, “to Drug Safety Determination Determination,” IEEE Intell. Syst., vol. 21,
no. 1, pp. 82–86, 2006.!
[38] C. Matuszek, J. Cabral, M. Witbrock, and J. Deoliveira, “An introduction to the syntax and content of Cyc,” Proc.
2006 AAAI Spring Symp. Formaliz. Compil. Backgr. Knowl. Its Appl. to Knowl. Represent. Quest. Answering, vol.
3864, no. 1447, pp. 44–49, 2006.!
[39] F. B. Rogers, “Medical subject headings.,” Bull. Med. Libr. Assoc., vol. 51, pp. 114–116, 1963. !
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!
!

!
Appendix A:!
Sr.#
Title
Richness
Requirem
ent
Approach
Domain
Method
Sources
Formali
zation
(langua
ge)
Availability
Limitat
ions
Consisten
cy check
Mergin
g
suppor
t
1.
Drug
Abuse
Ontolog
y (DAO)
43 classes,
20
properties
Epidemiol
ogical
study of
prescriptio
n drug
abuse
practices
using
social
media
Top down
Lexical &
rule based
approach
1) Drugs
2)Narcot
ic
Data
Collecti
on,
Automat
ic
coding,
data
analysis
and
interpret
ation
1)Drug
Slang
2)
NIDA
3)
NDCP
4)DBpe
dia
5)Freeb
ase
6)Cyc[3
8]
7)MES
H[39]
OWL
✓
1)Lack
of
demogr
aphic
indicato
rs
2)unava
ilability
of
geograp
hical
informa
tion
✓
✓
2.
SimAR
C
-
Tackle
Alcohol
related
problems
Agent
Based
Simulation
1)Alcoho
l Abuse
A
particula
r
instance
of
compute
r
simulati
on,
called
Agent-
Based
Modelin
g
(ABM),
allows
building
artificial
societies
from the
bottom-
up;
whereby
individu
al
autonom
ous
agents
interact,
commun
icate
and
pursue
personal
-
-
✓
No
Limitati
on
-
-

!
goals
3.
Analyzi
ng
Twitter
data on
marijua
na
concent
rates
across
the U.S
-
-
Data
collection
using
twitter
streaming
API
Drug
Abuse
1)
Tweets
were
collecte
d using
Twitter's
streamin
g API
2)
Twitter
data
filtering
3)calcul
ation of
Raw and
adjusted
percenta
ges of
dabs-
related
tweets
4)
permutat
ion test
was
used
Twitter
-
-
-
-
-
4.
A Web-
Based
Study of
Self-
Treatme
nt of
Opioid
Withdra
wal
Sympto
ms with
Lopera
mide
-
-
Decision
approach
using
UGC
Drug
Abuse
1)Data
collectio
n
2)IRB
Procedu
res
3)Codin
g
4)Coder
reliabilit
y
assessm
ent
-
-
-
-
-
5.
Prescrip
tion
Drug
Abuse
(Google
play
store)
-
Comprehe
nsive
overview
covering
the
symptoms
and
treatment
of
prescriptio
n drug
abuse
-
Drug
Abuse
-
-
Android
-
1)Does
not
provide
slang
terms
2)
banned
drug
informa
tion is
not
provide
-
-

!
d
3)brand
names
of
drugs in
differen
t
countrie
s
6.
Drug
Addictio
n (
Google
Play
App)
-
It
provides
the
informatio
n of illicit
drug types
such as
heroin,
marihuana
, cocaine,
alcohol etc
-
Drug
Abuse
-
-
Android
-
Does
not
cover
any
propert
y of E-
DAO
-
-
7.
Drugs.c
om
(Google
play
app)
-
It provides
the
informatio
n of drugs
list along
with its
side
effects and
price
guide; it
also
provides
the
symptom
checker.
-
Drug
Abuse
-
-
Android
-
1)Does
not
provide
slang
terms
2)
banned
drug
informa
tion is
not
provide
d
3)brand
names
of
drugs in
differen
t
countrie
s
-
-
8.
Pharma
pedia
Pakista
n
(Google
play
app)
-
Provides
comprehe
nsive
informatio
n on drugs
brands
prices and
more.
-
Drug
Abuse
-
-
Android
-
1)
banned
drug
informa
tion is
not
provide
d
3)brand
names
of
drugs in
differen
-
-

!
t
countrie
s
!