Knowledge representation of drug using ontology alignment and mapping techniques

Drug searches are still based on drug names and brands, making it difficult for patients to come looking for a cure by saying that they feel sick. Likewise, when looking for drugs and information about their content to avoid overdose errors when changing drugs when drugs are supposed to be unavailable. Based on the issues raised, a study was conducted on applying semantic web ontology to search for drugs that can appear based on patients’ names, compositions, or complaints of diseases. Protégé 5.5 serves to represent drug information based on knowledge. The application uses Netbeans with Jena API as a library and creates data and drug information on the semantic web. Drug search also uses similar in-formation meaning based on user knowledge. By representing knowledge on the search for drug and disease information with semantic web ontology technology, it can meet the purpose of research, namely to improve drug and disease information search following the user’s wishes.


Introduction
The semantic web [1], [2] uses technology that allows computers to know the meaning of information based on metadata. Metadata can provide an understanding of essential information [3] for computers with search results that can be detailed and precise. The semantic web manages metadata using semantic data modeling techniques such as OWL (Web Ontology Language) and RDF (resource description framework) [4]- [8]. Information search has used the semantic web based on meaning, not just word, syntax, or tagging. Likewise, the data of drugs in pharmacies have the same name but different manufacturers or have other names but contain the same composition [9].
The Regulation of the Minister of Health of the Republic of Indonesia (Number 9 of 2017) states that pharmacies are a means and place for pharmaceutical services for pharmacists to practice. Using semantic web ontology technology in drug searches in pharmacies [10] is more 2 Science in Information Technology Letters ISSN 2722-4139 Vol. 2., No. 1, May 2022, pp. 1-8 profitable than using the web with a regular database because it is still based on syntax. Search results depend on words typed, not based on knowledge.
In contrast, the search results will display drug information with different names but the same composition and similarities as a semantic web application [11], [12]. Another problem is that the search can only be based on the name of the drug, composition, or manufacturer, whereas when a patient or customer comes, they often ask for the drug based on the disease they are suffering from, not the name of the drug. Based on the problems, research has been conducted on applying semantic web ontology to look for drugs [13]- [15]. The application can search and show the name, composition, or complaints suffered, and the use of protégés to represent drug information based on knowledge and create data or information about webbased semantic drugs. Previous research was conducted by Afifa [16], establishing a semantic web searching for cures and diseases. Drug and disease data is integrated from 3 websitesdrug data obtained from CSV format. The data is then converted to RDF ontology using CSV2RDF, which is most suitable for the Python platform [16]. Similar research has used the semantic web in an integrated information retrieval application, with a knowledge base of drug product information registered with BPOM, which aims to build an ontology-based knowledge model. Drug product data and information storage with an ontology model make searching easy [17]. In addition, other research references are the application of the semantic web in drug product search applications according to BPOM standards [17] and the implementation of web scraping on semantic-based ontology webs for drug and disease data [16].

Fundamental
Ontology [18]- [21] in artificial intelligence has two definitions, namely ontology as a vocabulary (vocabulary) representation and ontology as a core of knowledge or body of knowledge [21], [22]. The ontology allows knowledge that plays a role in connecting related resources. Everyday examples that can be applied using semantic web ontology are like the sentence "I like pie," so some people with different perspectives and knowledge or routines can produce different meanings or meanings. The pie is a cake or sweet dessert with various decorations for the general public and food lovers. For people who struggle with data and statistics daily, the pie is a round curve with its distribution, and people relate it for distribution. Based on technology, especially smartphones, someone thinks that pie is the name of the android nine series. It is an example that everyone has different knowledge.
RDF (Resource Description Framework) is a method for modeling information created by the W3C [18]- [20], [23] with the (primary) idea of creating a web resource statement using subject, predicate, and objects (after this referred to as N-Triple) [5], [23]- [25]. The subject in question is the subject of the resource to be described. Meanwhile, the predicate describes the behavior (characteristics) of the source and the subject and object relationship. RDF is an extensible way to represent web source information flexibly. It represents several components such as personal information, social networks, and metadata about digital artifacts such as music and images and integrates various information sources. The RDF standard query language offered developers and end-users a way to write and use query results across a specific range of information. In particular, RDF can publish and link data on the web [4]- [6], [22]. The RDF schema uses the notion of classes to define categories in classifying resources. The relationship between an instance and its class uses the type property. RDF schema can be used as a reference for creating class and subclass hierarchies and properties and subsets. Domains and constraints define a particular triple object's subject and object types [24]. XML (Extensible Markup Language) [21], [25] is a design language that makes it easy to send documents using the web. Unlike the Hypertext Markup Language (HTML), XML has custom tags for the user. XML can describe the class of data objects (XML documents) and the behavior of computer processing programs. XML is an application profile or a limited form of the Standard Generalized Markup Language (SGML), with a corresponding document [25]. Ontology Web Language (OWL) is one of the languages used to access documents on certain content.
OWL explicitly represents the meaning of terms in the vocabulary and their relationship, called ontology. OWL has more facilities for expressing meaning and semantics, such as XML, RDF, and RDFS. Thus, OWL surpasses these languages in its ability to represent machineinterpretable content on the web. OWL has been designed to meet the requirements of RDF, RDFS, and XML schemas [25]. Protégé [18] supports creating and editing one or more ontologies in a single workspace via a fully customizable user interface. Visualization tools allow for interactive navigation of ontology relationships. Available refactor operations include joining ontology, moving axioms between ontologies, renaming multiple entities, etc.

Research Design and Methodology
This research uses a quantitative research methodology in its design and preparation. The quantitative method was chosen because it was the most appropriate for data based on sample data used and processed back into the system. It was supported by some additional knowledge from brief questions and answers with relevant pharmacy officers and the processes carried out during the research process. According to Sugiyono, the quantitative research methodology steps are as follows [26]. The stages in the data collection process include literature study, observation, and interviews [27]. The data that has been collected is then analyzed and processed to obtain information. This study used the waterfall method with five steps: specification, conceptualization, formalization, implementation, and implementation [28]. Specifications relate to system development (why the system was created, its purpose, and its users) [29]. Conceptualization represents an intermediary with text or diagrams. Formalization transforms 'conceptual models' from domain experts to formal or semi-computing models, implemented by an ontology language (implementation).
Maintenance can be used for corrections and upgrades. Direct observations determine the pattern when a person buys drugs at a pharmacy. This activity is carried out for approximately two weeks for 4 hours daily. Patients can also confirm their complaints by visiting or consulting with a doctor on duty while visiting the doctor's office for routine checkups. Interviews were conducted with related people at the research site, namely interviews with two pharmacists, three assistant pharmacists, and one general practitioner. Information from interviews is used in developing the system.

Results and Discussion
This section describes the problem we solved with several troubleshooting approaches.
People have special knowledge about it. Based on that fact, they may have different views, use different datasets, and have different perceptions. People can have different perceptions (Perception of drugs_1, Perception of drugs_2, and Perceptions_N). Perception is used as a database in the form of data and stored in separate storage. Repositories (db1, db2, and dbN) contain different data, concepts, terms, and semantics. It depends on the group that sees reality (policymakers) and the people who create and store data (users who use technology). Computer users play an essential role in controlling and changing the terminology and semantics of data.
Everyone uses technology to search for data. It is hard for different groups to get the same results. Problems arise when people use data from other groups to share, collaborate, and find more effective solutions-the generation of ontologies from relational sources of drugs. We use ontology alignment and ontology matching techniques. The main contributions of the line of work introduced in this paper are the techniques that aim to improve the steps of creating the relational-to-ontology mapping. This study presents solutions based on different knowledge about the same reality (drugs) based on different perceptions. It uses a mechanism that works with a standard set of concepts, common terms, semantics, a common language, and a standard set of queries. Users can still use their ideas, terms, and perceptions to input system queries.
This study attempts to provide similar answers (outputs) based on the standard layer between different systems and users based on these solutions. Ontology Alignment creates a new ontology from two or more ontologies overlapping the common parts. The source ontology domain is different from the resulting ontology domain but is related. The standard part is a common word recognized and used with the same meaning by a different user. Example: Diabetic, Diabetic neuropathy, diabetic retinopathy, Diabetes mellitus, diabetes. Semantic Web Search Engines such as google accept queries in a format that varies from one tool to another [30]. The concomitant diseases, complaints, the name of the drug, and the manufacturer will also be different in the user's mind, all arranged in one unit of knowledge.
Example: Dizziness, Faintness, Giddiness, Unsteadiness, Vertigo, Lightheadedness, Wooziness. When a person with diabetes wants to buy a drug, he will say the symptoms are felt using many terms. Example: frequent urination, urination. Other symptoms of Diabetes, Examples: Tired, overworked, drained, exhausted, sleepy, and exasperated. Wide selection of Diabetes mellitus medications from doctors, Examples: Metformin (biguanide), Sulfonylurea, Meglitinide, Thiazolidinediones (glitazone), and Inhibitor DPP-4 (gliptin). Prefix below is one example of knowledge management with the same disease name but written in different terminology. Users and pharmacists can use both terminologies following the common term to search for a drug. Example: Diabetic and Diabetic Neuropathy ( Fig. 1 and Fig. 2). SPARQL is an RDF (semantic) query language used for databases. This query can manipulate (fetching and processing) data stored in the Resource Description Framework (RDF) format. SPARQL [3] has "key-value" data that allows the user to query something specific or data with the RDF specification of the W3C. Thus, the entire database is a threefold set of "subjectpredicate-object." It is analogous to the NoSQL database use of "key-value document." Dizziness and frequent urination are certainly not the only symptoms that occur only in people with diabetes. Other diseases may have symptoms such as diabetes; of course, the connectedness of knowledge is indispensable in searching for the right results. In this study, we have successfully created synonyms of terminology traceability for pharmaceuticals by using ontology alignment. The data we tested are shown in Table 1.  Gastric, gastrocolic, stomachic, sakit perut, melilit, mulas -√ Table 1 has synonyms in mentioning the name of the disease. There are five people in the pharmacy, and they will likely use different terminology to say the exact disease name. Table 1 is spoken in two languages, English and Bahasa Indonesia. Table 2 shows synonyms of symptoms of the disease. Not only frequent urination (Table 2), the symptoms of diabetes include increased thirst, tiredness, fatigue, drastic weight loss, itching around the genitals, slow healing wounds, and blurred vision. The same symptoms can appear in different diseases. We connect this knowledge by using relationships to form a complex understanding of diseases and symptoms, in contrast to the expert system where the symptoms must be the same. Someone will use the semantic web and start searching: "How I Fought Neuropathy." The system will search for the connectedness of the symptoms already present in this ontology. Knowledge in this ontology includes Diabetes symptoms: Random blood sugar (RBS): Measurement of sugar levels using a random blood sample and fasting blood sugar (FBS): Measurement of blood glucose levels by fasting for at least 8 hours. Post-prandial blood sugar (PPBS): Measurement of blood glucose levels after 2 hours of food intake. Glycated hemoglobin test: measuring average blood glucose over 2-3 months. Oral glucose tolerance test (OGTT): measurement of blood glucose levels before and 2 hours after consuming a glucose solution. Urine test: detects the excess amount of glucose excreted by the body (Fig. 3).  Drugs are not yet available to cure diabetes. The first step in controlling blood sugar levels is a healthy diet and regular exercise. Available medications are used to manage specific conditions, which work differently. Some are used to increase insulin production, while some can increase insulin resistance. Implementing an information system is usually only an information provider, while an ontology can search based on knowledge.  Table 3 shows the information of searching, such as Biguanide: helps in sensitizing body tissues to insulin. Sulphonylureas: increase insulin production in the body. Alpha-glucosidase inhibitors: Delay the absorption of carbohydrates. Thiazolidinediones: Increases sensitivity to insulin. It has many side effects, so it takes a prescription to get it. DPP-4 inhibitors: Help lower blood sugar levels without side effects. SGLT2 Inhibitors: Prevents the absorption of excess glucose in the kidneys. In effect, the urine will excrete excessive levels of sugar. In addition, recurrent urinary tract infections increase urination. Insulin injection is helpful for efficiently managing sugar levels. In its use, it is necessary to prescribe if the sugar level is long or very high, and there are many types of insulin. With ontology alignment, symptoms, diagnoses, and treatments are connected in a complex relationship and knowledge of medicine and disease. This system displays not only information but knowledge.

Conclusion
In this study, we used ontology alignment to solve differences in terminology when consumers buy drugs in pharmacies. The drug is referred to as the same disease, often using different terms and naming: Diabetes and diabetics. Through ontology, alignment between systems allows sharing of conceptualizations, terminology, and meanings. Ontology alignment is the process of determining the correspondence between concepts. New ontologies are obtained from the commonly overlapping sections of two or more ontologies. Using a standard set of terms (general ontology) is one way to make the system better understand to process information. It is expected that this research will provide additional knowledge for developing drug search engines to accommodate differences in knowledge from consumers. By representing knowledge about drug and disease information based on semantic web ontology technology, it can meet the purpose of research, which is to facilitate the search for drug and disease information following the user's wishes. The use of semantic web ontology technology in search of drug and disease information can improve the effectiveness of the search. The significance of a search depends on how it represents data and information in knowledge representation.