User centered ontology for Sri Lankan farmers

Ecological Informatics 26 (2015) 140–150

Contents lists available at ScienceDirect

Ecological Informatics journal homepage: www.elsevier.com/locate/ecolinf

User centered ontology for Sri Lankan farmers Anusha Indika Walisadeera a,b,c,⁎, Athula Ginige a, Gihan Nilendra Wikramanayake b a b c

School of Computing, Engineering & Mathematics, University of Western Sydney, Parramatta Campus, NSW, Australia University of Colombo School of Computing, Colombo 07, Sri Lanka University of Ruhuna, Matara, Sri Lanka

a r t i c l e

i n f o

Article history: Received 3 January 2014 Received in revised form 20 July 2014 Accepted 21 July 2014 Available online 1 August 2014 Keywords: Agricultural information/knowledge Contextual information Knowledge representation Ontology Ontology development

a b s t r a c t Farmers in Sri Lanka are badly affected by not being able to get vital information required to support their farming activities in a timely manner. Some of the required information can be found in government websites, agriculture department leaflets, and through radio and television programs on agriculture. This knowledge is not reaching the farmers due to its unstructured, incomplete, varied formats, and lack of targeted delivery methods. Thus finding the right information within the context in which information is required in a timely manner is a challenge. The information and knowledge needs to be provided not only in a structured and complete way, but also in a contextspecific manner. For instance, farmers need agricultural information within the context of location of their farm land, their economic condition, their interest and beliefs, and available agricultural equipment. To investigate some of the underlying farmer centric research challenges an International Collaborative Research Project to develop mobile based information systems for people in developing countries has been launched. Farmer centered ontology was developed as part of this project. Agricultural information has strong local characteristics in relation to climate, culture, history, languages, and local plant varieties. These local characteristics as well as the need to provide information in a context-specific manner made us develop this user centered ontology for Sri Lankan farmers. Because of the complex nature of the relationships among various concepts we selected an ontological approach that supports description logic to create the knowledge repository. For this we developed a new approach to model the domain knowledge to meet particular access requirements of the farmers in Sri Lanka. Through this approach, we have investigated how to create a knowledge repository of agricultural information to respond to user queries taking into account the context in which information is needed by farmers at various stages of the farming life cycle. The Delphi Method and the OOPS! (web-based tool) were used to validate the ontology. Initial system was trialed with a group of farmers in Sri Lanka. The online knowledge base with a SPARQL endpoint was created to share and reuse the domain knowledge that can be queried based on farmer context. © 2014 Elsevier B.V. All rights reserved.

1. Introduction In many developing countries, agriculture plays a major role in the country's economy; Sri Lanka is no exception. The agriculture sector in Sri Lanka is the main source of livelihood for the rural population, which accounts for 70% of the total population. Often we hear news about farmers in Sri Lanka not being able to sell their harvest due to oversupply, not getting the planned harvest, selecting the wrong seed types, or not being aware of market prices (Hettiarachchi, 2011). These problems badly affect the farmers. A major contributing factor is lack of necessary information at the right time. From time to time farmers need information such as accurate market prices, current supply and demand, seasonal weather, best varieties (or cultivars) and seeds, fertilizers and pesticides, information on pest and diseases and their control methods, harvesting and post-harvesting methods, and information on farming machinery and practices, to make ⁎ Corresponding author. Tel.: +61 0470113551. E-mail address: [email protected] (A.I. Walisadeera).

http://dx.doi.org/10.1016/j.ecoinf.2014.07.008 1574-9541/© 2014 Elsevier B.V. All rights reserved.

informed decisions at various stages of the farming life cycle (De Silva et al., 2012; Lokanathan and Kapugama, 2012). Some of this information is available from government websites (DOA, 2013; MOA, 2013), leaflets, and mass media in several different formats (i.e. text, audio and video). Sometimes different terminologies to express the same concept have been used. Due to its unstructured, incomplete, and varied formats, general nature of information, and lack of appropriate delivery methods this knowledge is not reaching the farmers. Thus there are many issues to be investigated to achieve a successful delivery of information from agricultural experts to rural farmers. Some important issues are: what information is required, what should be the delivery methods, and how to customize the information to meet the needs of farmers in different regions. By addressing the above issues the flow of information in the agriculture sector can be strengthened which will result in farmers being able to achieve better outcomes for their effort. This will contribute towards the overall economic development of the farming industry as well as the country. Glendenning et al. (2010) have shown the importance of contextualized information and knowledge for the farmers in India. They further

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150

explained how effective this knowledge on their productivity and income since this information is more relevant to their farm enterprises and better reflects the needs of the farmers. Thus they have recommended that context-specific and relevant information should be considered when developing approaches for farmers as an agricultural extension. According to the above analysis, we have identified that, not having an agricultural knowledge repository that can be easily accessed by farmers within their own context is a major problem because farmers need agricultural information relevant to their situation such as the location of their farm land, their economic condition, their interest and belief, need and available equipment. Such information would be more relevant and appropriate to farmers' needs and also could have a greater impact on their decision-making process. Social Life Networks for the Middle of the Pyramid (www.sln4mop. org) is an International Collaborative research project aiming to develop mobile based information system to support livelihood activities of people in developing countries (Ginige, 2013). Our research work is part of the Social Life Network project, aiming to provide agricultural information and knowledge to farmers based on their own context. The term context is treated in different ways in the literature (Brown et al., 1997; Dey, 1988; Dey and Abowd, 2000). One such definition is as follows (Dey and Abowd, 2000): “Context is any information that can be used to characterize the situation of an entity. An entity is a person, place or object that is considered relevant to the interaction between a user and an application, including the user and applications themselves”. This definition describes context clearly and generally as it can be used to describe the situation of a participant in meaningful way. We first developed an ontological approach to represent the necessary agricultural information and knowledge within the farmers' context (Walisadeera et al., 2013a). Using this approach we designed the ontology to include information needs identified for the first stage of farming life cycle (Walisadeera et al., 2013b). Next we extended the ontology to include events associated with the farming life cycle such as fertilizers, growing problems and their control methods (Walisadeera et al., 2014). This paper presents a revised and enhanced version of previous work and the new work done to create an end-to-end system that includes creation of an online knowledge base and an information retrieval interface. It further discusses methodologies that were used for designing, technology selection for implementation, validation and evaluation techniques. The remainder of the paper is organized as follows. Section 2 describes the modeling of farmer context for this application. Section 3 presents related research in this field. Section 4 introduces design approach of the ontology for the farmers in Sri Lanka and also presents a generalization of the approach that evolved from this work. Section 5 provides a summary of ontology validation and evaluation techniques used to test the ontology. Finally, Section 6 concludes the paper and describes the future directions. 2. Contextual information Our starting point was to gain a better understanding of information contexts specific to the farmers in Sri Lanka. Our main target groups are farmers and people associated with agriculture in Sri Lanka such as researchers, agricultural officers and instructors, and information specialists. To identify farmers' needs, farmer's context, and domain knowledge clearly, we have extracted domain specific knowledge using the following reliable knowledge sources: • Subject matter experts from the Faculty of Agriculture, University of Ruhuna, Sri Lanka; agricultural offices in Labuduwa (Agriculture Research Station), Sri Lanka; and National Science Foundation, Colombo, Sri Lanka (by unstructured interviews and group discussions);

141

• One research group working on the Social Life Network project conducted two surveys, interviewed 5 agricultural officers and 25 farmers (farmers in Dambulla area in Sri Lanka) to identify issues faced by farmers at different stages of farming life cycle (De Silva et al., 2012); • The Crop Knowledge Database created by the Agriculture Department in Sri Lanka; • Research articles (De Silva et al., 2013; Kawtrakul, 2012; Lokanathan and Kapugama, 2012; Narula and Nainwal, 2010), Fertilizer use by crop in India (2005); and books (Decoteau, 2000); • Published authoritative online data sources (the Department of Agriculture — Sri Lanka (DOA, 2013), the Ministry of Agriculture — Sri Lanka (MOA, 2013), and NAVAGOVIYA (2013)); • Mass media (newspapers, radio and television) and meteorological data. By analyzing information gathered from various sources, we have identified what information is required by the farmers at various stages to make better decisions. The required information can be broadly divided into two types: static that changes very slowly over time and dynamic that rapidly changes — sometime on a daily or hourly basis. The required static information includes varieties and seed types with corresponding properties, fertilizers, pesticides, weather patterns, soil factors, disease management, and post-harvest issues and management. The dynamic information consists of market prices, consumer behavior and demand, information related to places to buy and sell products and services, and information on what other farmers grow in different regions. As a result of this analysis, information important to farmers was identified in the form of questions. Some examples are given in Table 1. When responding to these questions several factors need to be considered. Thus response can vary from farmer to farmer. We have identified these factors that can be grouped into four major categories; • Farm environment: information about environment based on location of farm such as elevation, rainfall, climate zone, temperature, humidity, sunlight, wind, soil, etc. • Types of farmers: farmers are classified based on size of the cultivated farm land and estimated budget for cultivation. There are two main categories: garden farmers and commercial farmers. Commercial farmers can be further categorized as small-scale farmers, mediumscale farmers, and large-scale farmers. For example, when applying the fertilizers, information varies based on size of the farm land, budget, and number of employees. • Preferences of farmer: farmers have their own preferences such as high yielding crops and varieties, preferred control methods (chemical, cultural, and biological) and fertilizers (organic, chemical, or bio), low labor cost crops, high disease and insect resistance crops and varieties, desired farming systems (shifting cultivation and bush cultivation) and techniques, etc. • Farming stages: required information varies based on different stages of the farming life cycle. We have reviewed existing farming stages to identify suitable farming stages for our application. Table 1 Farmers' Information Needs. Farmers' information needs What are the suitable crops to grow? What are the best varieties (or cultivars)? What are the best fertilizers for selected crops and in what quantities? When is the appropriate time to apply fertilizer? What are the types of pests or crop diseases? How to protect crops from diseases? Which are the most suitable control methods to a particular disease? What are the symptoms of a specific disease? What are the best techniques for harvesting? What are the important factors to maintain quality of harvested crops? Which post-harvest method is best for a particular crop? What are the crops cultivated by other farmers and in what quantities?

142

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150

Lokanathan and Kapugama (2012) have categorized the information needs of farmers based on six stages of a crop cycle identified in De Silva and Rathnadiwakara (2010). It includes Deciding stage, Seeding stage, Preparing and Planting stage, Growing stage, Harvesting, packing and storing stage, and Selling stage. Information needs of farmers are grouped by Narula and Nainwal (2010) based on four stages from sowing the seeds to selling such as: Pre-sowing, Pre-harvest, Post-harvest, and Market Information. By analyzing each of the farming stages of the above studies, we have recognized that, according to the farmers' information needs identified in this section, the above two classifications of farming stages do not totally fit for our application. Selecting proper crops and varieties is paramount for successful farming. Therefore in our application, crop selection is selected as the first farming stage. According to the farmers' needs, post-harvesting is also an important factor. We therefore have included Post-harvesting as a separate stage because this information can help farmers to reduce their post-harvest losses. To improve overall decision-making in farming, we specially have identified other farming stages according to following basis using the information needs identified earlier: • by covering all required information needed by farmers; • by placing right information at the right stage. Fig. 1 shows the identified farming stages from crop selection to selling stage. • Crop selection: most important decision is selecting which crops to grow. Crop selection is a complex process because it depends on many factors. The environmental factors mostly affected this selection. Features of a crop (color, size, shape, flavor, and hardiness), farmer preferences, available resources, and market demand are other key determinants for this decision. • Pre-sowing: refers to planning different activities related to growing the selected crop(s). At this stage farmer needs information on quality agricultural inputs such as seed rate (seed quantity based on size of the farm land or number of plants according to the land size), required plant nutrients/growth regulators and fertilizers, types of irrigation facilities needed for the selected crop(s), best planting methods and new techniques for field preparation. • Growing: includes information related to planting and managing the crop during the growth phase. Information on good agriculture

Crop Selection Pre-Sowing Growing Harvesting Post-Harvesting Selling Fig. 1. Farming stages.

practices (traditional practices and new technologies), common growing problems (diseases, weeds, insects, and nematode pests), and their symptoms, methods for management of growing problems such as chemical, cultural, and biological methods are required at this stage. • Harvesting: at this stage, farmer needs information related to harvesting such as maturity time, methods and techniques of harvesting, expected average yield, and labor cost. • Post-harvesting: refers to proper handling after harvesting. Required information includes post-harvesting issues and management, packaging, grading, storing, standardization, transportation, and value added products that can be prepared from the harvested crop. • Selling: refers to preparation for selling. Mainly includes information related to the market such as market prices, consumer behavior and demand, and alternative marketing channels. The nature of information required by farmers very much depends on what stage they are in the farming life cycle. For instance, when selecting crops, farmers are more interested to receive extra information about “diseases and its control methods”. In this stage, we can recommend to choose suitable varieties (or cultivars) as a control method for a particular disease. In the crop selection stage, farmers can select this variety to avoid a particular disease. If the farmers have actually faced a particular disease then farmers have gone beyond choosing a variety and now in the growing stage. In this stage we want to provide suitable disease control methods to manage the specific attack, for example apply suitable fungicide sprays. This analysis suggests that, farm environment, types of farmers, farmers' preferences, and farming stages are the important factors that need to be considered when delivering agricultural information and knowledge to farmers. These form the context model related to the farmers for Social Life Networks. If we provide agricultural information according to identified context model, then this information is more relevant and will satisfy farmers' information needs. For this we need to reorganize agricultural information and knowledge for farmers based on the identified model. 3. Related work During the literature review we have found that many agricultural information systems have been developed to assist farmers' in decision-making and problem-solving by providing the required information or knowledge through computers and mobile phones. Some systems include information about weather forecast, pest and diseases, and systems like Fertilizer Expert which provides the tailor-made fertilizer to farmers (Sriswasdi et al., 2008). We can see that, these systems address only a particular problem and also the focus is not holistic. There are few ontologies in the domain of agriculture, such as Thai Rice ontology (Thunkijjanukij et al., 2009) and Soil Science ontology (Heeptaisong and Srivihok, 2010). Thai Rice ontology covers the domain of rice production from cultivation to harvesting in Thailand (Thunkijjanukij et al., 2009). This is a prototype ontology developed for plant production using Thai rice. Since existing, research knowledge repositories for plant production are not well organized; research policy administrators and researchers face many problems in finding relevant previous studies for research and development. Therefore, Thai Rice ontology has been designed in a manner to facilitate the process of knowledge acquisition and information retrieval for research purposes. In agriculture domain, there are many well-established and authoritative controlled vocabularies. Thesaurus can be interpreted as a simple ontology and it is useful to build domain ontology. Fisseha (2002) has identified several limitations and drawbacks with current vocabularies such as semantic ambiguity in definitions and usage of vocabularies; lack of high-level cross-domain concepts; and meaning of their relationships not being precisely defined. One of the most well-established and authoritative controlled vocabulary in agriculture is the AGROVOC

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150

143

together those elements from the global repository that participate to a given goal. In our application, each knowledge module has related information and knowledge to answer the farmers' information needs. For example, crop module has information about crops and fertilizer module has fertilizer information and knowledge to handle the fertilizer knowledge needed by farmers. Next we identified the relationships among them. Fig. 2 shows the generic crop knowledge module. To represent agricultural knowledge within the farmers' context we need to associate this generic crop knowledge with characteristics that describes the farmers' context. This modularization also helps us to reduce the complexity of real-world scenario in the application domain. We organized the farmers' list of information requirements according to the farming life cycle stages. We begin our detail design process with the first question in the list; “What are the suitable crops to grow?” During initial interviews with vegetable farmers they also identified this as a very important question. Choosing the best crop for individual situations is difficult since one has to consider many factors such as environmental conditions which can vary based on region and time period, preferences of farmer, and resources available for them for cultivation. We therefore have reviewed existing literature on crop selection to identify a suitable criterion which can be used to make better decisions. According to the Decoteau's (2000), the crop selection especially for vegetable crops depends on four considerations: Crop Consideration, Farmer Consideration, Labor Consideration and Marketing Consideration. This criterion is designed only for vegetable crops and for farmers in developed countries. Therefore this criterion is not a good fit for our application. Bareja (2012) has identified following as major crop selection factors for successful farming; Farm conditions, Crop or varieties adaptability, Available technology, Marketability and profitability, Resistance to pests and diseases, Farming systems and Security. This criterion is designed for selection of multiple crops and varieties. However, it has not considered the important factors such as labor cost and the farmer types. We next reviewed factors described in the NAVAGOVIYA (2013) web site which is one of the important web sources in agriculture domain in Sri Lanka for selecting suitable crops. These factors are Climatic requirements, Soil properties, Growing season, Labor availability and cost, Raw material availability and Market demand. In our application, the environmental conditions were given the first priority as it has been identified as the most important factor in the above analysis. Also these conditions cannot be controlled by the farmers. Next from the crops that meet the environmental conditions farmers can choose the best varieties by considering factors such as high yielding varieties, the special characteristics of a crop, maturity and disease resistance (Decoteau, 2000). These factors come under farmer consideration because farmer can decide importance of each of the factors according to their interests. Based on various crop selection

Thesaurus of the Food and Agriculture Organization (FAO) (2014). Even though the AGROVOC is organized hierarchically; we cannot query AGROVOC in context since it is not organized to satisfy the farmers' needs. Bansal and Malik (2011) are proposing an agricultural ontology for the crop production cycle to provide relevant information to farmers based on AGROVOC vocabulary in Semantic Web. The aim of this work seems to help individual farmers to get relevant and contextual agricultural information by searching concepts on the user interfaces. However, it cannot be seen as presenting information in context and moreover there is no clear design approach presented to describe how to represent information in context. Organizing information so that it can be queried in a context-specific way is more resource intensive as it requires procedures, methods, staff, and professional expertise to provide this information which is required at the farm level (Garforth et al., 2003). Based on the literature review, we can see that the existing vocabularies, information systems, and ontologies in the domain of agriculture are crop-specific and general thus not specific enough to satisfy the farmers' needs for timely information in context. We did not come across any agricultural information system or ontology that has been developed to represent the agriculture domain knowledge in the context of farmer information needs. Having discovered this research gap we have focused our attention to find an approach to develop an agricultural artifact for the farmers (farmer centric model in the domain of agriculture) to represent farmers' information needs in identified context. The term artifact is broadly defined in Information Systems research as construct (vocabulary and symbols), model (abstractions and representations), method (algorithms and practices), and instantiation (implemented and prototype systems) (Hevner et al., 2004). 4. Design approach As an initial step, our study focuses only on the static information to represent the knowledge which is important for the farmers to manage their farming activities. We took this decision because the static information represents the data that rarely change over time while dynamic information such as market prices changes frequently and hard to obtain without an elaborate network to gather market data in real time. Firstly we identified areas of generic crop knowledge required to answer the farmers' information needs (see Table 1). We have called these broad areas of knowledge as “knowledge modules”. The generic crop knowledge consists of modules such as nursery management, harvesting, post-harvesting, growing problems, control methods, fertilizer, environmental factors, crops and basic characteristics of crops, variety, etc. According to the description of Parent and Spaccapietra (2009), knowledge module is expected to show a similar unit of purpose, gluing

Variety Nursery Management

Basic Characteristics hasCropCharacteristics

Environmental Factors

hasVariety

uses

dependsOn involves

contains

hasHarvestInfor

Crop isAffectedBy

Post Harvesting

Growing Problems

Control Methods

Harvesting

hasPostHarvestTechniques

Growing Practices

isControlledBy

Fertilizer

hasSymptom

Symptoms

Fig. 2. Generic crop knowledge module.

144

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150

criteria reviewed earlier we can see that only a few preferences have been considered. However, we have included wide range of preferences because this will help farmers to make better decisions. According to the collected information through the interviews with farmers and agricultural specialist in Sri Lanka, information about what other farmers grow in different regions and its quantities is also an important factor when selecting crops because from this information farmers can get an idea about whether there is going to be an oversupply or not. Farmers also consider the market information when selecting crops; therefore, we take the market information as a final consideration of our crop selection criteria. In the initial stage we did not consider the security aspect since it is not a noticeable factor for farmers in Sri Lanka but appropriate security with access rights would be considered at a later stage. Finally farmers select the suitable crops and varieties by considering all the necessary factors according to their needs. The following Table 2 provides a summary of the crop selection criteria reviewed above. We have defined a crop selection module (see Fig. 3(a)) based on our crop selection criteria to deliver information and knowledge related to crop selection stage. Environment, Crop, Variety and Basic Characteristics are the same modules identified in the generic crop knowledge module. The knowledge modules such as Farmer, Farmer Types, Preferences and Market Information are additional modules needed for crop selection. Fig. 3(a) shows that the farmer is the central concept and this very much motivated us in selecting a farmer centered approach to develop our design artifact. In here environment is designed with regard to farm environment and crop environment. In this stage we do not consider market information and other farmers' information as factors affecting crop selection because our initial effort was focused only on the static information. In similar way, we can identify the criteria for each item in the list of farmer information requirements. For example, we defined the criteria for applying fertilizers (see Fig. 3(b)) based on related literature (used knowledge sources in Section 2) to deliver fertilizer knowledge related to second stage of the farming life cycle. When applying a fertilizer for a specific crop a farmer needs to know the fertilizer quantity and its unit. A fertilizer quantity depends on many factors; especially it depends on the location, water source, soil Ph range, time of application, application method, and fertilizer type. Here we also include specific sources of a fertilizer such as nitrogen, phosphorus, potassium, etc. Thus fertilizer quantity needs to be specified in relation to all these information. To do that, we introduced a new information module; Fertilizer Event to represent this additional information and new relationships to describe this event. The details of modeling the events associated with the farming life cycle and the associated challenges are outside the scope of this paper and are explained in Walisadeera et al. (2014). 4.1. Contextualized information The next step is formulation of a set of contextualized (or personalized) information based on the farmers' information needs. For this Table 2 Summary of crop selection criteria. Criteria (factors)

Environment Farmer types Labor Crop characteristics Market information Farmer preferences Security Other farmers' information

Different sources Decoteau (2000)

Bareja (2012)

NAVAGOVIYA (2013)

Our criteria

√ √ √ √ √ Limited range × ×

√ × × × √ Limited range √ ×

√ × √ × √ Limited range × ×

√ √ √ √ √ Wide range × √

we had to develop our own approach to formulate the contextualized information. With the help of the domain experts we first identified the breadth of information required by farmers. Next based on earlier identified user context we identified the conditions we can use to obtain a subset of information that can satisfy a specific information need of users. Based on this, we expanded the questions in the farmer information need list to include the farmer context as follows: • included all information/knowledge of the farming stages (crop selection to selling stage) and its constraints (restrictions) — it represents information/knowledge needed by the farmers • included farmers' conditions based on the identified context model — it provides information/knowledge in user context. Fig. 4 shows our basis for formulating contextualized information. We can see from the above that the formulation of contextualized information for crop selection depends on multiple criteria such as the farmers' context, general crop knowledge, crop selection criteria (select suitable task modeling criteria such as criteria for selecting crops, applying fertilizers, selecting control methods, etc.) and the farmers' constraints. This serves as a basis for formulating information in a user context for our application. Some examples of contextualized information related to each category of crop selection and fertilizer applying are given in Table 3. We have identified the farmer constrains based on each criterion factor. For example under the Environmental factor (identified as a criterion factor for crop selection) we need to select suitable crops based on different locations, different seasons, different soil factors, etc. or combination of these constraints. We have identified these different constraints related to this application, for example, the location as Zone, Agro Zone, Elevation based location and so on. Through this process we have formulated the contextualized questions covering all constrains relevant to each criterion. We also generalized these questions. These are the range of questions that our designed artifact should be able to answer. When we are providing specific answers to these questions, we also need to provide additional information (background knowledge) related to the questions. For example necessary environmental conditions and application methods which are relevant to applying fertilizer, because these additional information will also help farmers to make better decisions. 4.2. Why ontology? Next we need to identify an optimum way to organize the information and knowledge in context. There are many methods to organize information, for instance using relational data models, web pages, etc. However, these methods of information organization are not effective to answer the ranges of questions identified in Table 3. To answer these types of questions we need more expressive relationships among concepts to be able to effectively represent the knowledge, for example to express semantic representation of the farmers' location such as Agro-ecology Zone, Climatic Zone, Elevation based location, Province, District, and Regional Area. An ontology is an explicit specification of a conceptualization (Gruber, 1993). It provides a structured view of domain knowledge and acts as a repository of concepts in the domain. This structured view is essential to facilitate knowledge sharing, knowledge aggregation, information retrieval, and question answering (Gruber, 1995). In addition, ontology provides the means of deduction capabilities provided by an inference mechanism and reasoning support in order to generate further knowledge (i.e. not explicitly known but can be deduced based on the existing knowledge of the domain) (Fox et al., 1996). Thus, ontology represents a better data model (richer knowledge) than a relational data model. To represent the domain knowledge in user context, we needed an approach which relies on the expressive features to represent appropriate description of the context. For this purpose ontology based approaches are better suited compared to

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150

145

a) Farmer Types

Farmer Preferences

hasFarmerType

Environmental Factors

hasPreferences

Market Information

Farmer needs

hasFarmEnvironment dependsOn

Basic Characteristics

Crop

Variety

hasCropCharacteristics

b)

hasVariety

FertilizerType

uses

Crop

Fertilizer Source

hasFertilizerEvent

Unit

hasFertilizer hasFertilizerQuantity

hasFertilizerUnit

FertilizerEvent

Quantity hasSoilPhValue

hasApplicationMethod

SoilPhValue

ApplicationMethod

hasWaterSource hasTimeOfApplication

hasLocation

TimeOfApplication

hasSp.Information

WaterSource Location

SpecialInformation

Fig 3. (a) Criteria for selecting crops. (b) Criteria for applying fertilizers.

rule based approaches (Toninelli et al., 2005). We therefore use an ontological approach to represent contextualized knowledge that can be used to find a response to queries within a specified context in agriculture domain. We reviewed ontology development methodologies and techniques to identify a suitable ontology development approach. Grüninger and Fox (1995) have published a formal approach to design ontology while providing a framework for evaluating the adequacy of the developed ontology. Its main strength is high degree of formality and focuses on building ontology based on first-order logic (FOL) by providing strong semantics. Being a formal ontology it is structurally and functionally rich enough for the description of the domain knowledge in context. It also provides a mechanism to address the drawbacks of terminological ambiguity in domain by defining rigorous, scientific, and meaningful terms. We therefore selected Grüninger and Fox's methodology, a FOL based approach to develop a farmer centric ontology for agriculture domain.

Our ontology creation begins with the definition of a set of farmers' information needs identified in Section 2. We take these farmers' information needs as the main motivation scenario of our application to provide information in context. Competency questions (CQs) determine the scope of the ontology and use to identify the contents of the ontology. The ontology should be able to represent the CQs using its terminologies, axioms and definitions. Then, a knowledge base based on the ontology can provide answers to these questions (Grüninger and Fox, 1995). These questions are benchmarks in the sense that the ontology is necessary and sufficient to satisfy the requirements specified by the CQs (Fox et al., 1996). Therefore, formulation of the CQs is a very important step because these questions guide the development of the ontology. In our application, the contextualized information (see Table 3) has been used as the CQs to develop the ontology because it satisfies the expressiveness and reasoning requirements of the ontology. 4.3. Identifying ontology components

Generic Crop Knowledge

Farmer Context Model

Additional Knowledge Modules

Task Modeling Criteria

Contextualized or Personalized Information Fig. 4. Basis for modeling contextualized information.

In order to answer the CQs, we need to identify the ontology components. There are three main ontology components; concepts, relationships and constraints (capture additional knowledge about the domain and it can be represented as axioms). In our ontology design, we use the middle-out strategy to identify the main concepts (Uschold and Gruninger, 1996). The main advantage of this approach is that it starts with most important concepts first. Once the higher level concepts are defined the specialized and generalized hierarchies get identified. Thus, these concepts are more likely to be stable. This results in less re-work and less overall effort. There are few concepts which we can directly elicit, for instance, we identified Crop as a main concept of our ontology based on scope of the

146

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150

Table 3 Farmers' information needs in context. Farmers' information needs

Farmers' information needs in context

Generalizing contextualized information

What are the suitable crops to grow?

Suitable crops based on the Environment: • Which crops are suitable to grow in the ‘LowCountryDryZone’ agro-ecology region? • Which crop's varieties are the most appropriate for ‘WetZone’ and ‘Maha’ season? • What are the suitable vegetable crops for ‘UpCountry’, applicable to the ‘Well-drained Loamy’ soil, and average rainfall N2000 mm? Suitable crops based on Preferences of Farmers: • What are the crops involving in high labor cost? • What Brinjal's varieties are good for the ‘Bacterial Wilt’ disease? Suitable crops based on Environment, Preferences and Other Information: • What is the best Brinjal's variety which is suitable for ‘DryZone’ and high-resistance to the ‘Bacterial Wilt’ disease? Suitable fertilizers based on the Environment: • What are the suitable fertilizers and in what quantities for Banana which are grown in Dry Zone? • Which fertilizers are the most appropriate for Chili under Rain-fed Condition? • What are the suitable fertilizers and in what quantities for farmers in Badulla district who cultivate Tomatoes? Suitable fertilizers based on Preferences of Farmers: • What are the suitable organic fertilizers which are used to Basal dressing for Tomato?

• Which crops are suitable to grow in specified Location (Agro Zone)? • Which Crop's varieties are the most appropriate for specified Location (Climatic Zone) and specified Season? • What are the suitable Types of Crops for specified Location (Elevation), applicable to the specified Soil types/characteristics, and Conditions (Rainfall or Temperature)? • What are the crops involving in Labor requirement/cost? • What Crop's varieties are good for the specified Disease? • What is the best Crop's variety which is suitable for specified Location (Climatic Zone) and Resistance conditions to the specified Disease?

What are the suitable fertilizers for selected crops and in what quantities?

ontology. Next, we need to identify other major concepts by analyzing each competency question. For example the main concept in the query which crops are suitable to grow in the ‘LowCountryDryZone’ agro-ecology region? (refer Table 3) is Zone. Once we identify the concept we need to define specialized and generalized (if necessary) hierarchies based on the following criteria: • concept properties, • nature of the instances (instances are used for denoting specific members of a concept and represented by constants or variables), • generic crop knowledge structure, and • the farmers' context To represent the information in context we need to identify the details of each concept in multiple levels. Therefore, designing this type of ontology is an extremely complex task. It needs to be done in a very systematic way. Since a variety is a group of crops that share common qualities of crops of the same species (Decoteau, 2000), we have identified a variety as a subset of a crop (i.e. variety is a subconcept of crop). Farmers need information related to their location. We have identified the location as zones, provinces, districts, regional areas and elevation based locations. Therefore we have introduced a concept as Location to represent those locations (generalization). Then the concept Location is a super concept of Zone, Province, District, RegionalArea, and Elevation concepts (taxonomic hierarchy). The concept Zone has properties such as maximum rainfall and minimum rainfall. By specializing Zone concept we have defined the concept AgroZone (agro-ecology zone) as a subclass of Zone, because there are several additional properties specific to AgroZone such as maximum and minimum temperature, and maximum and minimum elevation. The properties of concept Zone can be inherited by the AgroZone concept because of the taxonomic hierarchy (is_a relationship). AgroZone is a subclass of Zone if and only if every instance of AgroZone is also an instance of Zone (specialization). Then, AgroZone is also a subclass of Location. Based on the definition of the concept Zone, we can categorize the instances as DryZone, IntermediateZone and WetZone. If there is no further categorization (has only first level categorization) and/or need to restrict conditions then we can restrict this categorization as a property

• What are the suitable fertilizers and in what quantities for the Crops which are grown in specified Location (Climatic Zone)? • Which fertilizers are the most appropriate for Crops under specified Conditions? • What are the suitable fertilizers and in what quantities for farmers in specified Location (Districts) who cultivate specified Crops? • What are the suitable Types of fertilizers which are based on Method of application for specified Crops?

of a concept (e.g. ConceptType) to reduce complexity of the design of the ontology. This decision of data property identification depends on the application, for example on the type of information which we want to retrieve and kind of relationships between them. In our application, the Zone concept has three properties such as ZoneType, maximum and minimum rainfall. The Table 4 shows few concepts and their properties. If there is a more than one level categorization, we have organized this as a taxonomic hierarchy. The associative relationships (non-taxonomic hierarchy) are specified as follows: • identify the concepts and relationships with meaningful relations, • define the relationships and their inverse relationships (if applicable). For example, there is an associative relationship with inverse between Crop and Variety: Crop hasVariety Variety, Variety isVarietyOf Crop. Table 5 shows some associative relationships including their inverse. The set of relationships describes the semantics of the domain. Based on our analysis, the main properties of Crop are Crop Family, Hardiness, Nutrition, etc. Since Variety is a subclass of Crop these properties can be inherited. Other than these properties Variety has properties such as Length, Color, Shape, etc. We have paid special attention to properties specific to Variety because, when selecting crops farmers primarily consider basic features of a Variety and not just of a Crop. We have defined the environmental factors to be Sunlight, Wind, Humidity, WaterSource, etc., which are subclasses of the EnvironmentalFactor (superclass). In our application the union of Table 4 Concept properties. Concept

Properties

AgroZone

hasMaximumElevation, hasMinimumElevation, hasMaximumTemperature, hasMinimumTemperature hasCropType, hasCropFamily, hasHardiness, hasNutrition, hasValueAddedProduct, hasSpecialCropCharacteristics hasLength, hasColor, hasShape, hasFlavor, hasSize, hasQuality, hasWeight, hasDroughtResistanceRate hasFertilizerType, hasSource hasControlMethodType, hasApplicationStage

Crop Variety Fertilizer ControlMethod

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150 Table 5 Associative relationships with inverse.

147

Domain Knowledge from Reliable Knowledge Sources and Outcomes of Interviews with Domain Users and Domain Experts

Concept

Relationship

Concept

Crop Crop Crop Crop Season GorwingProblem

growsIn, grows isAffectedBy, affects dependsOn, isDependedOn hasGrowingSeason, isGrowingSeasonOf consists, isConsistedOf isControlledBy, controls

Location Disease EnvironmentalFactor Season Month ControlMethod

these subclasses form the environmental factors which need to be specified for instances of crops as well as for farms. In here we have used the subclasses as a set of mutually-disjoint classes which covers EnvironmentalFactor. Every instance of EnvironmentalFactor is an instance of exactly one of the subclasses in the union. Once the hierarchies and relationships have been identified, the next step is to define the informal CQs in a formal way using formal terminologies (see following example). Note that, here we use unary predicates for representing concepts, binary predicates for properties and binary relationships. The interpretation of C(x) is that x is an individual that belongs to concept C. CQ: What are the suitable organic fertilizers which are used as Basal dressing for Tomato? (∃x) (Fertilizer(x)) ∧ (Crop(Tomato)) ∧ (∃y) (FertilizerEvent(y)) ∧ hasFertilizerEvent(Tomato, y) ∧ hasFertilizer(y, x) ∧ hasApplicationMethodForFertilizerEvent(y, Basal_dressing) ∧ hasFertilizerType(x, Organic); The definitions of the terms, constrains and their interpretation related to the query are specified using a set of axioms in FOL. We have defined the axioms to express these definitions and constraints. For example, we have expressed uses relationship (see below) by using hasFertilizerEvent and hasFertilizer relationships through FertilizerEvent concept: (∃x)(Crop(x)) ∧ (∃y)(FertilizerEvent(y)) ∧ (∃z)(Fertilizer(z)) such that hasFertilizerEvent (x, y) ∧ hasFertilizer(y, z) → uses(x, z); The CQs drive the development of the ontology. The answers to the CQs can be retrieved from the knowledge base (i.e. ontology populated with instances). The implicit knowledge derived from the ontology by using inheritance and the FOL based axioms. We can use this ontology to organize domain knowledge by combining the farmers' context and can also deduce answers to queries based on the context. We also have reviewed the existing standard vocabularies and ontologies to reuse the concepts. For example we checked the concepts of AGROVOC, SWEET (Semantic Web for Earth and Environmental Terminology), CSIRO's phenonet (Agriculture Meteorology sensor network ontology), OWL Time ontology, NeoGEo and GEOSPARQL vocabularies which cover the geographical features and geometry. Since some ontologies are too complex and/or general their concepts are not enough to serve our requirements (context mismatch) for the EnvironmentalFactor, Time, and Location concepts of our ontology. However we can extend our ontology by designing it to interoperate with other ontologies such as AGROVOC. For example, once we have identified suitable crops to grow in specified conditions from our ontology, then we can get the crop related general information such as crop type and its hierarchy from the AGROVOC. 4.4. Generalizing approach We have now generalized the specific approach that was developed to create the farmer centric ontology for Social Life Networks. Fig. 5

Users’ Information Needs (Formally Referred to as Motivation Scenarios)

Generic Crop Knowledge (in Agriculture Domain)

Task Criteria Modeling

User Context Model

Contextualized (or Personalized) Information (Formally Referred to as Competency Questions)

Main Ontology Components

Specialization and Generalization

Axioms

Classification

Ontology Fig. 5. Ontology design framework.

shows this generalized approach. According to this approach, we first identify a set of questions (Users' Information Needs) that reflect various motivation scenarios. Next we create a model to represent information in user context (see Fig. 4). Then we derive the contextualized information incorporating user context and task modeling (the criteria for each task, for example selecting crops, applying fertilizers, and selecting control methods) with generic knowledge module. We refer this contextualized information (refer Table 3.) as the informal CQs. These CQs are used to identify the ontology components according to the Grüninger and Fox's methodology to develop the ontology. Developing an approach combining all this to reorganize the information so that it can be queried in the user context, is the new contribution we have made in this study. Using this framework, we can extend the ontology for different scenario problems. For example, when answering scenario question like “Which are the most suitable control methods to a particular disease?” we need to take into account suitable criteria for selecting control methods and farmers' context. We can then formulate the contextualized information based on this systematic approach. These questions drive the development of the ontology and can represent contextual information and knowledge to satisfy the user needs. 5. Ontology validation and evaluation 5.1. Ontology validation It is very important to check the validity of the ontology. In our application, two aspects need to be validated; correctness of the contents of the ontology and correctness of the construction of the ontology. After designing the ontology, the content of the ontology need to be validated by domain experts against the users' requirements. Otherwise, defects in the ontology will spread to subsequent design and implementation activities. The content correctness depends on definitions of concepts, relationships between concepts, hierarchical structures, concept properties,

148

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150

and information constraints of the ontology. The Delphi method is a research technique that is used to obtain the responses to a problem from a group of domain experts (Mattingley-Scott, 2013). We selected the Delphi method to obtain expert advice and responses to validate the content of the ontology. The validation process is done by agricultural experts by examining the correctness, relevancy and consistency of the ontology components and a set of predefined criteria. The structured paper-based questionnaires have been used for validation of the content correctness. The contents have been refined based on domain experts' feedbacks and comments. One approach for checking the correctness of the construction is to analyze whether the ontology contain anomalies or pitfalls (PovedaVillalón et al., 2012). We first identified the common pitfalls before the implementation, for example defining synonyms as concepts, defining wrong inverse relationships, recursive definition, misuse of part-of and subclass relationships, etc. Next we identified the types of Ontology Design Patterns (ODPs) that help to avoid the pitfalls by means of adapting or combining existing ODPs (Poveda et al., 2010). Design patterns are shared guidelines that help to solve design problems. Some examples are Semantic Web Best Practices and Development under W3C (2013), NeOn Modelling Components under NeOn project (Suárez-Figueroa et al., 2007). The reasoner attached to ontology also used to check logical inconsistency of the ontology. We also used the web-based tool called OOPS! (Poveda-Villalón et al., 2012) to detect potential pitfalls in the ontology. Using the above methods we validated the quality of our ontology. 5.2. Ontology evaluation After implementing the ontology, we need to examine the deficiencies of the ontology in use. We have used two techniques for this evaluation. Section 5.2.1 explains how we tested the usefulness of the ontology during the internal design process. Section 5.2.2 explains how we measured the effectiveness of the ontology using applicationbased approach. 5.2.1. Internal evaluation Creating ontology manually is a tedious and time-consuming task. There are many ontology development tools and languages available to build a new ontology. By doing comparative study of ontology languages and tools, we have selected protégé (Knublauch et al., 2004) as ontology development environment and Web Ontology Language (OWL) (Patel-Schneider et al., 2013) as ontology language. We also compared the operators and features of the conjunctive queries, Description Logics (DL), and FOL. We have identified that DL is more expressive than conjunctive queries but FOL offers a good expressive power in comparison to DL. However, DL is a fully decidable fragment of the FOL (Baader et al., 2008) and also reduces the complexity when compared with FOL. In our application, decidability is very important as we need to retrieve agricultural information in context. We therefore selected the DL based (OWL 2-DL) approach to implement our ontology. The implemented ontology using protégé is available at http://www. sln4mop.org/ontologies/2014/SLN_Ontology (Turtle version: http:// www.sln4mop.org/ontologies/2014/SLN_Ontology.ttl). It consists of 83 concepts, 191 object properties, and 45 data properties. Currently it has 23 vegetable crops, 8 fertilizers, 9 diseases, and 20 pesticide instances. This implementation is used to evaluate the ontological commitments internally and also used to test the consistency and inferences using reasoners (used FaCT++ reasoner plug-in with Protégé 4.2). We used the CQs to evaluate the ontological commitments to see whether the ontology meets the farmers' requirements using DL queries (DL expressions) and SPARQL queries (refer to information in Section 5.2.2.2). For this we used DL and SPARQL query facilities already available in Protégé environment to query the ontology. For example,

farmers are interested to know suitable organic fertilizers used for Basal dressing for Tomato. The related DL query is shown below.

The SPARQL query below is to find crops that grow in Up Country, applicable to Sandy Loamy soil, and the temperature ranges between 9 and 28 °C.

5.2.2. Application-based evaluation 5.2.2.1. Field testing. Next we checked the user satisfaction (i.e. utility) of our ontology. We used a mobile based application for this evaluation. A mobile based application was developed to provide information to farmers using this ontology (De Silva et al., 2013). The first evaluation was done only for a crop selection with a group of 32 farmers in Sri Lanka. The farmers used the application to select a crop that they want to grow, and they were asked the question “Is all information for the crop selection stage provided”. They recorded the answer on a 1 to 5 Likert scale; strongly agree, agree, moderately agree, disagree, and strongly disagree. The responses were 7% strongly agree, 57% agree and 36% moderately agree. The farmers also suggested few areas where they would like to get more information. We have gathered these requirements for our future designs. 5.2.2.2. Online knowledge base. The knowledge base based on the ontology was created by populating the ontology with instances. To share and reuse the agricultural information and knowledge we need to access the knowledge base via the Web. The online knowledge base can also be used for evaluation process. We first have converted the inferred model of the ontology into the Resource Description Framework (RDF) model. RDF is the World Wide Web Consortium (W3C) standard for representing and storing information on the Web (Klyne and Carroll, 2004). The RDF data is a collection of triples or statements of the form (Subject: resource being described, Predicate: named property, Object: property value). A set of such triples is called a RDF graph. Next we compared the features of Semantic Web toolkits for different programming languages (Bizer and Westphal, 2005) to identify a suitable toolkit to manage the RDF data. We have identified that ARC2 (appmosphere RDF classes) toolkit as a good Semantic Web framework written in PHP language which allows easily management of structured data in RDF (reviewed in PHP and Semantic Web (2013)). However ARC2 provides less functionality than RAP (RDF API for PHP) but have a better performance on core tasks such as parsing RDF/XML to an inmemory array (Oldakowski et al., 2005). ARC2 also provides an RDF storage with support for the SPARQL query language and various RDF parsers for numerous formats (RDF/XML, Turtle, RSS, microformats, eRDF, RDFa, etc.). The SPARQL is a standard query language developed primarily to query the RDF graphs (Prud'hommeaux and Seaborne, 2008). We have developed the online knowledge base (SPARQL endpoint) using the Semantic Web technologies (refer http://webe2.scem.uws.edu.au/ arc2/select.php). We can query the contextualized information on the Web via this application using SPARQL queries. For example, the following SPARQL

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150

query lists organic fertilizers used for Basal dressing for Tomato. We have evaluated the ontology by evaluating outputs of the given queries.

6. Conclusions and future work Farmers in Sri Lanka need necessary and relevant information to make optimal decisions for successful farming. Currently, not having agricultural knowledge repositories that can be easily accessed by farmers within their own context is a major problem. In this research project we have addressed this need. In this study, we identified the user context related to the farmers in Sri Lanka. Next we developed a logic based ontological approach to meet the information needs to suite the identified context. We have achieved this by modifying how contextualized information is formulated in a well-established methodology. Designing this type of ontology is not a simple task, because it depends on many factors as shown in Fig. 3(a) and (b). In this paper we have explained how we designed and developed the ontology to organize domain knowledge by meeting particular access requirements effectively using the framework shown in Fig. 5. The validation and evaluation have been done separately. We validated the ontology in terms of accuracy and quality by using the Delphi Method and the OOPS! web-based tool. We evaluated the ontology internally and externally by examining the deficiencies of the artifact in use. The online knowledge base with a SPARQL endpoint to share and reuse the domain knowledge was created. Knowledge organized in this manner can better assist the decision-making process. Since the dynamic information is also important for making decisions, we are planning to provide the dynamic information in future versions. We aim to further populate the knowledge base by involving people in agriculture community. For this we plan to develop a system to capture agricultural information via relational databases and importing it into the ontology. This ontology management system will help us to manage ontology in the long term. We received very valuable feedback from the field trials. Based on this feedback we are now refining the application. The first field trial was done only for the crop selection. A more detailed evaluation that will cover the second and third stages of the farming life cycle is being planned. Acknowledgments We acknowledge the financial assistance provided to carry out this research work by the HRD Program of the HETC project of the Ministry of Higher Education, Sri Lanka (RUH/O-Sci/N2) and the valuable assistance from other researchers working on the Social Life Network project. Assistance from the National Science Foundation (NTRP/2012/FS/ PG-01/P-02) to carry out the field visits is also acknowledged. References Baader, F., Calvanese, D., McGuinness, D.L., Nardi, D., Patel-Schneider, P.F. (Eds.), 2008. The Description Logics Handbook — Theory and Applications, second ed. Cambridge University Press. Bansal, N., Malik, S.K., 2011. A framework for agriculture ontology development in semantic web. International Conference on Communication Systems and Network Technologies. IEEE Computer Society, pp. 283–286. Bareja, B.G., 2012. Crop farming. http://www.cropsreview.com/index.html (Accessed 2012).

149

Bizer, C., Westphal, D., 2005. Developers guide to semantic web toolkits for different programming languages. http://www.wiwiss.fu-berlin.de/suhl/bizer/toolkits/. Brown, P.J., Bovey, J.D., Chen, X., 1997. Context-aware applications: from the laboratory to the marketplace. IEEE Pers. Commun. 4 (5), 58–64. De Silva, H., Rathnadiwakara, D., 2010. ICT policy for agriculture based on a transaction cost approach: some lessons from Sri Lanka. Int. J. ICT Res. Dev. Afr. (IJICTRDA) 1 (1), 51–64. De Silva, L.N.C., Goonetillake, J.S., Wikramanayake, G.N., Ginige, A., 2012. Towards using ICT to enhance flow of information to aid farmer sustainability in Sri Lanka. 23rd Australasian Conference on Information Systems (ACIS), Geelong, Australia, pp. 1–10. De Silva, L.N.C., Goonetillake, J.S., Wikramanayake, G.N., Ginige, A., 2013. Farmer response towards the initial agriculture information dissemination mobile prototype. In: Murgante, B., et al. (Eds.), 2013 International Conference on Computational Science and its Applications (ICCSA 2013)Part I, Lecture Notes in Computer Science. 7971. Springer-Verlag, Berlin Heidelberg, pp. 264–278. Decoteau, D.R., 2000. Vegetable Crops. Prentice-Hall, Upper Saddle River, New Jersey. Department of Agriculture (DOA), Sri Lanka, http://www.agridept.gov.lk [Accessed 2013]. Dey, A.K., 1988. Context-aware computing: the CyberDesk Project. AAAI Spring Symposium on Intelligent Environments. AAAI Press, Menlo Park, pp. 51–54. Dey, A.K., Abowd, G.D., 2000. Towards a better understanding of context and contextawareness. Workshop on the What, Who, Where, When and How of ContextAwareness, affiliated with the CHI 2000 Conf. on Human Factors in Computer Systems, New York. Fertilizer use by crop in India, 2005. Published by Food and Agricultural Organization of United Nations (FAO), Rome. Fisseha, F., 2002. Towards better Semantic Standards for Information Management AGROVOC and the Agricultural Ontology Service (AOS). UN FAO, Rome, Italy. Food and Agricultural Organization of United Nations (FAO), http://www.fao.org/aims (Accessed 2014). Fox, M.S., Barbuceanu, M., Gruninger, M., 1996. An organisation ontology for enterprise modeling: preliminary concepts for linking structure and behaviour. Comput. Ind. 29, 123–134. Garforth, C., Angell, B., Archer, J., Green, K., 2003. Improving farmers' access to advice on land management: lessons from case studies in developed countries. ODI Agricultural Research and Extension Network PaperOverseas Development Institute, London. Ginige, A., 2013. Social life networks for the middle of the pyramid. http://www.sln4mop. org//index.php/sln/articles/index/1/3 (Accessed 2013). Glendenning, C.J., Babu, S.C., Asenso-Okyere, K., 2010. Review of agriculture extension in India: are farmers' information needs being met? Discussion paperInternational Food Policy Research Institute. Gruber, T.R., 1993. A translation approach to portable ontology specifications. Technical Report KSL 92-71Knowledge System Laboratory, Stanford University, California. Gruber, T.R., 1995. Toward principles for the design of ontologies used for knowledge sharing. Int. J. Hum. Comput. Stud. 43, 907–928. Grüninger, M., Fox, M.S., 1995. Methodology for the design and evaluation of ontologies. Workshop on Basic Ontological Issues in Knowledge Sharing, IJCAI-95. Montreal, Canada. Heeptaisong, T., Srivihok, A., 2010. Ontology development for searching soil knowledge. The 9th International Conference on e-Business (iNCEB2010), Bangkok, Thailand. Hettiarachchi, S., 2011. Leeks cultivators desperate as price drops to record low. Sunday Times, Sri Lanka. Hevner, A.R., March, S.T., Park, J., Ram, S., 2004. Design science in information systems research. MIS Q. 28 (1), 75–105. Kawtrakul, A., 2012. Ontology engineering and knowledge services for agriculture domain. J. Integr. Agric. 11 (5), 741–751. Klyne, G., Carroll, J.J., 2004. Resource Description Framework (RDF): Concepts and Abstract Syntax. World Wide Web Consortium. Knublauch, H., Fergerson, R.W., Noy, N.F., Musen, M.A., 2004. The protégé OWL plugin: an open development environment for semantic web applications. In: McIlraith, S.A., et al. (Eds.), ISWC 2004Lecture Notes in Computer Science. 3298. Springer, Heidelberg, pp. 229–243. Lokanathan, S., Kapugama, N., 2012. Smallholders and Micro-Enterprises in Agriculture: Information Needs & Communication Patterns. LIRNE Asia, Colombo, Sri Lanka, pp. 1–48. Mattingley-Scott, M., 2013. Delphi method. http://www.12manage.com/methods_ helmer_delphi_method.html (Accessed 2013). Ministry of Agriculture (MOA), Sri Lanka, http://mimrd.gov.lk [Accessed 2013]. Narula, S.A., Nainwal, N., 2010. ICTs and agricultural supply chains — opportunities and strategies for successful implementation information technology in developing countries. A Newsletter of the International Federation for Information Processing (IFIP) Working Group 9.4. 20. Centre for Electronic Governance, Indian Institute of Management, Ahmedabad, pp. 24–28. NAVAGOVIYA, CIC (Private Sector), . http://www.navagoviya.org (Accessed 2013). Oldakowski, R.,Bizer, C., Westphal, D., 2005. RAP: RDF API for PHP. Proceedings of the 1st Workshop on Scripting for the Semantic Web, 2nd European Semantic Web Conference (ESWC2005). Parent, C.,Spaccapietra, S., 2009. An overview of modularity. In: Stuckenschmidt, H., et al. (Eds.), Modular OntologiesLecture Notes in Computer Science. 5445. Springer-Verlag, Berlin Heidelberg, pp. 5–23. Patel-Schneider, P.F., Hayes, P., Horrocks, I., 2013. OWL web ontology language semantics and abstract syntax. W3C Recommendation. http://www.w3.org/TR/owl-semantics (Accessed 2013). PHP and Semantic Web, 2013. http://blog.m1k.info/category/semantic-web-2/?lang= en&lang=en (Accessed 2013).

150

A.I. Walisadeera et al. / Ecological Informatics 26 (2015) 140–150

Poveda, M.,Suárez-Figueroa, M.C.,Gómez-Pérez, A., 2010. Common pitfalls in ontology development. In: Meseguer, P., et al. (Eds.), Current Topics in Artificial Intelligence (CAEPIA 2009)Lecture Notes in Artificial Intelligence. 5988. Springer-Verlag, Berlin Heidelberg, pp. 91–100. Poveda-Villalón, M., Suárez-Figueroa, M.C., Gómez-Pérez, A., 2012. Validating ontologies with OOPS! In: Teije, A., et al. (Eds.), 18th International Conference on Knowledge Engineering and Knowledge Management (EKAW 2012)Lecture Notes in Computer Science. 7603. Springer-Verlag, Berlin Heidelberg, pp. 267–281. Prud'hommeaux, E., Seaborne, A., 2008. SPARQL query language for RDF. W3C Recommendation. http://www.w3.org/TR/rdf-sparql-query. Semantic web best practices and development. W3C Recommendation. http://www.w3. org/2001/sw/BestPrctices/ (Accessed 2013). Sriswasdi, W., Luengsrisagoon, S., Lorsuwansiri, N., Wuttilerdcharoenwong, S., Khunthong, V., Suksaengsri, T., Kawtrakul, A., Seebungkerd, N., Tananon, U., Narkwiboonwong, W., Pusittigul, A., 2008. A smart mobilized fertilizing expert system: 1-2-3 personalized fertilizer. World Conference on Agricultural Information and IT, Tokyo, Japan, pp. 397–404. Suárez-Figueroa, M.C.,Brockmans, S.,Gangemi, A.,Gómez-Pérez, A.,Lehmann, J.,Lewen, H., Presutti, V.,Sabou, M., 2007. NeOn modeling components. http://www.neon-project. org/deliverables/WP5/NeOn_2007_D5.1.1v3.pdf (Accessed 2014).

Thunkijjanukij, A., Kawtrakul, A., Panichsakpatana, S., Veesommai, U., 2009. Rice production knowledge management: criteria for ontology development. Thai J. Agric. Sci. 42 (2), 115–124. Toninelli, A., Bradshaw, J.M., Kagal, L., Montanari, R., 2005. ISWC2005 Semantic Web and Policy Workshop, Galway, Ireland. Uschold, M., Gruninger, M., 1996. Ontologies: principles, methods and applications. Knowl. Eng. Rev. 11 (2), 93–155. Walisadeera, A.I.,Wikramanayake, G.N.,Ginige, A., 2013a. An ontological approach to meet information needs of farmers in Sri Lanka. In: Murgante, B., et al. (Eds.), 2013 International Conference on Computational Science and its Applications (ICCSA 2013)Part I, Lecture Notes in Computer Science. 7971. Springer-Verlag, Berlin Heidelberg, pp. 228–240. Walisadeera, A.I., Wikramanayake, G.N.,Ginige, A., 2013b. Designing a farmer centred ontology for social life network. 2nd International Conference on Data Management Technologies and Applications (DATA 2013), Reykjavík, Iceland, pp. 238–247. Walisadeera, A.I., Ginige, A., Wikramanayake, G.N., 2014. Conceptualizing crop life cycle events to create a user centered ontology for farmers. In: Murgante, B., et al. (Eds.), 2014 International Conference on Computational Science and its Applications (ICCSA 2014)Part V, Lecture Notes in Computer Science. 8583. Springer International Publishing, Switzerland, pp. 791–806.