How To Develop An Orchid Cluster Platform For Orchid Knowledge Knowledge

Transcription

1 Ontology based Orchid Knowledge Platform for Knowledge Services in Orchid Cluster Thepchai Supnithi, Pisuth Paiboonrat, Marut Buranarach, Asanee Kawtrakul National Electronics and Computer Technology Center, Pathumthani, Thailand {thepchai.supnithi,pisuth.paiboonrat,marut.buranarach, Setapong Lekawatana, Ath Intalak Department of Agricultural Extension Ministry of Agriculture and Cooperatives, Nonthiburi, Thailand Suvichai Sangtien Rajburi Orchids Cluster, Rajburi Thailand Worapot Choptham Faculty of Arts, Silpakorn University, Nakornpathom, Thailand Abstract Orchid is one of the economic crops in world market. The varieties of shape, style, color and smell of orchids attract people to create products in many ways, such as cut orchid, orchid plant, bouquet, lei, corsage, and orchid loose bloom and so on. Thailand is one of the biggest exporters in orchid market. There are a lot of processes in the chain, such as reproduction, cultivation, market trend survey and analysis, transportation, international trade standard, traceability and so on. In order to provide a smart knowledge service, we have to 1) managing knowledge efficiently in terms of sharing and reusing knowledge and 2) providing knowledge which matches to stakeholders. In this paper, we propose a general framework on ontology base orchid cluster platform for developing a usercentric model of knowledge services in orchid cluster. We give some examples on semantic search model, recommended system model and tracking and monitoring model respectively. knowledge service, ontology and model, orchid cluster platform I. INTRODUCTION Because of suitable climate, most orchids are grown on benches under simple saran shading. Orchid production areas in Thailand are scattered around the central region mainly Bangkok, Nonthaburi, Nakorn Prathom, Ratchaburi, Samut Sakorn, Kanchanaburi and Ayudhaya. Tropical orchids are the most important crops that cover approximately 3,500 hectares. Cut orchid production area is more than 90% of the total orchid production area and less than 10% is for potted orchid production. Contrary to the current economic crisis, growers are expanding their farm areas and sites. Dendrobium is the most dominant crop. Besides, Mokara, Oncidium, Aranthera, Aranda, Vanda, Arachnis, Renanthera, Ascocenda, Phalaenopsis, Cattleya and Paphiopedilum are being grown as cut-flowers and potted plants. With suitable climate in combination with more than 40 years of orchid growing experience, Thailand can grow orchids at a competitive cost. Growers prefer to get their planting material from tissue culture rather than division or stem cutting. Therefore, there are many clean cultured laboratories in Thailand producing young plants from tissue culture and seedlings of good quality at reasonable prices, not only for local supply but for export as well. New varieties are also continuously being introduced by local breeders and one could expect Dendrobium and other orchids in many new colors and shapes. About half of the production will go to local market and the other half for export. Most Thai consider orchid as common flowers and usually used for religious purposes. Nowadays more and more orchids are valued and are used as gifts or as decorations at official or social functions. Thailand is currently the world largest exporter of Dendrobium orchid and other tropical orchids. However, with the world economic crisis, Thai orchid also suffered, the export value dropped about 6.3% (4.6% for cut orchids and 15.5% for orchid plants) in 2009 compared to the previous year. In 2009, Thai orchid export value was 80 million U$ accounting for 77% of all Thai floricultural products. Export of orchid flowers was 69 million U$ and 10.8 million U$ for orchid plants to Japan, USA, China, Italy, the Netherlands and Taiwan as major markets. Orchid products include cut flowers, loose blooms, bouquets, leis or garlands, orchid plants, seedlings and flasks, etc.

2 TABLE 1 TYPE OF ORCHID PLANTS EXPORTED FROM THAILAND IN 2009 Type Quantity Value (U$) Share (Value %) Major imported countries (Quantity Shared %) Major Orchid Genera (Quantity Shared %) Plant (Live) 29,699,037 Plants 9,312, USA (19%), NL (16%), Rep. of Korea (14%), Vietnam (12%), Japan (11%), Dendrobium (66%), Phalaenopsis (23%), Vanda (4%), Cattleya (2%), Mokara (2%), Oncidium (1%) Plant (Dried) 2,063,860 kg. 1,378, China (100%) Dendrobium (100%) Flask 811,924 Flasks 1,101, Seedling 2,280,092 Seedlings 125, Root (Dried) Root (Fresh) 1, ,425 3,329 28,312 kg. Roots kg. Roots 12, , Bulb 294 Bulbs Total 10,555, NL (31%), USA (22%), Taiwan (14%) Japan (9%), Vietnam (4%) USA (58%), Brazil (15%), Japan (10%), Israel (6%), Vietnam (4%) Germany (80%), NL (15%), UAE (4%) NL (36%), UAE (35%), Japan (18%) Japan (41%), Germany (41%), Czech Rep. (13%) Phalaenopsis (47%), Dendrobium (22%), Oncidium (16%), Cattleya (3%) Phalaenopsis (47%), Dendrobium (34%), Oncidium (12%), Vanda (2%), Cattleya (2%) Vanda (98%), Aranda (2%) Aranthera (64%), Dendrobium (27%), Aranda (9%) Pecteilis (54%), Habenaria (41%), Spathoglottis (3%) There are several factors on the growth of orchid in the market; (1) the increasing on competition in orchid market (2) the growth of orchid industry in Thailand, and (3) the global warming which effect to orchid cultivation environment. It is necessary to continuously improve the quality and strengthen orchid to alive in changing environment. We design a framework to study and understand all processes in the chain, such as reproduction, cultivation, market trend survey and analysis, transportation, international trade standard, traceability and so on. In order to provide a smart knowledge service, we have to 1) managing knowledge efficiently in terms of sharing and reusing knowledge and 2) providing knowledge which matches to stakeholders. We mainly focus on the orchid cluster platform for the interoperability among concepts for modeling the appropriate framework to support all stakeholders. We organize this paper as follows; Section 2 shows some information on Orchid in Thailand and our previous work on ORCHIDNET, a web portal for orchid cluster. Section 3 shows orchid knowledge platform, which are orchid supply chain management, orchid knowledge platform architecture. Section 4 gives an example on Orchid ontology construction. Section 5 illustrates orchid data management. Section 6 explains components in platform and some examples on model which can be applied in orchid cluster. These models are constructed from orchid knowledge platform. Section 7 gives the conclusion and future work. II. BACKGROUND In this section, we explain a background on orchid cluster information in Thailand and show our previous attempt to develop an ORCHIDNET which is a web portal on orchid. A. Orchid Cluster in Thailand The world exporting value for Floricultural Products comprising cut-flowers, cut-foliage, bulbs and plants was 18,314 million $ in 2008, of which 9,099 million U$ were from the Netherlands which was more than half of the world exporting value. Thailand exporting value for floriculture products was worth 111 million $ and was ranked the 20 th in the world in Considering world export value of orchid flower, the Netherlands ranked the first with the export value of million U$ (48.2%) while Thailand ranked second with export value of 73.3 million $ (31.6%). The Netherlands exports temperate cut orchids such as Cymbidium while Thailand exports tropical cut orchids such as Dendrobium. Therefore, Thailand is currently the world largest exporter of Dendrobium orchid and other tropical orchids. Most orchid flowers exported from Thailand were orchid inflorescences or stems with the value share of 91.4%. Orchid genera exported as stems were Dendrobium, which accounted for 96.4%. The rest were Mokara, Aranthera, Oncidium, Aranda, Vanda and Arachnic. Other forms of orchid products were loose blooms, garlands, dried

3 Figure 1. Orchid Production Area (in 2008) flowers, bouquets, and corsages. Considering import value, Japan was also the largest importers of Thai orchid plants at 2.4 million U$ worth followed by the Netherlands, USA, Germany, Rep. of Korea and Vietnam with the value of 1.6, 1.4, 1, 0.9 and 0.9 million U$ respectively. Thailand exported 211 genera of orchid plants in More than 75% of total export value was orchid hybrids and approximately 24% was orchid species. Table 1 shows type of orchid plants exported from Thailand in 2009 [2]. Figure 1 shows the information on orchid production area in Thailand (2008). It is obviously seen that most of orchid farms are located in Nakornpathom, Samutsakorn, Bangkok, Rajburi. We design this framework with Rajburi-Nakorn Prathom orchid cluster who is the leader between orchid growers in the Thailand. B. ORCHIDNET ORCHIDNET [1] is an orchid web portal which aims to manage orchid knowledge and provide knowledge service to all stakeholders, from orchid farmer to end user perspective. It is a project under Agricultural Technology Transfer and Service Center, Ministry of Agriculture and Cooperatives, Thailand. The objective of this work is to provide an orchid resource for all stakeholders to reproduce, cultivate, transport, export and support orchid supply chain to acquire a sustainable orchid business. Moreover, we can promote Thai orchid from distributed knowledge on the website. ORCHIDNET contains the following information. (a) Cultivate information : This section provides information on orchid family, orchid care both basic and advance technique, orchid protection act, production cost estimation plan, geographical information and total production, stakeholders list and others end products from orchid, such as wreath, bouquet and so on (b) Market information: This section provides both monthly and annual statistics on orchid import and export in Thailand and others in the world, i.e., exporter information, the movement of orchid market price. All Figure 2. A snapshot of ORCHITNET web portal necessary regulations which are provided for orchid market are also collected. (c) Customer information: This section provides information for orchid care to customers. This knowledge helps them to extend the lifetime of various types of orchid. Some examples on flower arrangement technique and gardening tips (d) Orchid research: This section provides Orchid research database from Kasetsart University. (e) Orchid e-library: This section provides orchid related books and journal. The information is provided from agriculture promotion and development center, Samutsakorn province. (f) Orchid festival and activity: This section gives all information of orchid festival and activity in each period.

4 Figure 3 A supply chain management of orchid (g) Orchid expert network: This section collects all orchid experts from government, university and public sector such as farmer, exporter (h) Orchid reference: This section shows orchid related organization, such as standard certification, orchid tissue culture research unit and their functions. (i) Orchid news: This section provides all orchid related news in the value chain from reproductive training news to orchid exhibition Figure 2 shows an example on ORCHIDNET web portal (in Thai). III. ORCHID KNOWLEDGE PLATFORM To utilize ORCHIDNET, we design an orchid knowledge platform which analyze based on all supply chain process. A. Supply chain management process Figure 3 shows the overview of orchid supply chain management. There are two levels; the bottom level represents all necessary infrastructures to develop My-Farm system and I-farm system. Geo Spatial Services are provided in both web and mobile to enable all stakeholders to collaborate with others. QR code is another technology which enables to retrieve information and follow the activity in each process flow. Sensor technology is one of the most important technologies assisting to develop a smart farm. It is applied to get the real time data to analyze, to plan for packaging with the highest efficiency. For instance, if we have to keep orchid product in a storage room, the suitable environment is required to extend the product life. In this case, sensor plays the important role to control, humidity, temperature and so on. Water control is the most important factor for orchid care. We collect data on controlling minerals, cleanliness and PH of water. Since orchid hurts easily and need to be careful with pest and disease. Smart logistics becomes another process to support flower-friendly transportation and without pest and disease. The top level represents stakeholders in the supply chain management. Orchid farmer plays a major role to reproduce, cultivate for a good quality of orchid. Currently, there is a gap between old and young generation because of the difference environment, the old generation apply his experience to make a good quality product, however, young generation has a powerful idea, theoretical knowledge but it is difficult to apply in the real situation. Digitizing tacit knowledge from old generation and co-create a good template for young generation is the most important task for the sustainability orchid cluster. Those templates might be explored to others in bigger scale in terms of common knowledge. Exporter and Retailer who understand all reproduction and cultivation process can apply this knowledge to add more value to products. Finally, customers who need not only products, but also history of the products can taste both good quality and experience from farmer, exporter and retailer.

5 Figure 4. An architecture of orchid knowledge platform B. Orchid Knowledge Platform Architecture Figure 4 illustrates the architecture of our system based on the supply chain management explained in Figure 2. To maintain the interoperability and sharing resource, we developed necessary ontology, orchid ontology, orchid market ontology, export ontology and standard ontology. The integration of these ontologies enables us to represent all related task. Figure 5 shows some example of our orchid ontology. Ontologies are applied into the data level and application level. In data level, currently we concentrate on Orchid map, which represents location based data, such as geographical data on orchid farm, the environment nearby, such as industrial factory, to investigate on the appropriateness of orchid farm. In application level, we currently focus on three applications, semantic search system, recommended system and decision support system, and tracking and monitoring system. Both data level and application level can be applied into a lot of services. Education service aims to educate orchid farmer to reproduce, cultivate the orchid and improve new techniques. Agriculture service aims to transfer experience, knowledge from orchid farmer as value-added information to end users. Tourism service applies orchid map to promote orchid farm to tourists. Commerce service provides all marketing related statistics to orchid farmer, exporters, retailers to generate alternatives decision based on the decision support system provided in this framework. All services can be applied into orchid community, orchid environment, orchid in agriculture and orchid in industrial respectively. IV. ORCHID ONTOLOGY Ontology is defined as a theory of conceptualization in artificial intelligence [7, 8]. It is increasingly known as a Figure 5. A snapshot on Orchid Ontology

6 conceptual level of knowledge that can reuse, sharing among related users in a group. It also can be a standard for a specific group, and finally extend to global standard. Ontology is applied into a lot of standard, such as, MPEG ISO meeting, e-learning standard such as LOM, SCORM and so on. Orchid Ontology is the first ontology designed to serve our orchid cluster. Orchid ontology developing process is explained as follows: 1) List all possible related word: All orchid related word is collected from both primary resource, such as expert interview, and secondary resource, such as book, electronic resources. All secondary resources have to be authorized by expert. 2) Group similar meaning words into concept: Experts will select words with similar meaning into the same group. 3) Define relation among concepts based on the primitive relations; is-a, part-of and attribute-of relation. A is-a relation is a class-subclass relation who is designed for hierarchical relation among concepts and enable to connecting among them. Taxonomy can be applied to construct this structure. A part-of relation which represents a whole-part relation and attribute-of relation which represents the properties of the class are designed to construct a concept definition. Is-a and part-of relation are the relation which connecting between classes, but attribute-of relation is defined by the primitive data type, such as string, integer, Boolean and so on. For example, root, leaves, flowers are part-of orchid. The orchid leaf is parallel vein properties. 4) Define the cardinality for each part-of relation: Cardinality will enable to know the number of pieces in each component in each concept. For example, the cardinality of orchid flower petal is three. 5) Scope the task and define task-related concept in detail: Task specific is an approach to enable the designed ontology to be applied efficiently. Moreover, it assists expert to design suitable concepts and focus only in the close-related concepts in the task. 6) Discuss among knowledge engineer and expertise to determine ambiguity relations: Some concepts has varieties of representation, the discussion results show the consensus between experts and knowledge engineer. 7) Repeat step 2-5 to revise ontology and finalize ontology As shown in figure 5, there are currently, more than 100 concepts in Orchid ontology. We also apply the same process to develop other ontologies shown in figure 4. Figure 6. Orchid farm (left) and cluster (right) in Banglen City V. ORCHID DATA MANAGEMENT A. Orchid Map Based on statistics on orchid production area in 2008, we found that Nakornpathom has the most orchid farm as shown in figure 1. We select Banglen city, Nakornpathom and survey orchid production area. Figure 6 shows an example of orchid farm, and orchid cluster in Banglen city. This technique is applied from the mapping location in Silpakorn university [5]. B. Knowledge sharing Environment Since orchid cluster has a strong community, young generation has IT skill. They communicate each other via social network, such as using chat message in Blackberry, facebook and so on. Currently we designed a localized facebook for orchid community. VI. ONTOLOGY BASED ORCHID KNOWLEDGE MANAGEMENT SYSTEM Ontology based Orchid Knowledge Management System is designed using an add-on component concept. First, we design basic components which relate to input, process and output. Each component can be integrated into a lot of models based on user's requirement. In this section, we give three examples on requirement to develop 1) semantic search system, 2) recommended system with IVR and 3) tracking and monitoring system. A. Basic Component for Application Construction We design the three types of main basic components, these are, input component, process component, and output component. The integration of each component can be modeled into various types of application. 1) Input component Input component is mainly developed to handle ontology construction, and integration between real data and concept. Currently, the following components are designed. a) Ontology Creation Support Component Developing a good quality of ontology is an important task. Currently, there are a lot of Ontology editors which provided for expertise, such as Protégé [3], Hozo Editor [9], and so on. Protege is the most active group and has a lot of plug-in module. Unfortunately, it is difficult for Thai expert

7 Figure 7. A template generation component Figure 8. A mapping instance and ontology component and its snapshot Figure 9. A snapshot on inference rule creation component who has very few experiences on IT skill. In this work, we apply Hozo editor which is a WYSIWYG-like framework. It is designed in graphical view and it allows users to edit, and share their concepts in the same collaborative environment. With this tool, we aim to construct the ontology and educate knowledge construction concept by self-evaluation their concepts and observing, diagnosing other s concept. b) Template Generation Component After we design ontology, the next step is to construct instances that follow ontology we designed. Normally, instances can be retrieved from real data. We define an instance of each class by generating its related template. The main propose of template generation is to model the target class. The template of all classes related to the target class will be generated to enable all related instances are stored. A Part-of relation in each class will be linked to another related class. The transitivity relation is applied in part-of relation. An attribute-of relation will be applied as constraints of instant value, such as integer, string, Boolean. Cardinality gives the criteria on the number of slots for each related part-of concept. Figure 7 shows the architecture on template generation component. c) Mapping Component If database and ontologies are developed separately, it is necessary to construct a semi-automatic mapping tool between ontology and database. In our approach, we classified the process into four types as illustrated in figure 8. 1O1D correspondence: ontology matches with a single database. 1OND correspondence: ontology matches with multiple databases. In this case we join those databases into one single database to represent the ontology. MO1D correspondence: Multiple ontologies match with a single database. In this case, we split the database into multiple databases by selecting related columns and use the related values those columns as criteria. MOND correspondence: Multiple ontologies match with multiple databases. In this case we apply MO1D and 1OND to the design process.

8 d) Inference Rule Creation Component Some applications such as semantic search, constraints on class, basic relation and properties can be applied automatically. However, recommended system or decision support system requires a rule engine to construct an additional criterion. Figure 9 shows a snapshot of inference rule creation support system. We apply JENA[4] and SPARQL [10] for developing the backend of this component. 2) Process Component a) Semantic Search Component This component is designed to extend the possibility to improve the quality of search. There are two major advantages on using semantic in search process. Firstly, semantic search enables us to find the most appropriate result which is not detected by syntax or keyword. Synonym set can be applied to detect other keywords with the same meaning and get more alternatives. For example, orchid has Thai name, common name and scientific name. We apply synonym set to combine all names with the same meaning. Secondly, if we cannot find the appropriate results from keywords, hierarchical representation enables the system to reply the most connected alternatives, such as parent-child relation, sibling s relation. b) Ontology based Recommended Component This component is designed to give the appropriate knowledge to advice users. It is the extension of semantic search. It is necessary to develop an inference engine to analyze input data and give some recommendation based on input from each user. Ontology will play a role as conceptual knowledge applied in inference rule. c) Ontology based Tracking and Monitoring Component This component is designed to collect all statistics information and represented in user-friendly format. Tables or graphs are automatically generated from database which related to concept from ontology. If significance patterns have been found, it can be applied to generate rule. This rule is finally it will be stored as a rule in inference engine for developing other system. d) IVR/Call center Component This component is an additional component for output representation. It applied speech technology to automatically answer basic question. The process starts from recognizing question, matching keyword with appropriate answer, apply rule from rule engine. And finally, reply the answer to end users. 3) Output Component The output component provides alternatives for end users. Currently, we designed two types, web based and mobile. Web based visualization is designed for users based on web browser. It can be represented in stand-alone website or connecting with social network environment, such as Facebook and MSN. Since the demand on using mobile environment is increasingly grown up, we apply the output of our process component in mobile environment too. Figure 10. A semantic search model Figure 11. A shapshot on semantic search system C. Semantic Search Model Figure 10 shows an example on applying our basic component to develop a semantic search system. In this model, user wants to improve the search quality from database search into semantic search. This model starts from developing an orchid ontology using Ontology Creation Support Tool. Next, we map data in database and our designed ontology using Mapping instance and Ontology Tool. We represent the output in web based system. Figure 11 shows a snapshot of our system; we search for the picture that related to Dendrobium, however, we get varieties of Dendrobium which is the subclass of Dendrobium.

9 D. Information Tracking and Monitoring Model Figure 12 shows a tracking and monitoring model. Ontology is constructed in the same way as in semantic E. Personalized recommended system and decision support system Model Figure 14 explains another model, which apply ontology to recommended suitable knowledge through mobile device. Buranarach [6] shows a framework on integrating between Figure 12. An ontology based recommended IVR model Figure 14 An ontology based tracking and monitoring model Figure 13 A snapshot on ontology based recommended system (web based) search model using ontology creation and support component. After that we prepare all related database using template generation component In this case, we receive real time data from sensor. Sensor will send us data and store in the generated database and follows all related constraints among database. We construct a summarized statistics graph and table based on the retrieved data. There are several ways to apply these data. In our case, we analyze data and construct the related knowledge. Those data will be compared to the situation in the orchid farm and the dominant feature will be extracted to generate procedural knowledge. In this case, we will get the rule knowledge to apply in interface rule engine component. Figure 13 shows the result of inappropriate location for orchid. We plot all orchid farms in Banglen and industrial factory nearby, with the interpretation on sensor data, we come up with the warning on locations that are inappropriate with orchid care (represented in blue circle). database and IVR system, but there is no report on applying ontology in the work. In this work, we develop ontology to be a backbone in the system. Similar to the previous two models, we start from ontology construction by Ontology Creation Support component and Template generation component to gather data based on our designed ontology. Experts have to construct inference rule from inference rule support component. Rule can be generated from two ways, his real experience, or knowledge from model 2. Retrieved data and rule will be applied together in recommended system component to generate the appropriate recommended result. We also integrate this component and IVR call center, to develop a mobile based framework for end users. VII. CONCLUSION AND FUTURE WORK In the paper, we design an ontology based knowledge platform and framework to develop user-centric model. We design orchid ontology to serve orchid cluster. Three models examples, i.e., semantic search, recommended system, tracking and monitoring system are introduced here. In the future, we plan to apply this work to Nakornpathom province, Banglen city. REFERENCES [1] [2] [3]

10 [4] [5] [6] Buranarach Marut, et. al. Development of a personalized Knowledge Portal to Support Diabetes Pateint Self-management, MEDES p.48 (2010) [7] Guarino, N.: Some Ontological Principles for Designing Upper Level Lexical Resources, Proceedings of the First International Conference on Language Resources and Evaluation, Granada, Spain, pp (1998) [8] Gruber TR. A Translation Approach to Portable Ontology Specification. Presented at the Knowledge Acquisition, (1993). [9] Kozaki, K., et al.: Hozo: An Environment for Building/Using Ontologies Based on a Fundamental Consideration of "Role" and "Relationship", Proc. of the 13th International Conference Knowledge Engineering and Knowledge Management(EKAW2002), pp , Siguenza, Spain, October 1-4, 2002 [10] "W3C Semantic Web Activity News - SPARQL is a Recommendation".W3.org n.