Providing Semantically-enabled Information for Smes Knowledge Workers: Multi-agent-based Middleware

The objective of this research is to present a Multi-agent-based middleware that provides semantically-enabled information for SmES knowledge workers. This middleware is based on the European project E! 9770 PrEmISES [1]. Companies and universities from two EU countries (Romania and Spain) are working in order to help small and medium enterprises to better exploit their information spaces.

One important feature of PrEmISES is its capability to couple with the existing data systems that are used by small and medium companies and in this way to enhance them with a semantic layer/engine. The engine is used to find organizational documents within companies and make the searches more accurate by the use of ontologies.

The main purpose is to help the companies to better exploit their available information space. The engine is semantically enriched meaning that it is searching for the specified words/query plus for semantically related concepts.

In this paper we present PrEmISES architecture. We will present the main components and the main steps that were followed in order to develop the ontologies that were used for the ontology component.

The objective of this research is to present a product, named PrEmISES, that is used in order to help SMEs, from all around the world, to exploit their informational space. Nowadays, each large corporation has its own information management departments and dedicated software used for these purposes. These software framework implementations are not appropriate for the needs of small and medium enterprises, due to their lower budget and their smaller number of employees. The framework that we are developing is able to couple with the existing legacy data systems that are already used within SME companies.

By the use of ontologies, this framework implementation adds semantically enabled information integration and also provides for the employees a work process embedded, context-sensitive information services.

In this paper, we explain the product's high-level architecture, the advantages of ontologies’ automated generation process and how we use these ontologies in semantic search. PrEmISES has been selected among hundreds of innovation projects by Eureka Eurostars and counts with the support of the National Funding Agencies CDTI (Spain) and ANCS (Romania).

PrEmISES project is funded by the Eurostars Programme and will last 24 months with a total budget of 1,3M€. The project is led by Anova IT Consulting. The aim of the project is to help SMEs to better exploit their information space. The project is developed by a consortium that is formed of members (from industry and academia) of four entities from Romania and Spain. PrEmISES is a market-oriented R&D project which will prototype an intelligent middleware solution targeted at supporting SME’s knowledge workers with their common tasks. Our software product can provide:

• technologies for the automatic knowledge structuring that don’t need the redundancy of huge information sets
• multi-agent system for the user profiling
• semantic provisioning of formalized company knowledge to employees (not only searching and retrieving, but also logic-based reasoning fed by the user profiling).

In terms of market analysis, a specific category of target client has been identified: medium size knowledge companies who wants to improve their overall business. PrEmISES will tackle the current market lacks. Market currently offers several commercial instruments for enabling KM such as KM systems, database management systems, technologies for information repositories structures, data warehouse and intranet and extranet knowledge portals. However, these technological solutions do not take into account that KM practices in SMEs are more congruous with apprenticeship-based learning rather than the formal training typical of big companies. This means that, in order to be effective for SMEs a KM system needs to be able to provide relevant company knowledge to users based on context (constant worker context estimation with an activity centric view of context).

In other words, PrEmISES is internationally commercialized as a software license. First it was developed for medium-size Spanish, Portuguese and Romanian enterprises willing to improve business performance through KM solutions.

This solution endows enriched standard information processing mechanisms of legacy systems with a semantically-enabled information integration layer. On the customer side it is implemented through a multi-platform user interface focusing on high usability and smooth workflow integration. The provision of this information layer follows the SaaS software delivery model.

The aim of the PrEmISES project is to develop a system capable of helping SMEs to better exploit their available information spaces.

Due to the fact that the number of employees that are eager to gain knowledge is increasing fast, knowledge also becomes more distributed, it is generated faster and in a higher quantity. In the next section we present some general data of premises and premises advantages over other knowledge management systems. The second section we describe the main functionality and premises high level architecture.

Our software is used in technical areas like processing, information system, knowledge management, process management, IT and telecoms technology. The framework is internationally commercialized as a software license in the market area of computer software and integrated software solutions.

Most of the existing frameworks that are available on the market relay in huge sets of date in order to identify what is important. SMEs are dealing with small or medium sets of data. The advantage of PrEmISES is its capability to work with medium and even small data sets. PrEmISES represents an innovative framework in the area of knowledge management solutions [4, 5, 6]. PrEmISES is internationally commercialized as a software license. It addresses initially the medium-size Spanish, Portuguese and Romanian enterprises willing to improve business performance through KM solutions. In other words, PrEmISES is an affordable knowledge management solution tailored on the actual needs of European SMEs. Our software product that provides:

• technologies for the automatic knowledge structuring that don’t need the redundancy of huge information sets;
• a multi-agent system for the user profiling;
• semantic provisioning of formalized company knowledge to employees (not only searching and retrieving, but also logic-based reasoning fed by the user profiling).

In contrast to big organizations with dedicated information management departments, SMEs face obstacles when attempting to exploit their information resources and perform sustained knowledge management (KM). The actual solutions on the market are not suited for the SMEs need of exploiting knowledge without big financial and time-consuming efforts [10]. PrEmISES solves this market gap and help SMEs to improve business performance.

In [x] we have presented a quantitative research for determining the functional and non-functional requirements of the PrEmISES search engine. In that research there were highlighted the most important features for premises according to a survey that was completed by 60 persons of different age and gender. We took into account those opinions when we have started the development of Premises. In other words, the premises was built as a easy to use framework, with a friendly UI and it is capable to return relevant results in only a short amount of time. More than that, based on our customers need, we have developed our framework by taking into account the implementation of high security features and the possibility to use premises as a portable software that is capable to run on many operating systems.

In [2] a similar framework was presented for the medical domain. The article presents a digital library creation based on ontologies. Through the use of ontologies the mentioned framework is helping patients to select relevant articles for their condition. The framework creates a personalized digital library with filtered medical literacy. The results and advantages are presented in this article by taking as example the asthma condition.

According to [x] Premises is intended to be a low budget software with high accuracy results due to its ontological component. In the same article the high-level architecture of the project was presented, focusing on the domain ontology development process used by the ontological.

Premises architecture was developed by the consortium based on the following processes:

• Initial Process Scanning (for a deep understanding of small and medium enterprises realities)
• Social Subsystem Analysis
• Technical Subsystem Analysis
• Analyses Interpretation
• Solution Design
• Implementation.

As mentioned above the most important components are:

• Search & Administrative System Component
• Analyzing & Indexing System Component
• Ontology component

Each of those three components were built with a complex design and each of them is responsible for a specific task in the overall architecture. Due to their complexity the components were developed by building and integrating many sub-components.

Ontology component

The main objectives for this component were:

• Domain ontology generation
• Developing queries based on the ontologies that were developed

An ontology defines a common vocabulary for researchers who need to share information in a domain. It includes machine-interpretable definitions of basic concepts in the domain and relations among them.

The purpose of authoring ontologies is also reusing of knowledge. Once ontology is created for a domain, it should be (at least to some degree) reusable for other applications in the same domain. To simplify both ontology development and reuse, modular design is beneficial. The modular design uses inheritance of ontologies - upper ontologies describe general knowledge, and application ontologies describe knowledge for a particular application.

Depending on the scope of the ontology, ontology may be classified as follows:

• upper, generic, top-level ontology - describing general knowledge, such as what is time and what is space
• domain ontology - describing a domain, such as medical domain or electrical engineering domain, or narrower domains, such as personal computers domain
• task - ontology suitable for a specific task, such as assembling parts together
• application - ontology developed for a specific application, such as assembling personal computers

At each level modularization can be used as well - for example, upper ontology may consist of modules for real numbers, topology, time, and space (these parts of the upper ontology are usually called generic ontologies). Ontologies at lower levels import ontologies from upper levels and add additional specific knowledge. In this way, ontologies form a lattice of ontologies defined by partial ordering of inheritance of ontologies. Task and domain ontologies may be independent and are merged for application ontology, or it is possible that for example task ontology imports domain ontology. The upper ontologies are the most reused ones while application ontologies may be suitable for one application only.

When developing new ontology, it is desirable to reuse existing ontologies as much as possible. The new ontology should be started by importing upper level ontologies when appropriate ontologies exist. This will simplify the development since one can focus at the domain or application specific knowledge only. It will also simplify integration between applications in the future since defined parts of ontologies will be shared.

The ontology produced by the ontology generation component will be used in PrEmISES in order to tag documents in the document repository with relevant concepts in the specific domain in which the enterprise is operating. This means that the type of ontology produced by the component is a domain ontology. For more complex queries the domain ontologies can be aligned to existing upper ontologies (such as Dolce, Cyc).

The ontology generation component is responsible for creating the enterprise specific domain ontologies which will be used for indexing the company documents and for constructing semantic queries. Another purpose of the component is to extract subontologies for each user profile in order to support context-aware semantic searching.

The ontological component is composed from the following modules:

• Crawlers - the set of dedicated crawlers for structured data, semi-structured data, unstructured data which gather the input data for the ontology learner component.
• Ontology learner - Discovers concepts, instances and properties from the data gathered by the crawlers and stores them in RDF/OWL format.
• Ontology editor - Allows ontology experts to make modifications to the generated ontology in a lightweight visual editor.

Subontology extractor - Uses data from the company user profiles and usage data to extract context-dependent subontologies. The resulted ontologies will be stored in an RDF database which is dedicated for storing triples and answering to SPARQL queries. One of the most used RDF databases is Sesame.

The ontologies that we have developed in [x] are validated using queries for knowledge extraction. We have used Sparql in order to test and validate our ontologies. SPARQL defines queries in terms of graph patterns that are matched against the directed graph representing the RDF data.

SPARQL contains capabilities for querying required and optional graph patterns along with their conjunctions and disjunctions. The result of the match can also be used to construct new RDF graphs using separate graph patterns.

Another important tool that we have used is Sesame. Sesame is an open source Java framework for processing RDF data. This includes parsing, storing, inferencing and querying of/over such data. It offers an easy-to-use API that can be connected to all leading RDF storage solutions. It allows us to connect with SPARQL endpoints and create applications that leverage the power of linked data and Semantic Web.