SASA- a Semi-Automatic Semantic Annotator for Personal Knowledge Management Kim Tighe, Sean O’Riain Semantic Infrastructure Research Group Hewlett-Packard, Galway, Ireland Tel.:+353-91-754901 {ktighe,sean.oriain}@hp.com ABSTRACT increasingly depend upon their ability to becoming a knowledge- Best practice organisations have realised that people and their rich knowledge managing organisation [2]. Organisations that knowledge remain their greatest assets and will continue to be the focus solely on the application of their collective intellectual largest contributory factor in obtaining future competitive capital to achieve objectives run the risk of neglecting the advantage. Knowledge fundamentally derived from people in the fundamental truth that knowledge is derived from people [3]. absence of their understanding, personal context and application Lacking the human element of understanding, personal context remains largely as obscure information. Current personal and application, knowledge within an organisation remains knowledge management (PKM) activities do not adequately largely as obscure information. Supporting individuals in their support the finding, reminding, reuse and collaboration of PKM is therefore vital and will be the single largest contributory information. In this paper we propose a novel PKM tool called factor in gaining future competitive advantage over the next 25 SASA, a semi-automatic semantic annotator of PDF documents, years [4]. which will enable collecting, connecting and collaborating of Enabling technologies for the WWW have provided knowledge discovered information to facilitate knowledge sharing and workers with rich information sources but have also resulted in personal content management within a business setting. SASA, a adding to the existing considerable volume of information that plug-in for Adobe Acrobat Professional, utilises Semantic Web can be searched and queried. The classical Information Retrieval technologies to enable building, augmenting and sharing of (IR) problem of identification and retrieval of current information ontologies amongst knowledge workers. Within an ontology for activities such as informed decision making remains named entities are connected to additional information such as problematic. The core focus of PKM is directed at improving Web pages, documents, mail messages, personal notes, and wikis. individual efficiency. Current activities however remain limited SASA automatically derives the context of the document, lacking adequate support for the finding, reminding, reuse and highlights named entities and applies the relevant additional collaboration of information [5]. There remains a growing need information. The business case for such a tool is outlined and user for intuitive processes and PKM based tools to assist the worker scenario development used to illustrate how SASA will assist in evaluating not only their own knowledge but also a means to Business Client Account Managers in the laborious process of augment it by exploration and learning from additional reviewing, annotating and gathering information from customer information sources. Maximising human capital on a personal documentation by enhancing their PKM. level leads to enhancing individual effectiveness in a manner that improves productivity for both the individual and enterprise [6]. It Categories and Subject Descriptors is our contention that PKM enhanced with Semantic Web I.I.7 [Computing Methodologies]: Document and Text technologies can be used to assist in achieving this productivity Processing – General. J.0 [Computer Applications]: General. gain. The Semantic Web [7] envisages annotating document content by General Terms assigning to entites in the text links to their semantic descriptions Management, Economics, Experimentation from domain ontologies to make it easier for machines to assist humans in finding, sharing, combining, and reusing information. Current semantic annotation tools (e.g. KIM [8], Trailblazer [16] 1 Keywords and tools based on Annotea 2 or CREAM [9]) cater for document Semantic Annotation, Personal Knowledge Management, PDF annotation of Web-native formats such as HTML and XML. None however cater for the Portable Document Format (PDF [10] 3 ), a 1. INTRODUCTION format prevalent in virtually all market segments and used In a rapidly changing global economy unified by improved extensively for document interchange and publishing. communication and transportation, people and their knowledge are an organisation’s greatest assets [1]. The constant emergence of new products and competitors combined with an increasing 1 global marketplace are challenges facing an organisation’s ability http://www.hp.com/ie/galway/sirg/trailblazer/ 2 to survive in an increasingly unpredictable and competitive http://www.annotae.org/ environment. An enterprise’s continued existence will 3 A de facto standard on the Web alongside HTML. With the advent of the Semantic Web this paper examines how comments that are difficult to identify and retrieve particularly for Semantic Web enabling technologies, namely semantic annotation new document versions. Increasing the customer base, scope can be applied to the area of PKM to enhance knowledge worker expansion, EUWM organisational expansion and having to productivity and efficiency. This paper proposes a novel tool, comply with standards such as ISO 7 frameworks has led to a SASA for semi-automatic semantic annotation of PDF considerable increase in documentation volume and the level of documents, which will enable collecting, connecting and manual effort required. collaborating amongst knowledge workers to facilitate knowledge There is a clear opportunity for an intuitive tool that would assist sharing and personal content management within a business the consultants in performing documentation review and in setting. information gathering process in order to improve both The remainder of this paper is structured as follows: Section 2 collaboration and traceability of document findings. Currently outlines the business case. Section 3 illustrates the scenario under active development SASA is such a tool that offers the development. Section 4 presents our proposed solution. Section 5 semi-automatic semantic annotation of PDF documents. Its usage compares related work. Section 6 concludes this paper and will contribute towards a reduction in the level of effort required outlines future work. in each new project stage, cost reduction and increased team productivity. 2. BUSINESS CASE HP 4 Services’ Managed Services (MS) provides customers 5 with 3. SCENARIO DEVELOPMENT strategic outsourcing services and solutions to manage their IT Take for example the situation where a EUWM client account infrastructures. The MS business unit itself is structured into a manager has to prepare and deliver a Project Definition Document number of areas of expertise known as towers. Each tower based on the requirements of the U.S. car manufacturing customer specializes in a particular area such as the End User Workplace “Customer X”, and the capabilities of the delivery organisation Management (EUWM) which focuses on the end users desktop and support structure. Material used in drawing this information environment. The EUWM Pre-Sales & Implementation Team is together is contained in a large number of key business assigned Customer Relationship Management and Project documents such as Statement of Work (SOW), HP Overall Scope Management activities during the pre-sales and document, Technical Solution Document (TSD), RCM Model, transition/transformation stages of any customer engagement. Overall HP-Customer Contract and Associated Schedules, Requests For Information (RFI), Requests For Proposals (RFP), In each of the above activities, the EUWM consultant’s task of etc. understanding, interpreting and producing all relevant support documentation is crucial for the successful proposal, The assigned account manager firstly accesses the implementation and delivery of any service. Failing to adequately transition\transformation and delivery documentation, begins the capture all customer requirements, service limitations and any analysis process in an attempt to identify what is of relevance to assumptions made will impact customer satisfaction level, the the EUWM tower and what contractually HP are obliged to delivery organisations ability to succeed, HP’s profitability, and deliver. The Overall HP-Customer Contract is opened with Adobe ultimately, HP’s ability to win further contracts. Underpinning all Acrobat Professional and using our plug-in SASA creates the activities is the consultant whom has to ensure that services category 8 “Customer X” for the customer and begins reviewing scoped in the solution are delivered efficiently and implemented the documentation. With reference to Figure 1, when the account in adherence to contractual obligations. For that reason, their manager identifies an item of interest such as ‘Application resulted outputs from reviewing customer documentation such as Packaging’ it is added to the category as a named entity. A note of Project Definition Document or Project Requirements Document “Due to ITAR 9 U.S. government regulations all Customer X are essential for the project to initially commence and to continue transmissions must be manufactured within North America” is on-going successfully. associated with that named entity. As the account managers’ analysis progresses, another document, which is part of the New EUWM customer undertaking will require the consultant to Associated Schedules documentation, is found to contain a key begin the laborious process of reviewing, annotating and stipulation regarding where UNIX application packaging must be gathering information from on average 50 or more substantial performed. Another note is then added to the named entity documents which typically are received in either Microsoft Word ‘Application Packaging’ along with a bookmark to document or Adobe PDF format. At present, each document is manually Schedule B, which was found to have the associated information. reviewed and annotated by the consultant. Central document In this manner peripheral information obtained from sources such repository systems such as SharePoint 6 are occasionally used for as emails, phones calls and HP-Customer group discussions can information sharing in addition to documentation notes capture in be used to filter information and associate it with suitable named an associated mail or Word documents. However, it is not a entities. Once the document review stage has concluded the standard practice and can lead to problems of omitting key Project Definition Document write up commences. Resulting from the review the accounts manager now has in effect a semantically annotated information source. 4 Hewlett-Packard Ltd. 5 Telecom/NSP, financial services, manufacturing and 7 government or public sector markets. International Standards Organisation 8 6 SharePoint is Microsoft collaborative management tool for ‘Category’ is used to represent an ontology. 9 document and information sharing. International Traffic in Arms Regulations Figure 2 shows that to retrieve the information the manager need Figure 1). The toolbar provides functionality allowing the user to only click ‘Find Entities’ to have all named entities such as apply or create ontologies and link associated information as ‘Application Packaging’ belonging to the category highlighted. semantic annotations to named entities within the PDF document’s text content. Dialog windows created using the Adobe Dialog Manager (ADM) 11 are used to facilitate the addition of the named entities and their associated information. SASA stores the ontologies on the users’ local machine in Resource Description Format (RDF) [11], which makes them portable amongst groups. Referring to Figure 3, the SASA application architecture will contain: 1) A Trainer component to train SASA using the text and the users’ selected ontology about the context of the currently viewed PDF document. SASA will extract the text from the document and use a Vector Space Model (VSM) [12] 12 to represent the collected training information by using the words from the document and their frequency of occurrence to augment the existing training data. 2) A Categorisation component to derive the context of the currently viewed document using the Cosine Similarity Measure [13] to compare the text of the Figure 1. Adding a note to named entity ‘Application document with the training data and calculate from the vectors the Packaging’ most likely match to the current ontologies. 3) A NEIO Selection of the named entity ’Application Packaging’ will component to add and delete named entities and their associated provide visibility of all additional information and annotations information to and from the ontology. 4) An Annotator from previous documentation reviews. The client account component to semantically annotate the text of the PDF document manager also has the ability to view a summary of all the named by finding and highlighting named entities of interest and entites and their associated information (see Figure 2). This assists applying their relevant additional information. 5) An the accounts manager in ensuring that issues regarding the like of Import/Export component to share ontologies amongst users. 6) UNIX application packaging and ITAR regulations are factored in A Viewer component to view an entire trail of annotations for a and captured in the Project Definition Document. Otherwise the selected ontology. potential knock on effects of overlooking this information could adversely affect the project timeline, project scope, level of effort required, and delivery model with ultimately negative commercial impact. Figure 3. Overview of SASA Architecture 5. RELATED WORK SemanticWord [14], a Microsoft Word-based environment, adds Figure 2. Document annotated showing summary of all named several toolbars to the interface which support the creation of entities and their associated information semantic annotations in documents and templates according to selected ontologies. Magpie [15] is a Web browser extension 4. PROPOSED SOLUTION which uses Named Entity Recognition (NER) based on a supplied SASA is implemented as a plug-in 10 for Adobe Acrobat Professional. SASA adds a toolbar to the standard interface (see 11 A cross-platform API for implementing dialog interfaces for 10 A dynamically-linked extension to Acrobat, which hooks into Adobe applications such as Acrobat, etc. 12 the user interface and adds functionality to Acrobat An algebraic model used for information filtering and Professional, Acrobat Standard, or Adobe Reader. information retrieval. ontology of the user’s choice to highlight and add links to named [2] Davenport, T. H., Prusak L. “Working Knowledge, How entities on a Web page. Table 1 shows an extract from a recent Organisations Manage What They Know” Harvard Business survey of semantic annotation tools. It was found that they cater School Press 1997. primarily for Web native formats such as HTML and XML. [3] Nonaka, I., Takeuchi, M. (1995). The Knowledge Creating SASA caters for PDF and can be integrated with HPs Mozilla Company – How Japanese Companies Create the Dynamics Firefox extension Trailblazer [16] to allow for HTML also. of Innovation. Oxford, The Oxford University Press. Table 1. Extract from comparison of annotation tools for [4] Drucker, P. F. “Managing Knowledge Means Managing requirements 1-7 [17]. Oneself” Leader to Leader. 16(Spring 2000):8-10. Annotation Semantic [5] Volkel, M., Oren, E. Personal Knowledge Management with Magpie Amaya Tool Word Semantic Wikis Technical Report, AIFB Karlsruhe. Standard DAML+OIL HTML OCML RDF(S) XLink, December 2005. formats XPointer User- [6] Ernst & Young Center for Business Innovation. (1995). The Microsoft Word Web browser Web browser & Financial and Non-Financial Returns to Innovative Work centered GUIs plug-in editor Practices. New York: Ernst & Young. March. design Ontology [7] Berners-Lee, T, Hendler, J & Lassils, O. The Semantic Web, - - Annotation server support Scientific American, May 2001. Document HTML, XHTML Word HTML formats and XML [8] Popov, B., Kirayakov, A., Ognyanoff, D., Manov, D., Document Mark-up tied to Kirilov, A. KIM-a semantic platform for information - XPointer evolution text regions extraction and retrieval, Nat. Lang. Eng. 10 (3/4) (2004) Annotation Local or 375-392 - None, real time storage annotation server [9] Handschuh, S., Staab, S., Studer, R. Leveraging metadata Automation Yes Yes No creation for the Semantic web with CREAM, KI ‘2003- advances in artificial intelligence, in: Proceedings of the 6. CONCLUSIONS AND FUTURE WORK Annual German Conference on AI, September 2003, 2003. In this paper we have proposed a plug-in for Adobe Acrobat [10] Portable Document Reference Manual, Fifth Edition, Adobe Professional called SASA, a novel PKM tool for semi-automatic Systems Incorporated. semantic annotation of PDF documents utilising Semantic Web enabling technologies. SASA allows the user to build, augment http://partners.adobe.com/public/developer/pdf/index_refere and share ontologies amongst knowledge workers. Within an nce.html ontology named entities are connected to additional information [11] Brickley, D., Guha. R.V. 2004. RDF Vocabulary Description such as Web pages, documents, mail messages, personal notes, Language 1.0: RDF Schema. W3C Recommendation 10 and wikis. SASA automatically derives the context of the February 2004. document, highlights named entities and applies the relevant [12] Salton G., McGill, M. Introduction to Modern Information additional information. The business case for such a tool is Retrieval. McGraw-Hill, 1983. outlined and user scenario development used to illustrate how SASA will enhance PKM. Our future work plans, aside from [13] Van Rijsbergen. C.J. Information Retrieval, 1979. continued implementation of our SASA plug-in, include detailed [14] Tallis, M. SemanticWord processing for content authors, in: definition of the case study. We also plan to carry out a Proceedings of the Knowledge Markup and Semantic systematic user evaluation – with the help of Client Account Annotation Workshop (SEMANNOT 2003) at 2nd Managers at HP Galway. Lastly, we are also working on International Conference on Knowledge Capture (K-CAP semantically annotating a number of PDF documents at the one 2003), October 26, 2003. Sanibel, Florida, USA, 2003. time and researching sub section document training. [15] Domingue, J., Dzbor, M., Motta. E. Collaborative Semantic Web Browsing with Magpie. In Proc. of the 1st European 7. ACKNOWLEDGMENTS Semantic Web Symposium (ESWS), May 2004. We would like to thank Robert Connolly and Richard Joyce from [16] Tighe, K., Johnston, A. Using Named Entities as a basis for the EUWM Pre-Sales & Implementation Team, Dara Keogh, and sharing associative trails between Semantic Desktops. 1st Colman O’Dywer from the Solutions Management Services Team International Semantic Desktop Workshop (ISWC) at HP Galway for their time and expertise in framing the business November 2005. case. [17] Uren, V., Cimiano, P., Iria, J., Handschuh, S., Vargas-Vera, M., Motta, E., Ciravegna F. Semantic Annotation for 8. REFERENCES Knowledge Management: Requirements and a Survey of the [1] Nonaka, I., (1991). The Knowledge Creating Company. State of the Art Journal of Web Semantics: Science, Services Harvard Business Review. and Agents on the World Wide Web (4): 14-28. 2006.