ADL Newsletter for Educators and Educational Researchers

Semantic Web

Bruce Forrester

Introduction

The semantic web is another vision of Tim Berners-Lee, notable inventor of the World Wide Web. This “linked data” web, which is being developed under the auspices of the World Wide Web Consortium, is also included in ADL planning (for more see http://www.w3.org/DesignIssues/LinkedData.html). Its goal is to improve cooperation between computers and human beings by imbuing Web information with meaning. It is intended to identify and expose semantic linkages between bodies of knowledge regardless of how disparate they appear to be. If there are semantic linkages between 3rd Grade geography and quantum mechanics, the Semantic Web is expected to find them. While most of the underlying computer languages and protocols exist, the semantic web, as envisioned by Berners-Lee(1998), is not yet fully realized, but holds enormous potential:

The Web was designed as an information space, with the goal that it should be useful not only for human-human communication, but also that machines would be able to participate and help. One of the major obstacles to this has been the fact that most information on the Web is designed for human consumption, and even if it was derived from a database with well defined meanings (in at least some terms) for its columns, that the structure of the data is not evident to a robot browsing the web. Leaving aside the artificial intelligence problem of training machines to behave like people, the Semantic Web approach instead develops languages for expressing information in a machine processable form.

Herein lies the problem. If the meaning of all words could be easily captured and decoded, the problem space would be manageable. However, as any married couple will attest, words do not have simple meanings, even words that most would agree should have such simple meaning. For example, if you ask a group of people to take a piece of paper and draw a picture of a chair, you are most likely to get as many different drawings of chairs as there are people. A very simple word, yet some will draw a simple kitchen chair, some Lazyboys, some a beach chair and some will draw a bean bag chair. A very simple word that has many interpretations and understandings. It is not hard to imagine the connotations and nuance involved with concepts such as freedom, independence, or love.

In addition, the machine needs sets of inference rules that allow for automated reasoning about words and data. How does one set of data relate to another set? Any child knows that to be a brother it takes another sibling, either a boy or girl. However, this meaning is not derived from the words in and of themselves. These inference rules are learned and as such must be explicitly coded for machines to understand. Most words derive their meaning from context to the extent that, it could be argued, without context words are meaningless. The multitude of possible meanings for every word literally makes a word meaningless without context. Further, words are in a constant state of evolution based on pop culture and street use. Notwithstanding, I firmly believe that the power that will be made available will be well worth the effort it will take to build a linked data, semantic web.

So What?

Dodds and Fletcher (2004) state:

If successful, the Semantic Web will integrate real-world knowledge and skills acquired through simulation, education, training, performance aiding, and experience. It will provide a foundation for building more comprehensive and substantive models of subject matter domains and learners’ levels of mastery than we now have and combine them with more precise discovery of the instructional objects needed to produce desired human competencies. Learners and practitioners will be presented with a constellation of related activities – learning, doing, trying, referencing. This integration combined with the already available functionalities of intelligent tutoring systems provides the basis for a next generation meta-architecture and learning environments based on instructional objects.

The Semantic Web will allow instructional programs to create more comprehensive and substantive models of subject matter domains and learners’ levels of mastery. Semantic Web services will enhance learner profiles accumulated while working on instructional objects, representing their skills, knowledge, and abilities, linking these representations to instructional objects, and managing their progress toward instructional ends. With appropriate linkages, the semantic web will enable the programming of true adaptive learning. This could be based on threshold testing or better yet, based on learner questions and the accumulation of the questions over time. Semantic Web Services will be language, platform, and object model independent. They will enable different applications running on different operating systems, developed with different object models using different programming languages to cooperate and become easily used Web applications. They will link these models precisely with the instructional objects needed to produce desired human competencies. They will do much to help realize the ADL vision.

Let’s look beyond the seemingly insurmountable problems presented in the introduction and assume that a semantic web exists. What else would be gained from a semantic web? A straightforward task that comes to mind would be the ability to ask the computer questions. Unlike our current Google search, where we must derive keywords that are likely to produce good hits, the semantic web would allow us to ask questions in plain English. So something like, “What were the names of the wives of the US presidents that served two terms in office?” would be possible. While certainly useful, such applications are probably not worth the enormous effort needed to develop the semantic web. However, much greater power is possible. An interesting and very useful scenario can be found in an article from Berners-Lee found in Scientific American (http://www.sciam.com/article.cfm?id=the-semantic-web). This article discusses semantic “agents” that are able to coordinate the schedules of family members, helping their mother to receive health care, based on many factors such as location, availability, and convenience.

In academia, linked data and the semantic connections between objects becomes a virtual goldmine for researchers which will be exponentially augmented once public and private company databases are made available. Let’s look at a scenario that is available today in commercial form, that we could easily apply to say - a college library. Amazon.com sells books and has a web 2.0 site. This means that besides Amazon’s own inputs (inventory database, pricing tables, etc) there is a lively part of the site that is populated by users. Users can write reviews of books and rate other peoples’ reviews as to how useful they were. In addition, there is another database that collects data on users such as what books they buy and how these relate to other books they purchase. All these sources of data allow the website to collect data on which books people are searching and make suggestions as to related books that others have purchased, such as “people who bought this book also bought these books,” followed by a list of books. Amazon also sends out occasional emails that recommend new books related to the ones already purchased. I consider this free research. Of course, Amazon benefits if books are bought, but I no longer have to check the website for new books in my area of interest every couple of weeks.

Imagine that these same types of linked data are available to students in a library context. As students and staff check out books and articles, their choices are entered in a database that starts to link related books and articles. Students and staff can rate the usefulness of materials which then serves as potential roadmaps for students exploring new domains of knowledge. Students could sign up for automatic notification of new materials in their area of research or interest. Advanced searches could be used to discover related materials in cross-domains. I consider this the early start of a semantic web, driven by economics but fairly easily adapted for academic use.

Now we can consider a scenario in a combined military and ADL context:

Company Commander Major Smith has just been handed a high priority task. His company will be the first to respond to a new emergency situation. While he considers his troops to be in a high state of readiness, they have never been to this particular country nor have they had such a task. Similar tasks to be sure, but Major Smith knew that there were some significant differences and challenges.

Smith knew that his troops would require some “just in time training” before they arrived in the country. In preparation, he quickly logged into the ADL semantic database. As in the past, he was able to type in some keywords to find the information but with the addition of semantics he was now able to ask some pertinent questions about the destination in plain English. In the past few years, the ADL site has been updated to integrate with various military and intelligence databases. It also “talks” with the joint service lessons learned database. An ontology was designed for this closed namespace domain which now made it possible to ask the database questions.

Based on Major Smith’s keywords and questions, the database agents (an automated program that know of Smith and his troops) went to work searching for relevant courseware objects, intelligence reports, lessons learned, and specific country reports. Smith’s agent was able to access areas, depending on his security clearance level, and also flagged more sensitive areas that were automatically sent to intelligent agents to consider for release, given Smith’s tasking. In comparing the training and competencies of Smith’s company, the agent determined one critical shortfall and was able to engage a real human expert who was immediately asked to accompany Smith and provide hands on training to key members of the company.

All this occurred while Smith was making preparations and ensuring that his company was ready to depart the next day. The next morning, he downloaded a comprehensive country briefing, some refresher material based on the forecast mission that included recent lessons learned from a similar mission, and several trade specific training packages based on his company’s personnel. His agent was able to convince others of the need to release some of the more sensitive intelligence reports that would prove invaluable. The training material was quickly wired to his troop’s handheld portables enabling them to read the material during the 14-hour flight. The accompanying expert was able to close the performance gaps in a quick hands on session upon arrival in country.

What makes this scenario exciting is that all of the needed knowledge, intelligence, and training material is automatically gathered from disparate databases spread across the military and intelligence services. The material is then organized and packaged by ADL protocols and delivery templates. That is the potential the semantic web has for ADL, and for searching on the Internet more generally.

What holds even greater potential is the power unleashed as data and databases are opened up. Tim Berners-Lee gave a presentation on Linked Data at the TED 2009 conference. If nothing else is accomplished today, spend the 15 minutes required to watch his presentation. He does an outstanding job on explalining how linked data will make an impact on the world.

Is this really possible?

In the case of the World Wide Web “for the semantic web to function, computers must have access to structured collections of information and sets of inference rules that they can use to conduct automated reasoning” (Berners-Lee., Hendler,, & Lassila, 2001). Financially and practically, this would need to be done in a decentralized way, thus complicating the problem exponentially. The semantic web uses URIs (Universal Resource Identifiers) to identify subjects and objects found on the web. This Resource Description Framework (RDF) encodes things in sets of triplets similar to a sentence – subject, verb, object (For an introduction to the mechanics behind the semantic web, I recommend “The Semantic Web: An Introduction” found at http://infomesh.net/2001/swintro/). However, due to their decentralized nature, databases could use different identifiers for the same concept. This led to the use of ontologies that formally describe relationships between objects and terms. One can easily see the complexity of establishing such a framework without centralised control, in unlimited knowledge domains, and on the scale of the WWW. An additional problem is the workforce required to encode the semantic web.

For example, Wikipedia uses a workforce of volunteer contributors that is similar to the workforce that would be required to create a semantic web. These contributors have many different motivations. Articles in the English version of Wikipeadia have reached 2,852,045 as of April 2009. Contributors can easily see the value of their input; there is a high level of reciprocity that is felt immediately. However, in the case of the semantic web, a much greater initial effort would be required before any tangible returns could be gained. Wikipedia has about 75,000 core contributors, but I would guess that the critical mass for the semantic web on a worldwide basis would be several factors greater.

A very interesting open project, Twine (http://www.twine.com) hopes to provide its users with an open and extensive platform for organizing, sharing and discovering information. Twine learns about you as you use it and becomes more useful the more you use it. It connects people with similar interests and is ideal for individuals, groups and communities. It could easily be used by academics and students to keep track of knowledge domains. For example, a class could enter and link all articles and reference material used throughout the semester. They could add reviews, ratings, and tags. This site would be passed to the next class and built upon as new reference material becomes available and different interpretations of older material is recorded. The Twine site (http://www.twine.com/technology) indicates that:

“The Semantic Web is the next step in the evolution of the Internet. But making this technology meaningful and accessible to everyday users is where Twine comes in. Twine is an application that helps people organize, share and discover information around their interests. And while users certainly don’t need to understand the Semantic Web in order to appreciate Twine, several technologies are hard at work behind the scenes of its simple user interface.”

The technologies referred to are RDF and OWL (Web Ontology Language) that define the properties and classes, and determine their uses.

Another open/commercial semantic web project is Calais (http://www.opencalais.com/). The Calais Web Service automatically produces the RDF triplets from any document submitted. It produces named entities, facts, and events from unstructured documents.

“This metadata gives you the ability to build maps (or graphs or networks) linking documents to people to companies to places to products to events to geographies to… whatever. You can use those maps to improve site navigation, provide contextual syndication, tag and organize your content, create structured folksonomies, filter and de-duplicate news feeds, or analyze content to see if it contains what you care about.” (from http://www.opencalais.com/about).

Again, a very powerful use can easily be seen for academics.

While Twine and Calais are starting to make some progress in the WWW namespace, linkages are still limited to the relatively small user populations. Nonetheless, they are excellent tools for personal knowledge management or to bring together the ideas and research of a class or a course. Conversely, in a closed namespace such as ADL, even including a multitude of other pertinent government agency databases, it turns out that the semantic web problems become greatly simplified. Not easy – but simpler. Within military and intelligence knowledge domains, terms and objects tend to have well-defined meanings and uses. This fact allows for significant automation of the process for defining RDF triplets and ontological inferences. Initial work could be contracted and policy could encourage new data entries to be correctly formatted. In such a closed space, critical mass could be reached and the benefits of the semantic web realized within a reasonable time frame. In an academic setting, links could be made by students as part of a requirement of courses – similar to how annotated bibliographies were assigned in the past. Not only would such links benefit the current class, they would be built upon by future classes and shared with others through the WWW.

Conclusion

While a useful implementation of a semantic web has not yet been possible on a large scale, the examples and possibilities presented above are a good start. The ADL repository also presents an interesting possibility due to its limited namespace and the nature of contents of the repository. The speed of information transfer and knowledge acquisition required in today’s complex battle spaces and in academic research could certainly benefit from the possibilities presented in the scenario above. At the very least, current tools such as Twine are ideal for personal knowledge management and in dealing with information overload. In a military application, to be truly useful, the ADL repository would need to be augmented by additional sources of intelligence, research databases, and lessons learned databases. Such an investment would certainly also lead to advancements in the research and development of the semantic web, thus contributing to the civilian development of a semantic web on the WWW.

Future

Bakshi and Karger (no date) discuss semantic web applications that will need to be developed in conjunction with the semantic web in order for users to take advantage of the power inherent in what some are calling Web 3.0. Important as the applications are to future success, having each of us doing our part to contribute to Tim Berners-Lee’s four principles of Linked Data, in his note Design Issues: Linked Data, is even more important. These are paraphrased along the following lines:

• URIs (Uniform Resource Identifiers) to identify things that you expose to the Web as resources.
• HTTP URIs so that people can locate and look up (dereference) these things.
• Useful information about the resource when its URI is dereferenced.
• Links to other, related URIs in the exposed data as a means of improving information discovery on the Web.

With each of us doing our part, we can together create a quantum leap in the power of the web.

References

Bakshi, K., & Karger. D.R. (nd). Semantic web applications. Downloaded 2 April 2009 from MIT Computer Science and Artificial Intelligence Laboratory, http://www.ifi.uzh.ch/ddis/fileadmin/events/iswc2005ws/CameraReady/
Bakshi_SWApplications_019.pdf

Berners-Lee , T. (1998). Semantic Web Road map. Downloaded 6 April 2009 from http://www.w3.org/DesignIssues/Semantic.html

Berners-Lee. T., Hendler, T., & Lassila, O. (2001). The Semantic Web. A new form of web that is meaningful to computers will unleash a revolution of new possibilities. Scientific American Magazine. May 17, 2001. Downloaded 2 April 2009 from http://www.sciam.com/article.cfm?id=the-semantic-web

Dodds, P. V. W., & Fletcher, J. D. (2004) Opportunities for new “smart” learning environments enabled by next generation web capabilities. Journal of Education Multimedia and Hypermedia, 13(4), 391-404.