|
|
|
|
|
|
Frequently customers ask me why semantic technologies should be used instead of conventional information technologies. In the areas of knowledge representation (KR) and knowledge management (KM), there are compelling reasons and benefits for selecting semantic technologies over conventional approaches. This article attempts to summarize these rationales from a layperson perspective.
It is important to recognize that semantic technologies are orthogonal to the buzz around some other current technologies, including cloud computing and big data. Semantic technologies are also not limited to open data: they are equivalently useful to private or proprietary data. It is also important to note that semantic technologies do not imply some grand, shared schema for organizing all information. Semantic technologies are not “one ring to rule them all,” but rather a way to capture the world views of particular domains and groups of stakeholders. Lastly, semantic technologies done properly are not a replacement for existing information technologies, but rather an added layer that can leverage those assets for interoperability and to overcome the semantic barriers between existing information silos.
The world is a messy place. Not only is it complicated and richly diverse, but our ways of describing and understanding it are made more complex by differences in language and culture.
We also know the world to be interconnected and interdependent. Effects of one change can propagate into subtle and unforeseen effects. And, not only is the world constantly changing, but so is our understanding of what exists in the world and how it affects and is affected by everything else.
This means we are always uncertain to a degree about how the world works and the dynamics of its working. Through education and research we continually strive to learn more about the world, but often in that process find what we thought was true is no longer so and even our own human existence is modifying our world in manifest ways.
Knowledge is very similar to this nature of the world. We find that knowledge is never complete and it can be found anywhere and everywhere. We capture and codify knowledge in structured, semi-structured and unstructured forms, ranging from “soft” to “hard” information. We find that the structure of knowledge evolves with the incorporation of more information.
We often see that knowledge is not absolute, but contextual. That does not mean that there is no such thing as truth, but that knowledge should be coherent, to reflect a logical consistency and structure that comports with our observations about the physical world. Knowledge, like the world, is constantly changing; we thus must constantly adapt to what we observe and learn.
These observations about the world and knowledge are not platitudes but important guideposts for how we should organize and manage information, the field known as “information technology.” For IT to truly serve the knowledge function, its logical bases should be consistent with the inherent nature of the world and knowledge.
By knowledge functions we mean those areas of various computer applications that come under the rubrics of search, business intelligence, competitive intelligence, planning, forecasting, data federation, data warehousing, knowledge management, enterprise information integration, master data management, knowledge representation, and so forth. These applications are distinctly different than the earliest and traditional concerns of IT systems: accounting and transactions.
A transaction system — such as calculating revenue based on seats on a plane, the plane’s occupancy, and various rate classes — is a closed system. We can count the seats, we know the number of customers on board, and we know their rate classes and payments. Much can be done with this information, including yield and profitability analysis and other conventional ways of accounting for costs or revenues or optimizations.
But, as noted, neither the world nor knowledge is a closed system. Trying to apply legacy IT approaches to knowledge problems is fraught with difficulties. That is the reason that for more than four decades enterprises have seen massive cost overruns and failed projects in applying conventional IT approaches to knowledge problems: traditional IT is fundamentally mismatched to the nature of the problems at hand.
What works efficiently for transactions and accounting is a miserable failure applied to knowledge problems. Traditional relational databases work best with structured data; are inflexible and fragile when the nature (schema) of the world changes; and thus require constant (and expensive) re-architecting in the face of new knowledge or new relationships.
Of course, often knowledge problems do consider fixed entities with fixed attributes to describe them. In these cases, relational data systems can continue to act as valuable contributors and data managers of entities and their attributes. But, in the role of organizing across schema or dealing with semantics and differences of definition and scope – that is, the common types of knowledge questions – a much different integration layer with a much different logic basis is demanded.
The first change that is demanded is to shift the logic paradigm of how knowledge and the world are modeled. In contrast to the closed-world approach of transaction systems, IT systems based on the logical premise of the open world assumption (OWA) mean:
Much more can be said about OWA, including formal definitions of the logics underlying it [1], but even from the statements above, we can see that the right logic for most knowledge representation (KR) problems is the open world approach.
This logic mismatch is perhaps the most fundamental cause of failures, cost overruns, and disappointing deliverables for KM and KR projects over the years. But, like the fingertip between the eyes that cannot be seen because it is too close at hand, the importance of this logic mismatch strangely continues to be overlooked.
Data exists in many forms and of many natures. As one classification scheme, there are:
Further, these types of data may be “soft”, such as social information or opinion, or “hard”, more akin to measurable facts or quantities.
These various forms may also be serialized in a variety of data formats or data transfer protocols, some using straight text with a myriad of syntax or markup vocabularies, ranging to scripts or forms encoded or binary.
Still further, any of these data forms may be organized according to a separate schema that describes the semantics and relationships within the data.
These variations further complicate the inherently diverse nature of the world and knowledge of it. A suitable data model for knowledge representation must therefore have the power to be able to capture the form, format, serialization or schema of any existing data within the diversity of these options.
The Resource Description Framework (RDF) data model has such capabilities [2]. Any extant data form or schema (from the simple to the complex) can be converted to the RDF data model. This capability enables RDF to act as a “universal solvent” for all information.
Once converted to this “canonical” form, RDF can then act as a single representation around which to design applications and other converters (for “round-tripping” to legacy systems, for example), as illustrated by this diagram:
Generic tools can then be driven by the RDF data model, which leads to fewer applications required and lower overall development costs.
Lastly, RDF can represent simple assertions (“Jane runs fast”) to complex vocabularies and languages. It is in this latter role that RDF can begin to represent the complexity of an entire domain via what is called an “ontology” or “knowledge graph.”
When representing knowledge, more things and concepts get drawn into consideration. In turn, the relationships of these things lead to connections between them to capture the inherent interdependence and linkages of the world. As still more things get considered, more connections are made and proliferate.
This process naturally leads to a graph structure, with the things in the graphs represented as nodes and the relationships between them represented as connecting edges. More things and more connections lead to more structure. Insofar as this structure and its connections are coherent, the natural structure of the knowledge graph itself can help lead to more knowledge and understanding.
How one such graph may emerge is shown by this portion of the recently announced Google Knowledge Graph [3], showing female Nobel prize winners:
Unlike traditional data tables, graphs have a number of inherent benefits, particularly for knowledge representations. They provide:
Graphs are the natural structures for knowledge domains.
Once built, graphs offer some analytical capabilities not available through traditional means of information structure. Graph analysis is a rapidly emerging field, but already some unique measures of knowledge domains are now possible to gauge:
As science is coming to appreciate, graphs can represent any extant structure or schema. This gives graphs a universal character in terms of analytic tools. Further, many structures can only be represented by graphs.
The nature of knowledge is such that relevant information is everywhere. Further, because of the interconnectedness of things, we can also appreciate that external information needs to be integrated with internal information. Meanwhile, the nature of the world is such that users and stakeholders may be anywhere.
These observations suggest a knowledge representation architecture that needs to be truly distributed. Both sources and users may be found in multiple locations.
In order to preserve existing information assets as much as possible (see further below) and to codify the earlier observation regarding the broad diversity of data formats, the resulting knowledge architecture should also attempt to put in place a thin layer or protocol that provides uniform access to any source or target node on the physical network. A thin, uniform abstraction layer – with appropriate access rights and security considerations – means knowledge networks may grow and expand at will at acceptable costs with minimal central coordination or overhead.
Properly designed, then, such architectures are not only necessary to represent the distributed nature of users and knowledge, but can also facilitate and contribute to knowledge development and exchange.
The items above suggest the Web as an appropriate protocol for distributed access and information exchange. When combined with the following considerations, it becomes clear that the Web is the perfect medium for knowledge networks:
It is not surprising that the largest extant knowledge networks on the globe – such as Google, Wikipedia, Amazon and Facebook – are Web-based. These pioneers have demonstrated the wisdom of WOA for cost-effective scalability and universal access.
Also, the combination of RDF with Web identifiers also means that any and all information from a given knowledge repository may be exposed and made available to others as linked data. This approach makes the Web a global, universal database. And it is in keeping with the general benefits of integrating external information sources.
Existing IT assets represent massive sunk costs, legacy knowledge and expertise, and (often) stakeholder consensus. Yet, these systems are still largely stovepiped.
Strategies that counsel replacement of existing IT systems risk wasting existing assets and are therefore unlikely to be adopted. Ways must be found to leverage the value already embodied in these systems, while promoting interoperability and integration.
The beauty of semantic technologies – properly designed and deployed in a Web-oriented architecture – is that a thin interoperability layer may be placed over existing IT assets to achieve these aims. The knowledge graph structure may be used to provide the semantic mappings between schema, while the Web service framework that is part of the WOA provides the source conversion to the canonical RDF data model.
Via these approaches, prior investments in knowledge, information and IT assets may be preserved while enabling interoperability. The existing systems can continue to provide the functionality for which they were originally designed and deployed. Meanwhile, the KR-related aspects may be exposed and integrated with other knowledge assets on the physical network.
These kinds of approaches represent a fundamental shift in power and roles with respect to IT in the enterprise. IT departments and their bottlenecks in writing queries and bespoke application development can now be bypassed; the departments may be relegated to more appropriate support roles. Developers and consultants can now devote more of their time to developing generic applications driven by graph structures [4].
In turn, the consumers of knowledge applications – namely subject matter experts, employees, partners and stakeholders – now become the active contributors to the graphs themselves, focusing on reconciling terminology and ensuring adequate entity and concept coverage. Knowledge graphs are relatively straightforward structures to build and maintain. Those that rely on them can also be those that have the lead role in building and maintaining them.
Thus, graph-driven applications can be made generic by function with broader and more diverse information visualization capabilities. Simple instructions in the graphs can indicate what types of information can be displayed with what kind of widget. Graph-driven applications also mean that those closest to the knowledge problems will also be those directly augmenting the graphs. These changes act to democratize the knowledge function, and lower overall IT costs and risks.
Elsewhere we have discussed the specific components that go into enabling the development of a semantic enterprise, what we have termed the seven pillars [5]. Most of these points have been covered to one degree or another in the discussion above.
There are off-the-shelf starter kits for enterprises to embrace to begin this process. The major starting requirements are to develop appropriate knowledge graphs (ontologies) for the given domain and to convert existing information assets into appropriate interoperable RDF form.
Beyond that, enterprise staff may be readily trained in the use and growth of the graphs, and in the staging and conversion of data. With an appropriate technology transfer component, these semantic technology systems can be maintained solely by the enterprise itself without further outside assistance.
Unlike conventional IT systems with their closed-world approach, semantic technologies that adhere to these guidelines can be deployed incrementally at lower cost and with lower risk. Further, we have seen that semantic technologies offer an excellent integration approach, with no need to re-do schema because of changed circumstances. The approach further leverages existing information assets and brings the responsibility for the knowledge function more directly to its users and consumers.
Semantic technologies are thus well-suited for knowledge applications. With their graph structures and the ability to capture semantic differences and meanings, these technologies can also accommodate multiple viewpoints and stakeholders. There are also excellent capabilities to relate all available information – from documents and images and metadata to tables and databases – into a common footing.
These advantages will immediately accrue through better integration and interoperability of diverse information assets. But, for early adopters, perhaps the most immediate benefit will come from visible leadership in embracing these enabling technologies in advance of what will surely become the preferred approach to knowledge problems.
Note: there is a version of this article on Slideshare:
Since its first release four years ago, the History of Information Timeline has been one of the most consistently popular aspects of this site. It is an interactive timeline of the most significant events and developments in the innovation and management of information and documents throughout human history.
Recent requests for use by others and my own references to it caused me to review its entries and add to it. Over the past few weeks I have expanded its coverage by some 30%. There are now about 115 entries in the period ranging from ca 30,000 BC (cave paintings) to ca 2003 AD (3D printing). Most additions have been to update the past twenty or thirty years:
The timeline works via a fast scroll at the bottom; every entry when clicked produces a short info-panel, as shown.
All entries are also coded by a icon scheme of about 20 different categories:
 |
Book Forms and Bookmaking |  |
Calendars |  |
Copyrights and Legal |  |
Infographics and Statististics |
 |
Libraries |  |
Maps |  |
Math and Symbology |  |
Mechanization |
 |
Networks |  |
New Formats or Document Forms |  |
Organizing Information |  |
Pre-writing |
 |
Paper and Papermaking |  |
Printing |  |
Science and Technology |  |
Scripts and Alphabets |
 |
Standardization and Typography |  |
Theory |  |
Timelines |  |
Writing |
You may learn more about this timeline and the technology behind it by referring to the original announcement.
One of the main reasons I am such a big fan of RDF as a canonical data model is its ability to capture information in structured, semi-structured and unstructured form [1]. These sources are conventionally defined as:
A major trend I have written about for some time is the emergence of the structured Web: that is, the exposing of structure from these varied sources in order for more information to be interconnected and made interoperable. I have posited — really a view shared by many — that the structured Web is an intermediate point in the evolution of the Web from one of documents to one where meaningful semantics occurs [2].
It is clear in my writings — indeed in the very name of my company, Structured Dynamics — that structure plays a major role in our thinking. The use and reliance on this term, though, begs the question: just what is structure in an informational sense? We’ll find it helpful to get at the question of What is structure? from a basis using first principles. And this, in turn, may also provide insight into how structure and information are in fact inextricably entwined.
According to Wikipedia, structure is a fundamental notion, of tangible or intangible character, that refers to the recognition, observation, nature, or permanence of patterns and relationships of entities. The concept may refer to an object, such as a built structure, or an attribute, such as the structure of society.
Structure may be abstract, or it may be concrete. Its realm ranges from the physical to ideas and concepts. As a term, “structure” seems to be ubiquitous to every domain. Structure may be found across every conceivable scale, from the most minute and minuscule to the cosmic. Even realms without any physical aspect at all — such as ideas and languages and beliefs — are perceived by many to have structure. We apply the term to any circumstance in which things are arranged or connected to one another, as a means to describe the organization or relationships of things. We seem to know structure when we see it, and to be able to discern structure of very many kinds against unstructured or random backgrounds.
In this way structure quite resembles patterns, perhaps could even be used synonymously with that term. Other closely related concepts include order, organization, design or form. When expressed, structure, particularly that of a recognizably ordered or symmetrical nature, is often called beautiful.
One aspect of structure, I think, that provides the key to its roles and importance is that it can be expressed in shortened form as a mathematical statement. One could even be so bold as to say that mathematics is the language of structure. This observation is one of the threads that will help us tie structure to information.
The natural world is replete with structure. Patterns in nature are regularities of visual form found in the natural world. Each such pattern can be modeled mathematically. Typical mathematical forms in nature include fractals, spirals, flows, waves, lattices, arrays, Golden ratios, tilings, Fibonacci sequences, and power laws. We see them in such structures as clouds, trees, leaves, river networks, fault lines, mountain ranges, craters, animal spots and stripes, shells, lightning bolts, coastlines, flowers, fruits, skeletons, cracks, growth rings, heartbeats and rates, earthquakes, veining, snow flakes, crystals, blood and pulmonary vessels, ocean waves, turbulence, bee hives, dunes and DNA.
The mathematical expression of structures in nature is frequently repeated or recursive in nature, often in a self-organizing manner. The swirls of a snail’s shell reflect a Fibonacci sequence, while natural landscapes or lifeforms often have a fractal aspect (as expressed by some of the figures in this article). Fractals are typically self-similar patterns, generally involving some fractional or ratioed formula that is recursively applied. Another way to define it is as a detailed pattern repeating itself.
Even though these patterns can often be expressed simply and mathematically, and they often repeat themselves, their starting conditions can lead to tremendous variability and a lack of predictability. This makes them chaotic, as studied under chaos theory, though their patterns are often discernible.
While we certainly see randomness in statistics, quantum physics and Brownian motion, it is also striking how what gives nature its beauty is structure. As a force separate and apart from the random, there appears to be something in structure that guides the expression of what is nature and what is so pleasing to behold. Self-similar and repeated structures across the widest variety of spatial scales seems to be an abiding aspect of nature.
Such forms of repeated patterns or structure are also inherent in that most unique of human capabilities, language. As a symbolic species [3], we first used symbols as a way to represent the ideas of things. Simple markings, drawings and ideograms grew into more complicated structures such as alphabets and languages. The languages themselves came to embrace still further structure via sentence structures, document structures, and structures for organizing and categorizing multiple documents. In fact, one of the most popular aspects of this blog site is its Timeline of Information History — worth your look — that shows the progression of structure in information throughout human history.
Grammar is often understood as the rules or structure that governs language. It is composed of syntax, including punctuation, traditionally understood as the sentence structure of languages, and morphology, which is the structural understanding of a language’s linguistic units, such as words, affixes, parts of speech, intonation or context. There is a whole field of linguistic typology that studies and classifies languages according to their structural features. But grammar is hardly the limit to language structure.
Semantics, the meaning of language, used to be held separate from grammar or structure. But via the advent of the thesaurus, and then linguistic databases such as WordNet and more recently concept graphs that relate words and terms into connected understandings, we also have now come to understand that semantics also has structure. Indeed, these very structural aspects are now opening up to us techniques and methods — generally classified under the heading of natural language processing (NLP) — for extracting meaningful structure from the very basis of written or spoken language.
It is the marriage of the computer with language that is illuminating these understandings of structure in language. And that opening, in turn, is enabling us to capture and model the basis of human language discourse in ways that can be codified, characterized, shared and analyzed. Machine learning and processing is now enabling us to complete the virtual circle of language. From its roots in symbols, we are now able to extract and understand those very same symbols in order to derive information and knowledge from our daily discourse. We are doing this by gleaning the structure of language, which in turn enables us to relate it to all other forms of structured information.
The continuation of structure from nature to language extends across all aspects of human endeavor. I remember excitedly describing to a colleague more than a decade ago what likely is a pedestrian observation: pattern matching is a common task in many fields. (I had observed that pattern matching in very different forms was a standard practice in most areas of industry and commerce.) My “insight” was that this commonality was not widely understood, which meant that pattern matching techniques in one field were not often exploited or seen as transferable to other domains.
In computer science, pattern matching is the act of checking some sequence of tokens for the presence of the constituents of some pattern. It is closely related to the idea of pattern recognition, which is the characterization of some discernible and repeated sequence. These techniques, as noted, are widely applied, with each field tending to have its own favorite algorithms. Common applications that one sees for such pattern-based calculations include communications [4], encoding and coding theory, file compression, data compression, machine learning, video compression, mathematics (including engineering and signal processing via such techniques as Fourier transforms), cryptography, NLP [5], speech recognition, image recognition, OCR, image analysis, search, sound cleaning (that is, error detection, such as Dolby) and gene sequence searching and alignment, among many others.
To better understand what is happening here and the commonalities, let’s look at the idea of compression. Data compression is valuable for transmitting any form of content in wired or wireless manners because we can transmit the same (or closely similar) message faster and with less bandwidth [6]. There are both lossless (no loss of information) and lossy compression methods. Lossless data compression algorithms usually exploit statistical redundancy — that is, a pattern match — to represent data more concisely without losing information. Redundancy in information theory is the number of bits used to transmit a message minus the number of bits of actual information in the message. Lossless compression is possible because most real-world data has statistical redundancy. In lossy data compression, some loss of information is acceptable by dropping detail from the data to save storage space. These methods are guided by research that indicates, say, how certain frequencies may not be heard or seen by people and can be removed from the source data.
On a different level, there is a close connection between machine learning and compression: a system that predicts the posterior probabilities of a sequence given its entire history can be used for optimal data compression (by using arithmetic coding on the output distribution), while an optimal compressor can be used for prediction (by finding the symbol that compresses best, given the previous history). This equivalence has been used as justification for data compression as a benchmark for “general intelligence.” On a still different level, one major part of cryptography is the exact opposite of these objectives: constructing messages that pattern matching fails against or is extremely costly or time-consuming to analyze.
When one stands back from any observable phenomena — be it natural or human communications — we can see that the “information” that is being conveyed often has patterns, recursion or other structure that enables it to be represented more simply and compactly in mathematical form. This brings me back to my two favorite protagonists in my recent writings — Claude Shannon and Charles S. Peirce.
Claude Shannon‘s seminal work in 1948 on information theory dealt with the amount of information that could be theoretically and predictably communicated between a sender and a receiver [7] [8]. No context or semantics were implied in this communication, only the amount of information (for which Shannon introduced the term “bits”) and what might be subject to losses (or uncertainty in the accurate communication of the message). In this regard, what Shannon called “information” is what we would best term “data” in today’s parlance.
The context of Shannon’s paper and work by others preceding him was to understand information losses in communication systems or networks. Much of the impetus for this came about because of issues in wartime communications and early ciphers and cryptography and the emerging advent of digital computers. But the insights from Shannon’s paper also relate closely to the issues of data patterns and data compression.
A key measure of Shannon’s theory is what he referred to as information entropy, which is usually expressed by the average number of bits needed to store or communicate one symbol in a message. Entropy quantifies the uncertainty involved in predicting the value of a random variable. The Shannon entropy measure is actually a measure of the uncertainty based on the communication (transmittal) between a sender and a receiver; the actual information that gets transmitted and predictably received was formulated by Shannon as R, which can never be zero because all communication systems have losses.
A simple intuition can show how this formulation relates to patterns or data compression. Let’s take a message of completely random digits. In order to accurately communicate that message, all digits (bits) would have to be transmitted in their original state and form. Absolutely no compression of this message is possible. If, however, there are patterns within the message (which, of course, now ceases to make the message random), these can be represented algorithmically in shortened form, so that we only need communicate the algorithm and not the full bits of the original message. If this “compression” algorithm can then be used to reconstruct the bit stream of the original message, the data compression method is deemed to be lossless. The algorithm so derived is also the expression of the pattern that enabled us to compress the message in the first place (such as a*2+1).
We can apply this same type of intuition to human language. In order to improve communication efficiency, the most common words (e.g., “a”, “the”, “I”) should be shorter than less common words (e.g., “disambiguation”, “semantics”, “ontology”), so that sentences will not be too long. As they are. This is an equivalent principal to data compression. In fact, such repeats and patterns apply to the natural world as well.
Shannon’s idea of information entropy has come to inform the even broader subject of entropy in physics and the 2nd Law of Thermodynamics [10]. According to Koelman, “the entropy of a physical system is the minimum number of bits you need to fully describe the detailed state of the system.” Very random (uncertain) states have high entropy, patterned ones low entropy. As I noted recently, in open systems, structures (patterns) are a means to speed the tendency to equilibrate across energy gradients [8]. This observation helps provide insight into structure in natural systems, and why life and human communications tend toward less randomness. Structure will always continue to emerge because it is adaptive to speed the deltas across these gradients; structure provides the fundamental commonality between biological information (life) and human information.
In the words of Thomas Schneider [11], “Information is always a measure of the decrease of uncertainty at a receiver.” Of course, in Shannon’s context, what is actually being measured here is data (or bits), not information embodying any semantic meaning or context. Thus, the terminology may not be accurate for discussing “information” in a contemporary sense. But it does show that “structure” — that is, the basis for shortening the length of a message while still retaining its accuracy — is information (in the Shannon context). In this information there is order or patterns, often of a hierarchical or fractal or graph nature. Any structure that emerges that is better able to reduce the energy gradient faster will be favored according to the 2nd Law.
The data that constitutes “information” in the Shannon sense still lacks context and meaning. In communications terms, it is data; it has not yet met the threshold of actionable information. It is in this next step that we can look to Charles Sanders Peirce (1839 – 1914) for guidance [9].
The core of Peirce’s world view is based in semiotics, the study and logic of signs. In his seminal writing on this, “What is in a Sign?” [10], he wrote that “every intellectual operation involves a triad of symbols” and “all reasoning is an interpretation of signs of some kind”. A sign of an object leads to interpretants, which, as signs, then lead to further interpretants. Peirce’s triadic logic of signs in fact is a taxonomy of sign relations, in which signs get reified and expanded via still further signs, ultimately leading to communication, understanding and an approximation of “canonical” truth. Peirce saw the scientific method as itself an ultimate example of this process. The key aspect of signs for Peirce is the ongoing process of interpretation and reference to further signs.
Ideograms leading to characters, that get combined into sentences and other syntax, and then get embedded within contexts of shared meanings show how these structures compound themselves and lead to clearer understandings (that is, accurate messages) in the context of human languages. While the Shannon understanding of “information” lacked context and meaning, we can see how still higher-order structures may be imposed through these reifications of symbols and signs that improve the accuracy and efficiency of our messages. Though Peirce did not couch his theory of semiosis on structure nor information, we can see it as a natural extension of the same structural premises in Shannon’s formulation.
In fact, today, we now see the “structure” in the semantic relationships of language through the graph structures of ontologies and linguistic databases such as WordNet. The understanding and explication of these structures are having a similarly beneficial effect on how still more refined and contextual messages can be composed, transmitted and received. Human-to-machine communications is (merely!) the challenge of codifying and making explicit the implicit structures in our languages.
The Peirceian ideas of interpretation (context) and compounding and reifying structure are a major intellectual breakthrough for extending the Shannon “information” theory to information in the modern sense. These insights also occur within a testable logic for how things and the names of things can be understood and related to one another, via logical statements or structures. These, in turn, can be symbolized and formalized into logical constructs that can capture the structure of natural language as well as more structured data (or even nature, as some of the earlier Peirce speculation asserts [13]).
According to this interpretation of Peirce, the nature of information is the process of communicating a form from the object to the interpretant through the sign [14]. The clarity of Peirce’s logic of signs is an underlying factor, I believe, for why we are finally seeing our way clear to how to capture, represent and relate information from a diversity of sources and viewpoints that is defensible and interoperable.
Common to all of these perspectives — from patterns to nature and on to life and then animal and human communications — we see that structure is information. Human artifacts and technology, though not “messages” in a conventional sense, also embody the information of how they are built within their structures [15]. We also see the interplay of patterns and information in many processes of the natural world [16] from complexity theory, to emergence, to autopoiesis, and on to autocatalysis, self-organization, stratification and cellular automata [17]. Structure in its many guises is ubiquitous.
We, as beings who can symbolically record our perceptions, seem to innately recognize patterns. We see beauty in symmetry. Bilateral symmetry seems to be deeply ingrained in the inherent perception by humans of the likely health or fitness of other living creatures. We see beauty in the patterned, repeated variability of nature. We see beauty in the clever turn of phrase, or in music, or art, or the expressiveness of a mathematical formulation.
We also seem to recognize beauty in the simple. Seemingly complex bit streams that can be reduced to the short, algorithmic expression are always viewed as more elegant than lengthier, more complex alternatives. The simple laws of motion and Newtonian physics fit this pattern, as does Einstein’s E=mc2. This preference for the simple is a preference for the greater adaptiveness of the shorter, more universal pattern to messages, an insight indicated by Shannon’s information theory.
In the more prosaic terms of my vocation in the Web and information technology, these insights point to the importance of finding and deriving structured representations of information — including meaning (semantics) — that can be simply expressed and efficiently conveyed. Building upon the accretions of structure in human and computer languages, the semantic Web and semantic technologies offer just such a prospect. These insights provide a guidepost for how and where to look for the next structural innovations. We find them in the algorithms of nature and language, and in making connections that provide the basis for still more structure and patterned commonalities.
Ideas and algorithms around loseless compression and graph theory and network analysis are, I believe, the next fruitful hunting grounds for finding still higher-orders of structure, which can be simply expressed. The patterns of nature, which have emerged incrementally and randomly over the eons of cosmological time, look to be an excellent laboratory.
So, as we see across examples from nature and life to language and all manner of communications, information is structure and structure is information. And it is simply beautiful.
Modularization Also Leads to Big Graph VisualizationWe are pleased to announce the release of version 1.05 of UMBEL, which now has linkages to schema.org [6] and GeoNames [1]. UMBEL has also been split into ‘core’ and ‘geo’ modules. The resulting smaller size of UMBEL ‘core’ — now some 26,000 reference concepts — has also enabled us to create a full visualization of UMBEL’s content graph.
The first notable change in UMBEL v. 1.05 is its mapping to schema.org. schema.org is a collection of schema (usable as HTML tags) that webmasters can use to markup their pages in ways recognized by major search providers. schema.org was first developed and organized by the major search engines of Bing, Google and Yahoo!; later Yandex joined as a sponsor. Now many groups are supporting schema.org and contributing vocabularies and schema.
I was one of the first to hail schema.org hours after its announcement [7]. It seemed only fair that we put our money where our mouth is and map UMBEL to it as well.
The UMBEL-schema.org mapping was manually done by, firstly, searching and inspecting the current UMBEL concept base for appropriate matches. If that mapping failed to find a rather direct correspondence between existing UMBEL concepts and the types in schema.org, the source concept reference of OpenCyc was then inspected in a similar manner. Failing a match from either of these two sources, the decision was to add a new concept to the ‘core’ UMBEL. This new concept was then appropriately placed into the UMBEL reference concept subject structure.
The net result of this process was to add 298 mapped schema.org types to UMBEL. This mapping required a further three concepts from OpenCyc, and a further 78 new reference concepts, to be added to UMBEL. Along with the new updates to UMBEL and its mappings, the section of Key Files below provides further explanatory links. We are reserving the addition of schema.org properties for a later time, when we plan to re-organize the Attributes SuperType within UMBEL.
Even in the early development of UMBEL there was a tension about the scope and level of what geographic information to include in its concept base. The initial decision was to support country and leading-country province and state concepts, and some leading cities. This decision was in the spirit of a general reference structure, but still felt arbitrary.
GeoNames is devoted to geographical information and concepts — both natural and human artifacts — and has become the go-to resource for geo-locational information. The decision was thus made to split out the initial geo-locational information in UMBEL and replace it with mappings to GeoNames. This decision also had the advantage of beginning a process of modularization of UMBEL. 
Two sets of reference concepts were identified as useful for splitting out from the ‘core’ UMBEL in a geo-locational aspect:
These removed concepts were then placed into a separate ‘geo’ module of UMBEL, including all existing annotations and relations, resulting in a module of 1,854 concepts. That left 26,046 concepts in UMBEL ‘core’. Because of some shared parent concepts, there is some minor overlap between the two modules. These are now the modular splits in UMBEL version 1.05.
GeoNames has a different structure to UMBEL. It has few classes and distinguishes its geographic information on the basis of some 671 feature codes. These codes span from geopolitical divisions — such as countries, states or provinces, cities, or other administrative districts — to splits and aggregations by natural and human features. Types of physical terrain — above ground and underwater — are denoted, as well as regions and landscape features governed by human activities (such as vineyards or lighthouses) [1]. We wanted to retain this richness in our mappings.
We needed a bridge between feature codes and classes, a sort of umbrella property generally equivalent to owl:sameAs in nature, but with some possible inexactitude or degree of approximation. The appropriate choice here is umbel:correspondsTo, which was designed specifically for this purpose [2]. This predicate is thus the basis for the mappings.
The 671 GeoNames feature codes were manually mapped to corresponding classes in the UMBEL concepts, in a manner identical to what was described for schema.org above. The result was to add another further three OpenCyc concepts and to add 88 new UMBEL reference concepts to accommodate the full GeoNames feature codes. We thus now have a complete representation of the full structure and scope of GeoNames in UMBEL.
There are three modes in which one can now work with UMBEL:
In the latter case, you may use SPARQL queries with the umbel:correspondsTo predicate to achieve the desired retrievals. If more logic is required, you will likely need to look to a rules-based addition such as SWRL [3] or RIF [4] to capture the umbel:correspondsTo semantics.
Because of the UMBEL modularization, it has now become tractable to graph the main ontology in its entirety. The core UMBEL ontology contains about 26,000 reference concepts organized according to 33 super types. There are more than 60,000 relationships amongst these concepts, resulting in a graph structure of very large size.
It is difficult to grasp this graph in the abstract. Thus, using methods earlier described in our use of the Gephi visualization software [5], we present below a dynamic, navigable rendering of this graph of UMBEL core:
Note: at standard resolution, if this graph were to be rendered in actual size, it would be larger than 34 feet by 34 feet square at full zoom !!! Hint: that is about 1200 square feet, or 1/2 the size of a typical American house !
This UMBEL graph displays:
You may also want to inspect a static version of this big graph by downloading a PDF.
Lastly, we fully updated the UMBEL Web site and re-released the UMBEL wiki.
x:coref predicate from the UMBC Ebiquity group; see further Jennifer Sleeman and Tim Finin, 2010. “Learning Co-reference Relations for FOAF Instances,” Proceedings of the Poster and Demonstration Session at the 9th International Semantic Web Conference, November 2010; see http://ebiquity.umbc.edu/_file_directory_/papers/522.pdf. In the words of Tim Finin of the Ebiquity group:
owl:sameAs may lead to contradictions. However, virtually merging the descriptions in a co-reference engine is fine — both provide information that is useful in disambiguating future references as well as for many other purposes. Our property (:coref) is a transitive, symmetric property that is a super-property of owl:sameAs and is paired with another, :notCoref that is symmetric and generalizes owl:differentFrom.When we look at the analog properties noted above, we see that the property objects tend to share reflexivity, symmetry and transitivity. We specifically designed the umbel:correspondsTo predicate to capture these close, nearly equivalent, but uncertain degree of relationships.
Some Quick Investigations Point to Promise, DisappointmentsIt has been clear for some time that Google has been assembling a war chest of entities and attributes. It first began to appear as structured results in its standard results listings, a trend I commented upon more than three years ago in Massive Muscle on the ABox at Google. Its purchase of Metaweb and its Freebase service in July 2010 only affirmed that trend.
This week, perhaps a bit rushed due to attention to the Facebook IPO, Google announced its addition of the Knowledge Graph (GKG) to its search results. It has been releasing this capability in a staged manner. Since I was fortunately one of the first to be able to see these structured results (due to luck of the draw and no special “ins”), I have spent a bit of time deconstructing what I have found.
What you get (see below) when you search on particular kinds of entities is in essence an “infobox“, similar to the same structure as what is found on Wikipedia. This infobox is a tabular presentation of key-value pairs, or attributes, for the kind of entity in the search. A ‘people’ search, for example, turns up birth and death dates and locations, some vital statistics, spouse or famous relations, pictures, and links to other relevant structured entities. The types of attributes shown vary by entity type. Here is an example for Stephen King, the writer (all links from here forward provide GKB results), which shows the infobox and its key-value pairs in the righthand column: 
Reportedly these results are drawn from Freebase, Wikipedia, the CIA World Factbook and other unidentified sources. Some of the results may indeed be coming from Freebase, but I saw none as such. Most entities I found were from Wikipedia, though it is important to know that Freebase in its first major growth incarnation springboarded from a Wikipedia baseline. These early results may have been what was carried forward (since other contributed data to Freebase is known to be of highly variable quality).
The entity coverage appears to be spotty and somewhat disappointing in this first release. Entity types that I observed were in these categories:
|
|
|
Entity types that I expected to see, but did not find include:
|
|
|
This is clearly not rigorous testing, but it would appear that entity types along the lines of what is in schema.org is what should be expected over time.
I have no way to gauge the accuracy of Google’s claims that it has offered up structured data on some 500 million entities (and 3.5 billion facts). However, given the lack of coverage in key areas of Wikipedia (which itself has about 3 million entities in the English version), I suspect much of that number comes from the local businesses and restaurants and such that Google has been rapidly adding to its listings in recent years. Coverage of broadly interesting stuff still seems quite lacking.
The much-touted knowledge graph is also kind of disappointing. Related links are provided, but they are close and obvious. So, an actor will have links to films she was in, or a person may have links to famous spouses, but anything approaching a conceptual or knowledge relationship is missing. I think, though, we can see such links and types and entity expansions to steadily creep in over time. Google certainly has the data basis for making these expansions. And, constant, incremental improvement has been Google’s modus operandi.
For some time, and at various meetings I attend, I have always been at pains to question Google representatives whether there is some unique, internal ID for entities within its databases. Sometimes the reps I have questioned just don’t know, and sometimes they are cagey.
But, clearly, anything like the structured data that Google has been working toward has to have some internal identifier. To see if some of this might now have surfaced with the Knowledge Graph, I did a bit of poking of the URLs shown in the searches and the affiliated entities in the infoboxes. Under most standard searches, the infobox appears directly if there is one for that object. But, by inspecting cross-referenced entities from the infoboxes themselves, it is possible to discern the internal key.
The first thing one can do in such inspections is to remove that stuff that is local or cookie things related to one’s own use preferences or browser settings. Other tests can show other removals. So, using the Stephen King example above, we can eliminate these aspects of the URL:
This actually conformed to my intuition, because the ‘&stick’ aspect was a new parameter for me. (Typically, in many of these dynamic URLs, the various parameters are separated by one another by a set designator character. In the case of Google, that is the ampersand &.)
By simply doing repeated searches that result in the same entity references, I was able to confirm that the &stick parameter is what invokes the unique ID and infobox for each entity. Further, we can decompose that further, but the critical aspect seems to be what is not included within the following: &stick=H4sIAAAAAAAAAONg . . [VuLQz9U3] . . AAAA. The stuff in the brackets varies less, and I suspect might be related to the source, rather than the entity.
I started to do some investigation on types and possible sources, but ran out of time. Listed below are some &stick identifiers for various types of entities (each is a live link):
| Type | Entity | Infobox Identifier |
| Place | Los Angeles | &stick=H4sIAAAAAAAAAONgVuLSz9U3MDYoTDIuAQD9ON7KDgAAAA |
| Place | Brentwook | &stick=H4sIAAAAAAAAAONgVuLUz9U3MKwsNEkDAN6nm-sNAAAA |
| Person | Arthur Miller | &stick=H4sIAAAAAAAAAONgVmLXz9U3qEwrAADRsxaaCwAAAA |
| Person | Joe DiMaggio | &stick=H4sIAAAAAAAAAONgVuLQz9U3SDFMqwAAAy8zdQwAAAA |
| Person | Marilyn Monroe | &stick=H4sIAAAAAAAAAONgVuLQz9U3MCkvLAIAW7x_LwwAAAA |
| Movie | Shawshank Redemption | &stick=H4sIAAAAAAAAAONgVuLQz9U3MM_KKwEAPYgNDQwAAAA |
| Movie | The Green Mile | &stick=H4sIAAAAAAAAAONgVuLUz9U3MC62zC4AAGg8mEkNAAAA |
| Movie | Dr. Strangelove | &stick=H4sIAAAAAAAAAONgVuLQz9U3MEopzwIAfsFCUQwAAAA |
| Animal | Toucan | &stick=H4sIAAAAAAAAAONgVuLQz9U3MM8wNAcA4g1_3AwAAAA |
| Animal | Wolverine | &stick=H4sIAAAAAAAAAONgUeLQz9U3sDAryQIAhJ3RUwwAAAA |
| Musicians/Groups | The Beatles | &stick=H4sIAAAAAAAAAONgVuLQz9U3ME82yAIAC_7r3AwAAAA |
| Albums | The White Album | &stick=H4sIAAAAAAAAAONgVuLSz9U3MMxIN0nKAADnd5clDgAAAA |
You can verify that this ‘&stick‘ reference is what is pulling in the infobox by looking at this modified query that has substituted Marilyn Monroe’s &stick in the Stephen King URL string: 
Note the standard search results in the lefthand results panel are the same as for Stephen King, but we now have fooled the Google engine to display Marilyn Monroe’s infobox.
I’m sure over time that others will deconstruct this framework to a very precise degree. What would really be great, of course, as noted on many recent mailing lists, is for Google to expose all of this via an API. The Google listing could become the de facto source for Webby entity identifiers.
Sort of like when schema.org was first announced, there have been complaints from some in the semantic Web community that Google released this stuff without once saying the word “semantic”, that much had been ripped off from the original researchers and community without attribution, that a gorilla commercial entity like Google could only be expected to milk this stuff for profit, etc., etc.
That all sounds like sour grapes to me.
What we have here is what we are seeing across the board: the inexorable integration of semantic technology approaches into many existing products. Siri did it with voice commands; Bing, and now Google, are doing it too with search.
We should welcome these continued adoptions. The fact is, semWeb community, that what we are seeing in all of these endeavors is the right and proper role for these technologies: in the background, enhancing our search and information experience, and not something front and center or rammed down our throats. These are the natural roles of semantic technologies, and they are happening at a breakneck pace.
Welcome to the semantic technology space, Google! I look forward to learning much from you.
