Posted:May 21, 2012

UMBEL Big GraphModularization Also Leads to Big Graph Visualization

We 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.

Modularization of the UMBEL Vocabulary

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. UMBEL Vocabulary and Reference Concept Ontology

Two sets of reference concepts were identified as useful for splitting out from the ‘core’ UMBEL in a geo-locational aspect:

  1. Geopolitical places and places of human activities and facilities
  2. Natural geographical places and features.

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.

Mapping to GeoNames

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:

  1. With UMBEL ‘core’ alone, recommended when your concept space is not concerned with geographical information
  2. UMBEL ‘core’ plus the UMBEL ‘geo’ module — equivalent to prior versions of UMBEL, or
  3. UMBEL ‘core’ plus GeoNames, recommended where geographical information is important to your concept space.

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.

New Big Graph Visualization

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 !

Note: If you are viewing this in a feed reader, click here to see the interactive graph.

This UMBEL graph displays:

  • All 26,000 concepts (“nodes”) with labels, and with connections shown (though you must must zoom to see)
  • The color-coded relation of these nodes to the 33 or so major SuperTypes in UMBEL, as well as the relative position of these clusters with respect to one another, and
  • When zooming (use scroll wheel or + icon) or panning (via mouse down moves), wait a couple of seconds to get the clearest image refresh:

You may also want to inspect a static version of this big graph by downloading a PDF.

Key Files and Links

Lastly, we fully updated the UMBEL Web site and re-released the UMBEL wiki.


[1] For more information on GeoNames, see http://www.geonames.org/. The complete mapping to GeoNames is based on its 671 feature codes, which describe natural, geopolitical, and human activity geo-locational information; see further http://www.geonames.org/statistics/total.html

[2] Approximate relationships are discussed in M.K. Bergman, 2010. “The Nature of Connectedness on the Web,” AI3:::Adaptive Information blog, November 22, 2010; see http://www.mkbergman.com/935/the-nature-of-connectedness-on-the-web/. One option, for example, is the 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:

The solution we are currently exploring is to define a new property to assert that two RDF instances are co-referential when they are believed to describe the same object in the world. The two RDF descriptions might be incompatible because they are true at different times, or the sources disagree about some of the facts, or any number of reasons, so merging them with 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.

[3] SWRL (Semantic Web Rule Language) combines sublanguages of the OWL Web Ontology Language (OWL DL and Lite) with those of the Rule Markup Language (Unary/Binary Datalog). SWRL has the full power of OWL DL, but at the price of decidability and practical implementations. See further http://www.w3.org/Submission/SWRL/.
[4] The Rule Interchange Format (RIF) is a W3C Recommendation. RIF is based on the observation that there are many “rules languages” in existence, and what is needed is to exchange rules between them. RIF includes three dialects, a Core dialect which is extended into a Basic Logic Dialect (BLD) and Production Rule Dialect (PRD). See further http://www.w3.org/2005/rules/wiki/RIF_FAQ.
[5] See further, M.K. Bergman, 2011. “A New Best Friend: Gephi for Large-scale Networks,” AI3:::Adaptive Information blog, August 8, 2011.
[6] schema.org lists its various contributing schema and also provides an OWL ontology of the system.
[7] See further, M.K. Bergman, 2011. “Structured Web Gets Massive Boost,” AI3:::Adaptive Information blog, June 2, 2011.

Posted by AI3's author, Mike Bergman Posted on May 21, 2012 at 12:26 am in Ontologies, Structured Web, UMBEL | Comments (4)
The URI link reference to this post is: http://www.mkbergman.com/999/new-umbel-release-gains-schema-org-geonames-capabilities/
The URI to trackback this post is: http://www.mkbergman.com/999/new-umbel-release-gains-schema-org-geonames-capabilities/trackback/
Posted:May 18, 2012

Google logoSome Quick Investigations Point to Promise, Disappointments

It 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.

Apparent Coverage

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: Google 'Stephen King' structured result

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:

  • People
  • Chemical Compounds
  • Directors
  • Some Companies
  • Some National Parks
  • Places
  • Musicians/Musical Groups
  • Actors
  • Some Government Agencies
  • Many Local Businesses
  • Animals
  • Movies
  • Albums
  • Notable Landmarks
  • Who Knows What Else

 

Entity types that I expected to see, but did not find include:

  • Products
  • Most Companies
  • Who Knows What Else
  • Songs
  • Most Government Agencies
  • Concepts
  • Non-government Organizations

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.

Deconstructing the URL

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:

https://www.google.com/search?num=100&hl=en&safe=off&q=stephen+king&stick=H4sIAAAAAAAAAONgVuLQz9U3MCvIKwQAcT1q1AwAAAA&sa=X&ei=Q-S2T6e1HYOw2wXR2OXOCQ&ved=0CPMGEJsTKAA&biw=1280&bih=827

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):

TypeEntityInfobox Identifier
PlaceLos Angeles &stick=H4sIAAAAAAAAAONgVuLSz9U3MDYoTDIuAQD9ON7KDgAAAA
PlaceBrentwook &stick=H4sIAAAAAAAAAONgVuLUz9U3MKwsNEkDAN6nm-sNAAAA
PersonArthur Miller &stick=H4sIAAAAAAAAAONgVmLXz9U3qEwrAADRsxaaCwAAAA
PersonJoe DiMaggio &stick=H4sIAAAAAAAAAONgVuLQz9U3SDFMqwAAAy8zdQwAAAA
PersonMarilyn Monroe &stick=H4sIAAAAAAAAAONgVuLQz9U3MCkvLAIAW7x_LwwAAAA
MovieShawshank Redemption &stick=H4sIAAAAAAAAAONgVuLQz9U3MM_KKwEAPYgNDQwAAAA
MovieThe Green Mile &stick=H4sIAAAAAAAAAONgVuLUz9U3MC62zC4AAGg8mEkNAAAA
MovieDr. Strangelove &stick=H4sIAAAAAAAAAONgVuLQz9U3MEopzwIAfsFCUQwAAAA
AnimalToucan &stick=H4sIAAAAAAAAAONgVuLQz9U3MM8wNAcA4g1_3AwAAAA
AnimalWolverine &stick=H4sIAAAAAAAAAONgUeLQz9U3sDAryQIAhJ3RUwwAAAA
Musicians/GroupsThe Beatles &stick=H4sIAAAAAAAAAONgVuLQz9U3ME82yAIAC_7r3AwAAAA
AlbumsThe 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: Google 'Stephen King' + 'Marilyn Monroe' structured results 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.

Some Concluding Thoughts

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.

Posted by AI3's author, Mike Bergman Posted on May 18, 2012 at 7:43 pm in Adaptive Information, Structured Web | Comments (3)
The URI link reference to this post is: http://www.mkbergman.com/1009/deconstructing-the-google-knowledge-graph/
The URI to trackback this post is: http://www.mkbergman.com/1009/deconstructing-the-google-knowledge-graph/trackback/
Posted:April 4, 2012

Tractricious Sculpture at Fermilab; picture by Mike KappelAdaptive Information is a Hammer, but Genes are Not a Nail

Since Richard Dawkins first put forward the idea of the “meme” in his book The Selfish Gene some 35 years ago [1], the premise has struck in my craw. I, like Dawkins, was trained as an evolutionary biologist. I understand the idea of the gene and its essential role as a vehicle for organic evolution. And, all of us clearly understand that “ideas” themselves have a certain competitive and adaptive nature. Some go viral; some run like wildfire and take prominence; and some go nowhere or fall on deaf ears. Culture and human communications and ideas play complementary — perhaps even dominant — roles in comparison to the biological information contained within DNA (genes).

I think there are two bases for why the “meme” idea sticks in my craw. The first harkens back to Dawkins. In formulating the concept of the “meme”, Dawkins falls into the trap of many professionals, what the French call déformation professionnelle. This is the idea of professionals framing problems from the confines of their own points of view. This is also known as the Law of the Instrument, or (Abraham) Maslow‘s hammer, or what all of us know colloquially as “if all you have is a hammer, everything looks like a nail [2]. Human or cultural information is not genetics.

The second — and more fundamental — basis for why this idea sticks in my craw is its mis-characterization of what is adaptive information, the title and theme of this blog. Sure, adaptive information can be found in the types of information structures at the basis of organic life and organic evolution. But, adaptive information is much, much more. Adaptive information is any structure that provides arrangements of energy and matter that maximizes entropy production. In inanimate terms, such structures include chemical chirality and proteins. It includes the bases for organic life, inheritance and organic evolution. For some life forms, it might include communications such as pheromones or bird or whale songs or the primitive use of tools or communicated behaviors such as nest building. For humans with their unique abilities to manipulate and communicate symbols, adaptive information embraces such structures as languages, books and technology artifacts. These structures don’t look or act like genes and are not replicators in any fashion of the term. To hammer them as “memes” significantly distorts their fundamental nature as information structures and glosses over what factors might — or might not — make them adaptive.

I have been thinking of these concepts much over the past few decades. Recently, though, there has been a spate of the “meme” term, particularly on the semantic Web mailing lists to which I subscribe. This spewing has caused me to outline some basic ideas about what I find so problematic in the use of the “meme” concept.

A Brief Disquisition on Memes

As defined by Dawkins and expanded upon by others, a “meme” is an idea, behavior or style that spreads from person to person within a culture. It is proposed as being able to be transmitted through writing, speech, gestures or rituals. Dawkins specifically called melodies, catch-phrases, fashion and the technology of building arches as examples of memes. A meme is postulated as a cultural analogue to genes in that they are assumed to be able to self-replicate, mutate or respond to selective pressures. Thus, as proposed, memes may evolve by natural selection in a manner analogous to that of biological evolution.

However, unlike a gene, a structure corresponding to a “meme” has never been discovered or observed. There is no evidence for it as a unit of replication, or indeed as any kind of coherent unit at all. In its sloppy use, it is hard to see how “meme” differs in its scope from concepts, ideas or any form of cultural information or transmission, yet it is imbued with properties analogous to animate evolution for which there is not a shred of empirical evidence.

One might say, so what, the idea of a “meme” is merely a metaphor, what is the harm? Well, the harm comes about when it is taken seriously as a means of explaining human behavior and cultural changes, a field of study called memetics. It becomes a pseudo-scientific term that sets a boundary condition for understanding the nature of information and what makes it adaptive or not [3]. Mechanisms and structures appropriate to animate life are not universal information structures, they are simply the structures that have evolved in the organic realm. In the human realm of signs and symbols and digital information and media, information is the universal, not the genetic structure of organic evolution.

The noted evolutionary geneticist, R.C. Lewontin, one of my key influences as a student, has also been harshly critical of the idea of memetics [4]:

 ”The selectionist paradigm requires the reduction of society and culture to inheritance systems that consist of randomly varying, individual units, some of which are selected, and some not; and with society and culture thus reduced to inheritance systems, history can be reduced to ‘evolution.’ . . . we conclude that while historical phenomena can always be modeled selectionistically, selectionist explanations do not work, nor do they contribute anything new except a misleading vocabulary that anesthetizes history.”

Consistent with my recent writings about Charles S. Peirce [5], many logicians and semiotic theorists are also critical of the idea of “memes”, but on different grounds. The criticism here is that “memes” distort Peirce’s ideas about signs and the reification of signs and symbols via a triadic nature. Notable in this camp is Terrence Deacon [6].

Information is a First Principle

It is not surprising that the concept of “memes” arose in the first place. It is understandable to seek universal principles consistent with natural laws and observations. The mechanism of natural evolution works on the information embodied in DNA, so why not look to genes as some form of universal model?

The problem here, I think, was to confuse mechanisms with first principles. Genes are a mechanism — a “structure” if you will — that along with other forms of natural selection such as the entire organism and even kin selection [7], have evolved as means of adaptation in the animate world. But the fundamental thing to be looked for here is the idea of information, not the mechanism of genes and how they replicate. The idea of information holds the key for drilling down to universal principles that may find commonality between information for humans in a cultural sense and information conveyed through natural evolution for life forms. It is the search for this commonality that has driven my professional interests for decades, spanning from population genetics and evolution to computers, information theory and semantics [8].

But before we can tackle these connections head on, it is important to address a couple of important misconceptions (as I see them).

Seque #1: Information is (Not!) Entropy

In looking to information as a first principle, Claude Shannon‘s seminal work in 1948 on information theory must be taken as the essential point of departure [9]. The motivation 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. (As a result, the Shannon paper is also intimately related to data patterns and data compression, not further discussed here.)

In a strict sense, Shannon’s paper was really talking about the amount of information that could be theoretically and predictably communicated between a sender and a receiver. No context or semantics were implied in this communication, only the amount of information (for which Shannon introduced the term “bits” [10]) 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 form that the uncertainty (unpredictability) calculation that Shannon derived:

 \displaystyle H(X) = - \sum_{i=1}^np(x_i)\log_b p(x_i)

very much resembled the mathematical form for Boltzmann‘s original definition of entropy (as elaborated upon by Gibbs, denoted as S, for Gibb’s entropy):

S = - k_B \sum p_i \ln p_i \,

and thus Shannon also labelled his measure of unpredictability, H, as entropy [10].

After Shannon, and nearly a century after Boltzmann, work by individuals such as Jaynes in the field of statistical mechanics came to show that thermodynamic entropy can indeed be seen as an application of Shannon’s information theory, so there are close parallels [11]. This parallel of mathematical form and terminology has led many to assert that information is entropy.

I believe this assertion is a misconception on two grounds.

First, as noted, what is actually being measured here is data (or bits), not information embodying any semantic meaning or context. Thus, the formula and terminology is not accurate for discussing “information” in a conventional sense.

Second, the Shannon methods are based on the communication (transmittal) between a sender and a receiver. Thus the Shannon entropy measure is actually a measure of the uncertainty for either one of these states. The actual information that gets transmitted and predictably received was formulated by Shannon as R (which he called rate), and he expressed basically as:

R = Hbefore – Hafter

R, then, becomes a proxy for the amount of information accurately communicated. R can never be zero (because all communication systems have losses). Hbefore and Hafter are both state functions for the message, so this also makes R a function of state. So while there is Shannon entropy (unpredictability) for any given sending or receiving state, the actual amount of information (that is, data) that is transmitted is a change in state as measured by a change in uncertainty between sender (Hbefore) and receiver (Hafter). In the words of Thomas Schneider, who provides a very clear discussion of this distinction [12]:

Information is always a measure of the decrease of uncertainty at a receiver.

These points do not directly bear on the basis of information as discussed below, but help remove misunderstandings that might undercut those points. Further, these clarifications make consistent theoretical foundations of information (data) with natural evolution while being logically consistent with the 2nd law of thermodynamics (see next).

Seque #2: Entropy is (Not!) Disorder

The 2nd law of thermodynamics expresses the tendency that, over time, differences in temperature, pressure, or chemical potential equilibrate in an isolated physical system. Entropy is a measure of this equilibration: for a given physical system, the highest entropy state is one at equilibrium. Fluxes or gradients arise when there are differences in state potentials in these systems. (In physical systems, these are known as sources and sinks; in information theory, they are sender and receiver.) Fluxes go from low to high entropy, and are non-reversible — the “arrow of time” — without the addition of external energy. Heat, for example, is a by product of fluxes in thermal energy. Because these fluxes are directional in isolation, a perpetual motion machine is shown as impossible.

In a closed system (namely, the entire cosmos), one can see this gradient as spanning from order to disorder, with the equilibrium state being the random distribution of all things. This perspective, and much schooling regarding these concepts, tends to present the idea of entropy as a “disordered” state. Life is seen as the “ordered” state in this mindset. Hewing to this perspective, some prominent philosophers, scientists and others have sometimes tried to present the “force” representing life and “order” as an opposite one to entropy. One common term for this opposite “force” is “negentropy[13].

But, in the real conditions common to our lives, our environment is distinctly open, not closed. We experience massive influxes of energy via sunlight, and have learned as well how to harness stored energy from eons past in further sources of fossil and nuclear energy. Our open world is indeed a high energy one, and one that increases that high-energy state as our knowledge leads us to exploit still further resources of higher and higher quality. As Buckminster Fuller once famously noted, electricity consumption (one of the highest quality energy resources found to date) has become a telling metric about the well-being and wealth of human societies [14].

The high-energy environments fostering life on earth and more recently human evolution establish a local (in a cosmic sense) gradient that promotes fluxes to more ordered states, not lesser unordered ones. These fluxes remain faithful to basic physical laws and are non-deterministic [15]. Indeed, such local gradients can themselves be seen as consistent with the conditions initially leading to life, favoring the random event in the early primordial soup that led to chemical structures such as chirality, auto-catalytic reactions, enzymes, and then proteins, which became the eventual building blocks for animate life [16].

These events did not have preordained outcomes (that is, they were non-deterministic), but were the result of time and variation in the face of external energy inputs to favor the marginal combinatorial improvement. The favoring of the new marginal improvement also arises consistent with entropy principles, by giving a competitive edge to those structures that produce faster movements across the existing energy gradient. According to Annila and Annila [16]:

“According to the thermodynamics of open systems, every entity, simple or sophisticated, is considered as a catalyst to increase entropy, i.e., to diminish free energy. Catalysis calls for structures. Therefore, the spontaneous rise of structural diversity is inevitably biased toward functional complexity to attain and maintain high-entropy states.”

Via this analysis we see that life is not at odds with entropy, but is consistent with it. Further, we see that incremental improvements in structure that are consistent with the maximum entropy production principle will be favored [17]. Of course, absent the external inputs of energy, these gradients would reverse. Under those conditions, the 2nd law would promote a breakdown to a less ordered system, what most of us have been taught in schools.

With these understandings we can now see the dichotomy as life representing order with entropy disorder as being false. Further, we can see a guiding set of principles that is consistent across the broad span of evolution from primordial chemicals and enzymes to basic life and on to human knowledge and artifacts. This insight provides the fundamental “unit” we need to be looking toward, and not the gene nor the “meme”.

Information is Structure

Of course, the fundamental “unit” we are talking about here is information, and not limited as is Shannon’s concept to data. The quality that changes data to information is structure, and structure of a particular sort. Like all structure, there is order or patterns, often of a hierarchical or fractal or graph nature. But the real aspect of the structure that is important is the marginal ability of that structure to lead to improvements in entropy production. That is, processes are most adaptive (and therefore selected) that maximize entropy production. Any structure that emerges that is able to reduce the energy gradient faster will be favored.

However, remember, these are probabilistic, statistical processes. Uncertainties in state may favor one structure at one time versus another at a different time. The types of chemical compounds favored in the primordial soup were likely greatly influenced by thermal and light cycles and drying and wet conditions. In biological ecosystems, there are huge differences in seed or offspring production or in overall species diversity and ecological complexity based on the stability (say, tropics) or instability (say, disturbance) of local environments. As noted, these processes are inherently non-deterministic.

As we climb up the chain from the primordial ooze to life and then to humans and our many information mechanisms and technology artifacts (which are themselves embodiments of information), we see increasing complexity and structure. But we do not see uniformity of mechanisms or vehicles.

The general mechanisms of information transfer in living organisms occur (generally) via DNA in genes, mediated by sex in higher organisms, subject to random mutations, and then kept or lost entirely as their host organisms survive to procreate or not. Those are harsh conditions: the information survives or not (on a population basis) with high concentrations of information in DNA and with a priority placed on remixing for new combinations via sex. Information exchange (generally) only occurs at each generational event.

Human cultural information, however, is of an entirely different nature. Information can be made persistent, can be recorded and shared across individuals or generations, extended with new innovations like written language or digital computers, or combined in ways that defy the limits of sex. Occasionally, of course, loss of living languages due to certain cultures or populations dying out or horrendous catastrophes like the Spanish burning (nearly all of) the Mayan’s existing books can also occur [18]. The environment will also be uncertain.

So, while we can define DNA in genes or the ideas of a “meme” all as information, in fact we now see how very unlike the dynamics and structures of these two forms really are. We can be awestruck with the elegance and sublimity of organic evolution. We can also be inspired by song or poem or moved to action through ideals such as truth and justice. But organic evolution does not transpire like reading a book or hearing a sermon, just like human ideas and innovations don’t act like genes. The “meme” is a totally false analogy. The only constant is information.

Some Tentative Implications

The closer we come to finding true universals, the better we will be able to create maximum entropy producing structures. This, in turn, has some pretty profound implications. The insight that keys these implications begins with an understanding of the fundamental nature — and importance — of information. According to Karnani et al [19]:

“. . . the common contemporary consent, the second law of thermodynamics, is perceived to drive disorder. Therefore, it may appear, at first sight, inconceivable that this universal law could possibly account for the existence and orderly characteristics of information, as well as for its meaningful content. However, the second law, or equivalently the principle of increasing entropy, merely states that difference among energy densities tends to vanish. When the surrounding energy density is high, the system will evolve toward a stationary state by increasing its energy content, e.g, by devising orderly machinery for energy transduction to acquire energy. . . . Syntax of information, when described by thermodynamics, is associated with the entropy of the physical representation, and significance of information is associated with the entropy increase in the receiver system when it executes the encoded information.”

All would agree that the evolution of life over the past few billion years is truly wondrous. But, what is equally wondrous is that the human species has come to learn and master symbols. That mastery, in turn, has broken the bounds of organic evolution and has put into our hands the very means and structure of information itself. Via this entirely new — and incredibly accelerated — path to information structures, we are only now beginning to see some of its implications:

  • Unlike all other organisms, we dominate our environment and have experienced increasing wealth and freedom. Wealth increases and their universal applicability continue to increase at an exponential rate [20]
  • We no longer depend on the random variant to maximize our entropy producing structures. We can now do so purposefully and with symbologies and bit streams of our own devising
  • Potentially all information variants can be recorded and shared across all human individuals and generations, a complete decoupling from organic boundaries
  • Key ideas and abstractions, such as truth, justice and equality, can operate on a species-wide basis and become adopted without massive die-offs of individuals
  • We are actively moving ourselves into higher-level energy states, further increasing the potential for wealth and new structures
  • We are actively impacting our local environment, potentially creating the conditions for our species’ demise
  • We are increasingly engaging all individuals of the human species in these endeavors through literacy, education and access to global information sources. This provides a still further multiplier effect on humanity’s ability to devise and manipulate information structures into more adaptive and highly-ordered states.

The idea of a “meme” actually cheapens our understanding of these potentials.

Ideas matter and terminology matters. These are the symbols by which we define and communicate potentials. If we choose the wrong analogies or symbols — as “meme” is in this case — we are picking the option with the lower entropy potential. Whether I assert it to be so or not, the “meme” concept is an information structure doomed for extinction.


[1] Richard Dawkins, 1976. The Selfish Gene, Oxford University Press, New York City, ISBN 0-19-286092-5.
[2] This phrase was perhaps first made famous by Mark Twain or Bernard Baruch, but in any case is clearly understood now by all.
[3] According to Wikipedia, Benitez-Bribiesca calls memetics “a dangerous idea that poses a threat to the serious study of consciousness and cultural evolution”. He points to the lack of a coding structure analogous to the DNA of genes, and to instability of any mutation mechanisms for “memes” sufficient for standard evolution processes. See Luis Benitez Bribiesca, 2001. “Memetics: A Dangerous Idea”, Interciencia: Revista de Ciencia y Technologia de América (Venezuela: Asociación Interciencia) 26 (1): 29–31, January 2001. See http://redalyc.uaemex.mx/redalyc/pdf/339/33905206.pdf.
[4] Joseph Fracchia and R.C. Lewontin, 2005. “The Price of Metaphor”, History and Theory (Wesleyan University) 44 (44): 14–29, February 2005.
[5] See further M. K. Bergman, 2012. “Give Me a Sign: What Do Things Mean on the Semantic Web?,” posting on AI3:::Adaptive Information blog, January 24, 2012. See http://www.mkbergman.com/994/give-me-a-sign-what-do-things-mean-on-the-semantic-web/.
[6] Terrence Deacon, 1999. “The Trouble with Memes (and what to do about it)”. The Semiotic Review of Books 10(3). See http://projects.chass.utoronto.ca/semiotics/srb/10-3edit.html.
[7] Kin selection refers to changes in gene frequency across generations that are driven at least in part by interactions between related individuals. Some mathematical models show how evolution may favor the reproductive success of an organism’s relatives, even at a cost to an individual organism. Under this mode, selection can occur at the level of populations and not the individual or the gene. Kin selection is often posed as the mechanism for the evolution of altruism or social insects. Among others, kin selection and inclusive fitness was popularized by W. D. Hamilton and Robert Trivers.
[8] You may want to see my statement of purpose under the Blogasbörd topic, first written seven years ago when I started this blog.
[9] Claude E. Shannon, 1948. “A Mathematical Theory of Communication”, Bell System Technical Journal, 27: 379–423, 623-656, July, October, 1948. See http://cm.bell-labs.com/cm/ms/what/shannonday/shannon1948.pdf.
[10] As Shannon acknowledges in his paper, the “bit” term was actually suggested by J. W. Tukey. Shannon can be more accurately said to have popularized the term via his paper.
[12] See Thomas D. Schneider, 2012. “Information Is Not Entropy, Information Is Not Uncertainty!,” Web page retrieved April 4, 2012; see http://www.lecb.ncifcrf.gov/~toms/information.is.not.uncertainty.html.
[13] The “negative entropy” (also called negentropy or syntropy) of a living system is the entropy that it exports to keep its own entropy low, and according to proponents lies at the intersection of entropy and life. The concept and phrase “negative entropy” were introduced by Erwin Schrödinger in his 1944 popular-science book What is Life?. See Erwin Schrödinger, 1944. What is Life – the Physical Aspect of the Living Cell, Cambridge University Press, 1944. A copy may be downloaded at http://old.biovip.com/UpLoadFiles/Aaron/Files/2005051204.pdf.
[14] R. Buckminster Fuller, 1981. Critical Path, St. Martin’s Press, New York City, 471 pp. See especially p. 103 ff.
[15] The seminal paper first presenting this argument is Vivek Sharma and Arto Annila, 2007. “Natural Process – Natural Selection”, Biophysical Chemistry 127: 123-128. See http://www.helsinki.fi/~aannila/arto/natprocess.pdf. This basic theme has been much expanded upon by Annila and his various co-authors. See, for example, [16] and [19], among many others.
[16] Arto Annila and Erkki Annila, 2008. “Why Did Life Emerge?,” International Journal of Astrobiology 7(3 and 4): 293-300. See http://www.helsinki.fi/~aannila/arto/whylife.pdf.
[17] According to Wikipedia, the principle (or “law”) of maximum entropy production is an aspect of non-equilibrium thermodynamics, a branch of thermodynamics that deals with systems that are not in thermodynamic equilibrium. Most systems found in nature are not in thermodynamic equilibrium and are subject to fluxes of matter and energy to and from other systems and to chemical reactions. One fundamental difference between equilibrium thermodynamics and non-equilibrium thermodynamics lies in the behavior of inhomogeneous systems, which require for their study knowledge of rates of reaction which are not considered in equilibrium thermodynamics of homogeneous systems. Another fundamental difference is the difficulty in defining entropy in macroscopic terms for systems not in thermodynamic equilibrium.
The principle of maximum entropy production states that the in comparing two or more alternate paths for crossing an energy gradient that the one that creates the maximum entropy change will be favored. The maximum entropy (sometimes abbreviated MaxEnt or MaxEp) concept is related to this notion. It is also known as the maximum entropy production principle, or MEPP.
[18] The actual number of Mayan books burned by the Spanish conquistadors is unknown, but is somewhere between tens and thousands; see here. Only three or four codexes are known to survive today. Also, Wikipedia contains a listing of notable book burnings throughout history.
[19] Mahesh Karnani, Kimmo Pääkkönen and Arto Annila, 2009. “The Physical Character of Information,” Proceedings of the Royal Society A, April 27, 2009. See http://www.helsinki.fi/~aannila/arto/natinfo.pdf.
[20] I discuss and chart the exponential growth of human wealth based on Angus Maddison data in M. K. Bergman, 2006. “The Biggest Disruption in History: Massively Accelerated Growth Since the Industrial Revolution,” post in AI3:::Adaptive Information blog, July 27, 2006. See http://www.mkbergman.com/250/the-biggest-disruption-in-history-massively-accelerated-growth-since-the-industrial-revolution/.
Posted:December 12, 2011

State of SemWeb Tools - 2011Number of Semantic Web Tools Passes 1000 for First Time; Many Other Changes

We have been maintaining Sweet Tools, AI3‘s listing of semantic Web and -related tools, for a bit over five years now. Though we had switched to a structWSF-based framework that allows us to update it on a more regular, incremental schedule [1], like all databases, the listing needs to be reviewed and cleaned up on a periodic basis. We have just completed the most recent cleaning and update. We are also now committing to do so on an annual basis.

Thus, this is the inaugural ‘State of Tooling for Semantic Technologies‘ report, and, boy, is it a humdinger. There have been more changes — and more important changes — in this past year than in all four previous years combined. I think it fair to say that semantic technology tooling is now reaching a mature state, the trends of which likely point to future changes as well.

In this past year more tools have been added, more tools have been dropped (or abandoned), and more tools have taken on a professional, sophisticated nature. Further, for the first time, the number of semantic technology and -related tools has passed 1000. This is remarkable, given that more tools have been abandoned or retired than ever before.

Click here to browse the Sweet Tools listing. There is also a simple listing of URL links and categories only.

We first present our key findings and then overall statistics. We conclude with a discussion of observed trends and implications for the near term.

Key Findings

Some of the key findings from the 2011 State of Tooling for Semantic Technologies are:

  • As of the date of this article, there are 1010 tools in the Sweet Tools listing, the first it has passed 1000 total tools
  • A total of 158 new tools have been added to the listing in the last six months, an increase of 17%
  • 75 tools have been abandoned or retired, the most removed at any period over the past five years
  • A further 6%, or 55 tools, have been updated since the last listing
  • Though open source has always been an important component of the listing, it now constitutes more than 80% of all listings; with dual licenses, open source availability is about 83%. Online systems contribute another 9%
  • Key application areas for growth have been in SPARQL, ontology-related areas and linked data
  • Java continues to dominate as the most important language.

Many of these points are elaborated below.

The Statistical Picture

The updated Sweet Tools listing now includes nearly 50 different tools categories. The most prevalent categories, each with over 6% of the total, are information extraction, general RDF tools, ontology tools, browser tools (RDF, OWL), and parsers or converters. The relative share by category is shown in this diagram (click to expand):

Since the last listing, the fastest growing categories have been SPARQL, linked data, knowledge bases and all things related to ontologies. The relative changes by tools category are shown in this figure:

Though it is true that some of this growth is the result of discovery, based on our own tool needs and investigations, we have also been monitoring this space for some time and serendipity is not a compelling explanation alone. Rather, I think that we are seeing both an increase in practical tools (such as for querying), plus the trends of linked data growth matched with greater sophistication in areas such as ontologies and the OWL language.

The languages these tools are written in have also been pretty constant over the past couple of years, with Java remaining dominant. Java has represented half of all tools in this space, which continues with the most recent tools as well (see below). More than a dozen programming or scripting languages have at least some share of the semantic tooling space (click to expand):

Sweet Tools Languages

With only 160 new tools it is hard to draw firm trends, but it does appear that some languages (Haskell, XSLT) have fallen out of favor, while popularity has grown for Flash/Flex (from a small base), Python and Prolog (with the growth of logic tools):

PHP will likely continue to see some emphasis because of relations to many content management systems (WordPress, Drupal, etc.), though both Python and Ruby seem to be taking some market share in that area.

New Tools

The newest tools added to the listing show somewhat similar trends. Again, Java is the dominant language, but with much increased use of JavaScript and Python and Prolog:

Sweet Tools Languages

The higher incidence of Prolog is likely due to the parallel increase in reasoners and inference engines associated with ontology (OWL) tools.

The increase in comprehensive tool suites and use of Eclipse as a development environment would appear to secure Java’s dominance for some time to come.

Trends and Observations

These dry statistics tend to mask the feel one gets when looking at most of the individual tools across the board. Older academic and government-funded project tools are finally getting cleaned out and abandoned. Those tools that remain have tended to get some version upgrades and improved Web sites to accompany them.

The general feel one gets with regard to semantic technology tooling at the close of 2011 has these noticeable trends:

  • A three-tiered environment – the tools seem to segregate into: 1) a bottom tier of tools (largely) developed by individuals or small groups, now most often found on Google Code or Github; 2) a middle-tier of (largely) government-funded projects, sometimes with multiple developers, often older, but with no apparent driving force for ongoing improvements or commercialization; and 3) a top-tier of more professional and (often) commercially-oriented tools. The latter category is the most noticeable with respect to growth and impact
  • Professionalism – the tools in the apparent top tiers feel to have more professionalism and better (and more attractive) packaging. This professionalism is especially true for the frameworks and composite applications. But, it also applies to many of the EU-funded projects from Europe, which has always been a huge source of new tool developments
  • More complete toolsets – similarly, the upper levels of tools are oriented to pragmatic problems and problem-solving, which often means they embody multiple functions and more complete tooling environments. This category actually appears to be the most visible one exhibiting growth
  • Changing nature of academic releases – yet, even the academic releases seem to be increasing in professionalism and completeness. Though in the lowest tier it is still possible to see cursory or experimental tool releases, newer academic releases (often) seem to be more strategically oriented and parts of broader programmatic emphases. Programs like AKSW from the University of Leipzig or the Freie Universität Berlin or Finland’s Semantic Computing Research Group (SeCo), among many others, tend to be exemplars of this trend
  • Rise of commercial interests and enterprise adoption – the growing maturity of semantic technologies is also drawing commercial interest, and the incubation of new start-ups by academic and research institutions acts to reinforce the above trends. Promising projects and tools are now much more likely to be spun off as potential ventures, with accompanying better packaging, documentation and business models
  • Multiple languages and applications – with this growing complexity and sophistication has also come more complicated apps, combining multiple languages and functions. In fact, for some time the Sweet Tools listing has been justifiably criticized by some as overly “simplifying” the space by classifying tools under (largely) single applications or single languages. By the 2012 survey, it will likely be necessary to better classify the tools using multiple assignments
  • Google code over SourceForge for open source (and an increase in Github, as well) – virtually all projects on SourceForge now feel abandoned or less active. The largest source of open source projects in the semantic technology space is now clearly Google Code. Though of a smaller footprint today, we are also seeing many of the newer open source projects also gravitate to Github. Open source hosting environments are clearly in flux.

I have said this before, and been wrong about it before, but it is hard to see the tooling growth curve continue at its current slope into the future. I think we will see many individual tools spring up on the open source hosting sites like Google and Github, perhaps at relatively the same steady release rate. But, old projects I think will increasingly be abandoned and older projects will not tend to remain available for as long a time. While a relatively few established open source standards, like Solr and Jena, will be the exception, I think we will see shorter shelf lives for most open source tools moving forward. This will lead to a younger tools base than was the case five or more years ago.

I also think we will continue to see the dominance of open source. Proprietary software has increasingly been challenged in the enterprise space. And, especially in semantic technologies, we tend to see many open source tools that are as capable as proprietary ones, and generally more dynamic as well. The emphasis on open data in this environment also tends to favor open source.

Yet, despite the professionalism, sophistication and complexity trends, I do not yet see massive consolidation in the semantic technology space. While we are seeing a rapid maturation of tooling, I don’t think we have yet seen a similar maturation in revenue and business models. While notable semantic technology start-ups like Powerset and Siri have been acquired and are clear successes, these wins still remain much in the minority.


[1] Please use the comments section of this post for suggesting new or overlooked tools. We will incrementally add them to the Sweet Tools listing. Also, please see the About tab of the Sweet Tools results listing for prior releases and statistics.

Posted by AI3's author, Mike Bergman Posted on December 12, 2011 at 8:29 am in Open Source, Semantic Web Tools, Structured Web | Comments (6)
The URI link reference to this post is: http://www.mkbergman.com/991/the-state-of-tooling-for-semantic-technologies/
The URI to trackback this post is: http://www.mkbergman.com/991/the-state-of-tooling-for-semantic-technologies/trackback/
Posted:July 18, 2011

Photo courtesy of levelofhealth.comA Decade of Remarkable Advances in Ten Grand IT Challenges

I’ve been in the information theory and technology game for quite some time, but believe nothing has matched the pace of advances of the past ten years. As one example, it was a mere eight years ago that I was sitting in a room with language translation vendors contemplating automated translation techniques for US intelligence agencies. The prospects finally looked doable, but the success of large-scale translation was not assured.

At about that same time, and the years until just recently, a whole slew of Grand Challenges [1] in computing hung out there: tantalizing yet not proven. These areas ranged from information extraction and natural language understanding to speech recognition and automated reasoning.

But things have been changing fast, and with a subtle steadiness that has caused it to go largely unremarked. Sure, all of us have been aware of the huge changes on the Web and search engine ubiquity and social networking. But some of the fundamentally hard problems in computing have also gone through some remarkable (but largely unremarked) advances.

We now have smart phones that speak instructions to us while we instruct them by voice in turn. Virtually all information conceivable is now indexed and made available through the Web; structure is now rapidly characterizing that information, making it even more useful to discover and organize. We can translate documents online with acceptable accuracy into more than 60 languages [2]. We can get directions to or see satellite views of virtually any place on earth. We have in fact become accustomed to new technology magic on a nearly daily basis, so much so that the pace of these advances seems to be a constant, blunting our perspective of just how rapid these advances have been progressing.

These advances are perhaps not the realization of artificial intelligence as articulated in the 1950s to 1980s, but are contributing to a machine-based ability to do tasks useful to humans heretofore impossible and at scales unimaginable. As Google and IBM’s Watson are showing, statistics (among other techniques) applied to massive knowledge bases or text corpora are breaking down all of the Grand Challenges of symbolic computing. The image that is emerging is less one of intelligent machines working autonomously than it is of computers working interactively or semi-automatically with humans to address previously unsolvable problems.

By using a perspective of the decade past, we also demark the seminal paper on the semantic Web by Berners-Lee, Hendler and Lassila from May 2001 [3]. Yet, while this semantic Web vision has been a contributor to the success of the Grand Challenge advances of the past ten years, I think we can also say that it has not been the key or even a primary driver. That day may still yet come. Rather, I think we have to look to natural language and statistics surrounding large-scale corpora as the more telling drivers.

Ten Grand Challenge Advances

Over the past ten years there have been significant advances on at least ten Grand Challenges in symbolic computation. As the concluding section notes, these advances can be traced in most part to broader advances in natural language processing, the logical and semiotic bases for interoperability, and standards (nominally in the semantic Web) for embracing them. Here are these ten areas of advance, all achieved over the past ten years:

#1 Information Extraction

Information extraction (IE) uses various forms of natural language processing (NLP) to identify structured information within unstructured or semi-structured documents. These documents are presented in machine-readable form (including straight text, various document formats or HTML) with the various types of information “tagged” or prompted for inclusion. Information types that can be extracted with one of the various techniques include entities, relations, topics, categories, and so forth. Once tagged or extracted, the information in the documents can now be included and linked to standard structured information (as might come from conventional databases) or to structure in other documents.

Most recently, a large number of online services and open source systems have also become available with strengths in one or more of these extraction types [4]. Some current examples include Yahoo! Term Extraction, OpenCalais, BeliefNetworks, OpenAmplify, Alchemy API, Evri, Extractiv, Illinois Tagger, and about 80 others [4].

#2 Machine Translation

Machine translation is the automatic translation of machine-readable text from one human language to another. Accurate and acceptable machine translation requires applying different types of knowledge including grammar, semantics, facts about the real world, etc. Various approaches have been developed and refined over time.

Especially helpful has been the availability of huge corpora in multiple languages to which large-scale statistical analysis may be applied (as is the case of Google’s machine translation) or human editing and refinement (as is the case with the more than 280 language versions of Wikipedia).

While it is true none of these systems have 100% accuracy (even human translators show much variation), the more advanced ones are truly impressive with remaining ambiguities flagged for resolution by semi-automatic means.

#3 Sentiment Analysis

Though sentiment analysis is strictly speaking a subset of information extraction, it has the more demanding and useful task of extracting subjective information, often across a group of documents or texts. Sentiment analysis can be applied to online reviews to determine the “polarity” about specific objects, and it is especially useful for identifying public opinion trends or evaluating social media for ranking, polling or marketing purposes.

Because of its greater difficulty and potential high value, many of the leading sentiment analysis capabilities remain proprietary. Some capable open source versions are available nonetheleless. There is also an interesting online application using Twitter feeds.

#4 Disambiguation

Many words have more than one meaning. Word sense disambiguation uses either machine learning, dictionaries (gazetteers) of known entities and concepts, ontologies or linguistic databases such as WordNet, or combinations thereof to evaluate ambiguous terms or phrases and resolve them based on context. Some systems need to be “trained” or some work automatically or others are based on evaulation and prompting (semi-automatic) to complete the disambiguation process.

State-of-the-art systems have greater than 90% precision [5]. Most of the leading open source NLP toolkits have quite capable disambiguation modules, and even better proprietary systems exist.

#5 Speech Synthesis and Recognition

Speech synthesis is the conversion of text to spoken speech and has been around for quite some time. Speech recognition is a far more difficult task in that a given sound clip or real-time spoken speech of a person must be converted to a textual representation, which itself can then be acted upon such as navigating or making selections. Speech recognition is made difficult because of individual voice differences, the variations of human languages and speech patterns, and the need to segment speech into a sequence of words. (In most spoken languages, the sounds representing successive letters blend into each other, so the conversion of the modulated wave form to discrete characters or tokens can be a very difficult process.)

Crude systems of a decade ago required much training with a specific speaker’s voice to show much effectiveness. Today, the range and ability to use these systems without training has markedly improved.

Until recently, improvements largely were driven by military and intelligence requirements. Today, however, with the ubiquity of smart phones and speech interfaces, the consumer market is greatly accelerating progress.

#6 Image Recognition

Image recognition is the ability to determine whether or not an electronic image contains some specific object, feature, or activity, and then to extract the image data associated with it. Today, under specific circumstances and for specific tasks, this can be done by computer. However, for the general case of arbitrary objects in arbitrary situations this challenge has not yet been fully met. The systems of today work best for simple geometric objects (e.g., polyhedra), human faces, printed or hand-written characters, or vehicles, and in specific situations, typically described in terms of well-defined illumination, background, and orientation of the object relative to the camera.

Auto license recognition at intersections, face recognition by security cameras, and greatly expanded and improved character recognition systems (machine vision) represent some of the current state-of-the-art. Again, smart phone apps are helping to drive advances.

#7 Interoperability Standards and Methods


Rapid Progress in Climbing the Data Federation Pyramid

Most of the previous advances are related to extracting structured information or mapping or deriving additional structured information. Once obtained, of course, the next challenge is in how to relate that information together; that is, how to make it interoperate.

We have been steadily climbing a data federation pyramid [6] — and at an impressively accelerating rate since the adoption of the Internet and Web. These network innovations gave us a common basis and protocols for connecting distributed devices. That, in turn, has freed us to concentrate on the standards for data representation and interoperability.

XML first provided a means for a common data serialization that encouraged various communities and industries to devise exchange vocabularies. RDF provided a means for a common data model, one that was both simple and extensible at the same time [7]. OWL built upon that basis to enable us to build common domain models (see next).

There are alternatives to the semantic Web standards of RDF and OWL such as common logic and there are many competing data exchange formats to XML. None of these standards is essential on its own and all have their communities and advocates. However, because they are standards and they share common network bases, it has also been relatively easy to convert amongst the various available protocols. We are nearly at a global level where everything is connected, machine-readable, and in structured form.

#7 Common Domain Models

Semantics in machine-readable form means that we can more confidently link and combine available information. We are seeing a veritable explosion of domain models to represent various domains and viewpoints in consensual, interoperable form. What this means is that we are now gaining the computing vocabularies and grammars — along with shared community models (world views) — to get this stuff to work together.

Five years ago we called this phenomena mashups, but no one uses that term any longer because these information brewpots are everywhere, including in our very hands when we interact with the apps on our smart phones. This glue of domain models is generally as invisible to us as is the glue in laminates or the resin in plastics. But they are the strength and foundations nonetheless that enable much of the computing magic unfolding around us.

#9 Virtual Apps (Cloud Computing)

Once the tyranny of physical separation was shattered between data and machine by the network, the rationale for keeping the data with the app or even the user with the app disappeared. Cloud computing may seem mysterious or sound to have some high-octave hum, but it really is nothing more than saying that the Web enables us to treat all of our computing resources as virtual. Data can be anywhere; machines and hard drives can be anywhere; and applications can be anywhere.

And, virtualness brings benefits in and of itself. Whole computing environments can be installed or removed nearly instantaneously. Peak computing demands can be met with virtual headrooms. Backup and rollover and redundancy practices and strategies can change. Web services mean tailored capabilities can be invoked from anywhere and integrated for local needs. Massive computing resources and server farms can be as accessible to the individual as they are to prior computing behemoths. Combined with continued advances in underlying computing hardware and chips, the computing power available to any user is rising exponentially. There is now even more power in the power curve.

#10 Big Data

One hears stories of Google or the National Security Agency having access and managing servers measured in the hundreds of thousands. Entirely new operating systems and computing environments — many with roots in open source — such as virtual operating systems and MapReduce approaches like Hadoop have been innovated to deal with the current era of “big data”.

MapReduce is a framework for processing huge datasets using a large number of servers. The “map” step partitions the problem into tractable sub-problems, organized in a tree structure. The “reduce” step then takes the answers to all the sub-problems and combines them to produce the final output.

Such techniques enable analysis of datasets of a size impossible before. This has enabled the development of statistics and analytical techniques that have been able to make correlations and find patterns for some of the Grand Challenge tasks noted before that simply could not be addressed within previous limits. The “big data” approach is providing a brute force alternative to previously intractable problems.

Why Such Progress?

Declining hardware costs and increasing performance (such as from Moore’s Law), combined with the adoption of the Internet + Web network, set the fertile conditions for these unprecedented advances in computing’s Grand Challenges. But the adaptive radiation in innovations now occurring has its own dynamics. In computing terms, we are seeing the equivalent of the Cambrian explosion in evolutionary history.

The dynamics driving this computing explosion are based largely, I believe, on the statistics of information retrieval and extraction needed to cope with the scale of documents on the Web. That, in turn, has impelled innovations in big data and distributed architectures and designs that have pried open previously closed computing lockboxes. As data from everywhere and from every provenance pours into the system, means for handling and interoperating with it have become imperatives. These forces, in turn, have been channeled and are being met through the open and standards-based approaches that helped lead to the development of the Internet and its infrastructure in the first place.

These powerful evolutionary forces in computing are clearly evident in the ten Grand Challenge advances above. But the challenges above are also silent on another factor, underpinning the interoperability initiatives, that is only now just becoming evident and exerting its own powerful force. That is the workable, intellectual foundations for interoperability itself.

Clearly, as the advances in the Grand Challenges show, we are seeing immense exposures of new structured information and impressive means for accessing and managing it on a global, distributed scale.  Yet all of this data and this structure begs the question of how to get the information to work together. Further, the sources and viewpoints and methods by which all of this data has been created also puts a huge premium on means to deal with the diversity. Though not evident, and perhaps not even known to many of the innovators and practitioners, there has been a growing intellectual force shaping our foundational views about the nature of things and their representations. This force has been, I believe, one of those root cause drivers helping to show the way to interoperability.

John Sowa, despite his unending criticism of the semantic Web in favor of common logic, has nonetheless been a very positive evangelist for the 19th century American logician and philosopher, Charles Sanders Peirce. Sowa points out that the entire 20th century largely neglected Peirce’s significant contributions in many areas and some philosophers appropriated Peircean insights without proper attribution [8]. Indeed, Peirce has only come to wider attention within the past decade or so. Much of his voluminous lifetime writings have still not yet been committed to publication.

Among many notable contributions, Peirce was passionate about signs and their triadic representations, in a field known as semiotics. The philosophical and logical basis of his triangle of signs deserves your attention, which can not be adequately treated here [9]. However, as summarized by Sowa [8], “A semiotic view of language and logic gets to the heart of the philosophical controversies and their practical implications for linguistics, artificial intelligence, and related subjects.”

In essence, Peirce’s triadic logic of semiotics helps clarify philosophical questions about things, how they are perceived and how they are named that has vexed philosophers at least since the time of Aristotle. What Peirce was able to put forward was 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.

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 [10]. As we plumb Peircean logics further, I believe we will continue to gain additional insights and methods for combining and relating information. The next phase of our advances on these Grand Challenges is likely to be fueled more by connections and interoperability than in basic extraction or representation.

The Widening Explosion

We are not seeing the vision of artificial intelligence unfold as posed three decades ago. Nor are we seeing the AI-complete type of problems being solved in their entirety [11]. Rather, we are seeing impressive but incomplete approaches. Full automation and autonomy are not yet at hand, and may be so far in the future as to never be. But we are nevertheless seeing advances across the board in all Grand Challenge areas.

What is emerging is a practical achievement of the Grand Challenges, the scale and scope of which is unprecedented in symbolic computing. As we see Peircean logic continue to take hold and interoperability grow in usefulness and stature, I think it fair to say we can look back in ten years to describe where we stand today as having been in the midst of an evolutionary explosion.


[1] Grand Challenges were United States policy objectives for high-performance computing and communications research set in the late 1980s. According to “A Research and Development Strategy for High Performance Computing”, Executive Office of the President, Office of Science and Technology Policy, 29 pp., November 20, 1987, “A grand challenge is a fundamental problem in science or engineering, with broad applications, whose solution would be enabled by the application of high performance computing resources that could become available in the near future.”
[2] For example, as of July 17, 2011, Google offered 63 different source or target languages for translation.
[3] Tim Berners-Lee, James Hendler and Ora Lassila, 2001. “The Semantic Web”. Scientific American Magazine; see http://www.scientificamerican.com/article.cfm?id=the-semantic-web.
[4] Go to Sweet Tools, and enter the search ‘information extraction’ to see a list of about 85 tools.
[5] See, for example, Roberto Navigli, 2009. “Word Sense Disambiguation: A Survey,” ACM Computing Surveys, 41(2), 2009, pp. 1–69. See http://www.dsi.uniroma1.it/~navigli/pubs/ACM_Survey_2009_Navigli.pdf.
[6] M.K. Bergman, 2006. “Climbing the Data Federation Pyramid,” AI3:::Adaptive Information blog, May 25, 2006; see http://www.mkbergman.com/229/climbing-the-data-federation-pyramid/.
[7] M. K. Bergman, 2009. “Advantages and Myths of RDF,” AI3:::Adaptive Information blog, April 8, 2009. See http://www.mkbergman.com/483/advantages-and-myths-of-rdf/
[8] John Sowa, 2006. “Peirce’s Contributions to the 21st Century”, in H. Schärfe, P. Hitzler, & P. Øhrstrøm, eds., Conceptual Structures: Inspiration and Application, LNAI 4068, Springer, Berlin, 2006, pp. 54-69. See http://www.jfsowa.com/pubs/csp21st.pdf.
[9] See, as a start, the Wikipedia article on Charles Sanders Peirce (pronounced “purse”), as well as the Arisbe collection of his assembled papers (to date). Also see John Sowa, 2010. “The Role of Logic and Ontology in Language and Reasoning,” from Chapter 11 of Theory and Applications of Ontology: Philosophical Perspectives, edited by R. Poli & J. Seibt, Berlin: Springer, 2010, pp. 231-263. See http://www.jfsowa.com/pubs/rolelog.pdf. Sowa also says, “Although formal logic can be studied independently of natural language semantics, no formal ontology that has any practical application can ever be developed and used without acknowledging its intimate connection with NL semantics.”
[10] While Peirce’s logic and clarity of conceptual relationships is compelling, I find reading his writings quite demanding.
[11] In the field of artificial intelligence, the most difficult problems are informally known as AI-complete or AI-hard, meaning that the difficulty of these computational problems is equivalent to solving the central artificial intelligence problem of making computers as intelligent as people. Computer vision, autonomous robots and understanding natural language are amongst challenges recognized by consensus as being AI-complete. However, practical advances on the Grand Challenges were never defined as needing to meet the AI-complete criterion. Indeed, it is even questionable whether such a hurdle is even worthwhile or meaningful on its own.

Posted by AI3's author, Mike Bergman Posted on July 18, 2011 at 10:00 pm in Adaptive Innovation, Semantic Web, Structured Web | Comments (3)
The URI link reference to this post is: http://www.mkbergman.com/965/in-the-midst-of-an-evolutionary-explosion/
The URI to trackback this post is: http://www.mkbergman.com/965/in-the-midst-of-an-evolutionary-explosion/trackback/