Archive for January, 2008

In part 1, I offered some context and my definition of ‘semantics’ as being the system of relationships that are important or significant to people and which are symptomatic of human experience. In this part, I will flush out what this means.

Though most people may be familiar with that famous quote: I think, therefore I am, from the chapter Discourse on Method in Rene Descartes’ philosophical classic, he changed his mind later in his life. Looking at later works of his, we can find that he really was not looking for truth through the lens of reason or inference. He was looking for a certainty so clear on its face that it is self-evident.

“So, after considering everything very thoroughly, I must finally conclude that the proposition, I am, I exist, is necessarily true whenever it is put forward by me or conceived in my mind.” –from: Meditations of First Philosophy (from Wikipedia entry Cogito Ergo Sum ).

This kind of certainty grounds us or anchors us in reality. It is only a small step to go from I am, I exist to I am, I physically exist— the body being self-evident. It makes sense and seems quite reasonable that if a computational model is to formulate an outlook similar to the human perspective it must reckon with the conditions imposed on the physical being. Being primarily a body positioned in space, is a condition most certain of being human.

An ontology of the human condition begins by addressing the conditions of being human that are symptomatic of the physical state of existence of the human being. For this we can leave all philosophy, linguistics and psychology behind for the time being. I can describe my physical existence in simple first-person terms. The description is simple and characteristic of people I know.

I have weight and mass and I exist in a three-dimensional space. Most human beings have a body with a head, neck and torso and limbs. I have two arms with hands and two legs with feet, this is considered normal for a human. I have a front and a rear. I have a left side and a right side. I have a face on the front side of my head, an ear on each side and only hair on the back of my head. Usually, I am sitting or standing vertically with my feet on the ground beneath me and the sky overhead. All of this is self-evident.

Range of motions/visionMoreover, I have power and I can control myself. I can move my eyes to see at wide angles. I can move my limbs and my entire body through space within limits. I can only move in one direction at a time, for example

I have the power to exert a physical influence in the space surrounding my body. I can reach out in front or to the side within limits. I can grasp things, pull them to me or push them or throw them away.

Given that description of physically being human, let’s consider the semantic elements of human experience from this perspective. These are none other than the several elements of our human experience that are symptoms, or characteristic signs, of individual physical existence.

Remember, we are looking for universal semantic elements. These will not be a part of us, like our faculties or our eyes or opposing thumbs. They must be so indispensable that if we were without these elements we could not exist. These elements will be part of the natural state or condition of being human. It should be something that most certainly affects our ways or has the power to influence our perspective. In my view, there should be two sets of universal semantic elements.

The first of these semantic elements stems from our body-centered horizontal and vertical reference in three-dimensional space (four- if you count time as a dimension). This quality is not unique to human beings though it is universal property of the existence and experience of human beings. The significant fact here is that by being a point or center of reference, one’s orientation in space influences one’s view and perspective. Orientation affects every person, i.e., it is symptomatic of anybody and it is important to physical and psychological and interpersonal relationships.

The second of these semantic elements is might. Being out-of-control, or even experiencing a loss of control can be one of the most dreadful experiences for anyone. Control is very significant to people. People go after and forcefully exercise control. They get control or take control and hold on to control. Personal power is symptomatic of anybody.

Again, though it is not unique, it is a universal property of the experience of being human. Having the power control our limbs, to stand upright, to move, is important to everyone. For most people self-control is necessary for the purpose of meeting personal goals, and controlling their relationships to situations as well as their relationships with others. Once again, all of this is self-evident.

So we can conclude that there are at least two sets of semantic elements that are symptomatic or characteristic signs of the human condition and all the things and events that make up the human experience: One set of elements pertains to power and control (over life and limb) and the other set pertains to body-centered orientation in a physical space as well as a psychological space. While these may be self-evident properties, they have never been formalized into a computational model by other scientists working in the field.

What is not self-evident is the ways thinking processes seem to be interrelated with movement of the body and parts of the body. Thinking processes seem to be related to stages and processes involved in motion and movement. Not many scientists investigating the mind, language and human psychology would endorse this view. One need only peruse the Wikipedia entries for linguistics, cognitive science, human psychology, studies of mind to see what I mean. Not that it matters much, when engineers get together and talk about methods and procedures, best practices and the development of standards, they tend to stick to more concrete science.

Computational engineering is an engineering discipline and so when we started, we were looking for concrete grounds that neither linguists nor philosophers could deliver. We were not looking for the foundation of interpersonal relations between people, though we were looking for abstract relations within human expressions in a natural language. We were looking for certainty and an anchor even though we did not know what it looked like. This required the faith that we would know it when we found it and it took years of dedication and hard work to turn it into an semantic search and information filtering system for computers.

When Tom Adi began his scientific investigation into natural languages he was determined to search for natural laws instead of relying on linguistics or philosophy or any social science. The original scientific study that led Tom to produce his semantic theory was recently published in the book Semiotics and Intelligent Systems Development. The natural laws Tom was looking for were needed to satisfy his original premise that the elements and relations of natural languages correspond with the elements and relations of other natural systems at all times.

As I demonstrated above, the notions of Orientation and Control can be abstracted directly from human nature. They abstractly name universal properties of a natural, physical and human system of relationships. They correspond to the idea of polarity in language and to the strata of power and control enumerated in Adi’s theory of semantics. So the result is that we have something very solid, a concrete basis, for enlisting computational assistance in the interpretation of perceptions and intelligence and interpersonal understanding.

I will get into specific examples in the next part though it should not be difficult to recognize these semantic elements as the basis for the way we see and perceive the world. They also influence the actions we take and the interactions we cause to take place. For those that cannot wait, there is a formalization of this framework we call the “semantic matrix” that we have been using since 1986. The semantic matrix formalizes these elements according to their phonetic correspondents and the linked paper is a good introduction into some of the mathematical aspects of Tom Adi’s semantic theory.

Understanding humans and human events, society and culture, is about understanding their semantic elements and their relationships to the symphony of life and survival in an unforgiving and non-subjective world. The motivation for searching for certainty and universal elements of human relationships are the benefits afforded by finding harmony, simplicity and clarity in one’s understanding.

Universal properties are common tools in mathematical, i.e., computational, algorithms. By understanding their abstract properties, one obtains information about any constructions and can avoid repeating the same analysis for each individual instance. Readware technology applies this concrete foundation in its indexing and search algorithms.

When we began writing our first search engine, in 1985, we were breaking new ground in information science. For the first time, we would index text not by keywords but by abstract concepts corresponding to the way we naturally interpret the world. By fixing the boundaries of the information to the space enveloped by relations framed by orientation and control and by mapping syntactic elements from text in corresponding ways, we predicated we could map queries onto text and classify documents with more relevance than other methods.

I cannot say that I knew the exact correspondence then, as I feel do now, though I could see how it would affect relevance because it better related to the human condition. We tested Readware’s capability for achieving relevance in independently judged recall and retrieval exercises. We also tested Readware for commercial endeavors and we have worked on several commercial ventures with big and small companies.

One of these, commercial ventures had us working on RSS news feeds while we hosted Feedster’s service for a short while. I wrote about this experience in an earlier post. Internet statistics show that Readware had a substantial impact on page views and reach for Feedster. Like all business ventures, this one was subject to forces that had little to do with technology and technology alone was not able to save it.

In the third and final part, I will present additional references and depending on the disciplines of those following this article, I will add some additional and anecdotal information. As usual, I would like to hear what you think. While you are welcome to email me your comments, you can leave your comments for everyone to see by clicking on comments at the top of the page.

Read Full Post »

I promised in the last post that I could offer a solution to the disconnect between what search engines locate and what people think is relevant. Now there is nothing wrong with search engines as long as you know what you are looking for and it has a uniquely relevant name or handle. Some search engines are better than others and most are very good at what they do. The disconnect happens when you do not know what to look for, or you are looking for interrelationships that are often ambiguously expressed.

Besides the search process itself, there are many applications that could benefit from more intelligent computer processing. The man-machine interface could be vastly improved if computers could only understand what is important to people, how they talk, what they mean. To most computer scientists this means that the terms we use need to be defined.

This is true, yet, sometimes a word definition is not enough to inform the computer process. Try looking for trends, for example. Knowing the definition of a trend and how it is used in a sentence is not sufficient for spotting a trend in in a series of reports or documents. We need to know the semantics, or the significant elements and relations, of trends. Matching the word or a synonym to the word in an index of terms is only somewhat helpful.

In this post I will tell you about the technology I have helped produce and mostly how the underlying techniques address the root problem of semantics, or rather the lack of semantics, that is holding back development of more intelligent systems. Most experts would agree that a unified semantic theory is necessary for progress to be made in intelligent computing platforms and programs.

Since the mid-1980’s, I have been working on this problem with my associate Dr. Tom Adi. Together, we have produced at least a half-dozen products from the semiotic and intelligent techniques we have developed. Most of these have to do with processing and classifying plain text, web pages, documents, messages, etc., The technology is real, proven scalable and adaptable, and we have paying customers using it in one form or another.

The techniques we are perfecting achieve some, not all, of the same objectives as the semantic web; particularly in the case of semantic search techniques. We support or intend to support all the W3C standards. As it stands today we have what can be called a RESTful interface delivered over http services. The technology is called Readware technology.

The Readware IpServer is a set of software services and an application programming interface (API) that provides text and data indexing, classification and a search and retrieval engine, for computers. It is a programmable semantic search engine with its own resources. Unlike semantic web technologies, the language and knowledge resources are provided in the package and the meta-data is automatically generated by the engine.

Unlike most uniformed search engines (ala Google) that match search terms to indexed terms, Readware categorizes terms and phrases while indexing and can equate them and inform the search with measures of fidelity, kinship and affinity between them. These measures encompass linguistic functions, such as synonymy and word disambiguation, and more. However, Readware does not use the kind of rudimentary logical inference or natural language processing that our modern counterparts Hakia or Power-Set do.

Instead, Readware simply and intelligently maps text according to a cognitive model with semantics that are reflective of interpersonal relationships. Readware search algorithms conduct a competitive exegesis to locate relevant bonds, connections and associations between the elements in the information space. We know the connections and associations found will be mostly relevant because they are reflective of interpersonal relationships that are important or significant to people and which are symptomatic of human experience.

Let me define what I mean by interpersonal relationships. Interpersonal relationships are the body of relationships that form (and break):

  1. between people, and;
  2. between their perceptions and beliefs about the events and happenings around them, (often set in text) and;
  3. on account of their own actions and interactions in the world.

Of course everyone thinks and acts differently, each person has their own experiences and point of view, beliefs and perceptions. Even so: in my experience there is a universal and unified framework, a system and semantics that underlies and rules the ways people tend to think, speak and act along with the interpersonal relationships that fall between people.

I am not saying I can tell what people are thinking and I am not talking about religion. I am talking about the fundamentals of interpersonal relationships. These are ontological relationships that stem from the human condition and what it means to be human. These relationships have more to do with personal control, affinity, consanguinity, engagement and interaction with others, trust, and the making and breaking of bonds, than the truth values of the conventional structure of sentences.

In any case, as I have stated quite clearly where our semantics are seated, and that they are rather different from those of our contemporaries, let me start back at the beginning.

I started working with computers in the late 70’s. I became involved in the design of “information-systems” that delivered actionable information. At that time computers were being used as a kind of store and forward device. It seems not much has changed. Initially, I was working on the input, editing and communications side. It was not until after 1981, when I became involved with the output side, that I realized computers should do more than only store and forward. They should classify, filter, route and perform other intelligent functions.

I began looking into what it would take to make more intelligent programs, particularly in text and language processing. If I had not been so naive at that time of my life, I might have realized that I did not even know what it means to be intelligent; neither did anyone else. I felt we were all looking for theories about intelligence, human psychology and language use.

When I met Tom Adi in the early 80’s, I was intrigued when he told me his hypothesis about meaning and interpretation of language: that the elements and relations of natural languages should abstractly correspond to elements and relations of other natural phenomena at all times. When I asked him how he defined the elements of the system, he said that he had not figured that out yet. I will come back to this later, let me just say a few things about the qualities of some systems that everyone reading will recognize.

One of the things that define a system is that it is comprised of a set or body of elements that have specific interconnecting relationships. Examples are our own internal circulatory or metabolic systems, and the solar system. The body, heart, veins and blood are interrelated. Each planet in the solar system has a relationship to each other and to the sun.

And of course nearly everyone reading should be familiar with the personal computers and the relationship between the movement of the mouse and the pointer on the computer screen, for example. These sorts of indispensable elements and consequential relationships are what I am talking about. These indispensable elements and consequential relations are significant or semantic elements and relations.

In natural systems the significant (semantic) relationships are fixed, or invariant, stable or constant, and they are universal in their application. That the planets continue to exist day after day and that they revolve around the sun in the order and way they do is inherited from the earliest days of planet formation. The initial configuration and relationships are inherited through time and space. There is a “natural” order to the solar system that does not seem to change over time– though everything changes in time, certain elements and relations, as I have described above, seem rather stable, constant and predictable, and universal.

It can be hard to think of human behavior as systematic, or to consider that human interactions and activities have specific elements and relationships like those between the planets, their moons, and the sun of our solar system. On the surface it seems not to be so. Yet the truth is that there are fundamental, ontological, elements and relations of interpersonal activities. Because the acts of speaking, reading and writing are natural phenomena and also forms of interpersonal activity, it makes sense that there are elements and relations of natural languages that correspond to elements and relations of interpersonal relationships.

It is probable that when systems of writing were first devised that the semantic relationships, between the sounds comprising a word and the abstract associations inspired by sound combinations, were well understood by those devising writing systems. Yet because sounds degenerate over time and natural languages change and exchange names and words, those relationships were confused and even forgotten, even though the systems of writing, though reformed, survive today.

This is because phonetic systems of writing, like our own western alphabet, were humanity’s first interpersonal memory. They were not designed to preserve the conventions, trappings and sophistry of the philosophers and keepers of human languages. They were designed to preserve personal (and abstracted) thoughts, experience and human memories.

With the advent of the alphabet, for the first time in history, people were able to transmit and exchange their abstract thoughts and their own desires with other people. The Romans became fond of writing and reading love poems and romance novels soon after they acquired their writing system. Western language, science, law, religion and most artifacts of “human culture” began to blossom and spread with the spread of written languages.

Yet, while written languages are widely viewed as “systems of writing,” many modern linguists have not considered the relations between elements of written language (letters) or their relationships to sounds and to words, or anything else, as worthy of scientific inquiry. This explains the disconnect between the ways computers compute semantic relationships in text and the ways people, read, recognize and think about the world at large.

Computer scientists and programmers really cannot be blamed. They depended and relied upon linguists. Linguists, being conventional in all regards to language, and being preoccupied with the relationships within sentences, were either unable or unwilling to consider the interrelationship between abstract thought and the ways people act, read and write.

Today, our modern writing systems are highly refined instruments and devices clothing the thoughts, opinions, politics and religions of the human race. That brings us back to interpersonal relationships, and beliefs and perceptions, which are the factors driving society and culture, and to our interactions in the world. The interrelationships between all these things can be found in text, in literature and in messages and instructions.

The Readware platform is based on scientific research into the interrelationship between natural languages like Arabic, English, German and Russian, for example, and the ways people use words to preserve ideas, culture and the human society. It is not the same thing as looking for relationships in external artifacts such as the truth values between an arbitrary sentence and a possible world. The technology implements the ontology and semantics in a computing platform that can be used by anyone that can use a computer.

In the next part I will tell you more about how we found the system of elements and relations in language and how they correspond to human thinking and activity. In the meantime, feel free to drop me a comment or ask questions.

Read Full Post »