User:Moanamontes/Pan Lingua

Disambiguation:

1. Panlingua is an artificial language developed by Xul Solar in 19xx.

2. Panlingua is a theory of language developed by Chaumont Devin during the years 1994-2004.

The Panlingua Theory of Language explains how all of the internal data structures of language can be modeled on machines using nothing but a set of simple links and nodes.

History: After working as a commercial software developer for some years, Chaumont Devin became seriously interested in artificial intelligence (AI) during the latter part of 1985. He quickly realized that conscious human intelligence was based upon language, and that in order to come up with true AI it would first be necessary to master the workings of language. Driven by this realization, he temporarily dropped all other aspects of AI to focus upon Language and experimented with text-based chatterbots from 1986 to 1994, after which it became apparent that he had reached a point of diminishing returns for ever larger investments of programming time. Something was wrong, and he knew the problem was that although he was able to make the computer better and better at manipulating words, it never actually UNDERSTOOD anything it was doing. He had pondered the problem of semantics for some time, but did not wish to introduce the higher level of complexity it entailed until he had exhaustively studied the major areas involved in purely text-based word manipulations. But realizing that he had reached what was popularly known as the "stone wall of intractability" in this field, he decided to experiment with the semantic aspects of language in early 1994. At first he tried using templates (as did Doug Lenat in his "Project Cyc"), but it soon became apparent that this approach led to yet another stone wall of intractability. It was then that he came up with the idea of a knowledge-representation system he named "Interlinguish," which has survived in modified form until now. Interlinguish was originally based upon two-dimensional (dependency grammar style) arrangements of "atoms," which he also called "universal atoms of meaning," which were similar to the words used in John Soa's system called "Predicate Calculus." He then became aware of the work others had been doing on ontologies, and obtained a copy of WordNet from Princeton. Perceiving the errors that had kept others back from perfecting these systems and integrating them into the bigger picture of language as a whole, he then developed an ontology of his own and the software to manage it which he called SEMLEX. Then he proceeded to integrate this ontology into the whole system he was building to process human language and saw that what he had must be a model of what was really happening within the human mind. He also saw that unlike other kinds of computer programming in which, as one of his instructors put it, "There is always more than one way to skin a cat," linguistic programming (computational linguistics) would simply not work unless everything was exactly right. It was thus that he became aware that there must be some universal theory of language governing how language workson computers as well as inside human minds, so he named this underlying theory Panlingua, from Greek "pan-," meaning "all," and Latin "lingua," meaning "language" or "tongue." At that time he was not aware that this name had already been used for an artificial language developed more than a half century before by the Brazilian philosopher, Xul Solar, which has nothing whatsoever to do with this theory.

Major Discoveries:

As he struggled to work out the details of this universal theory of language, Chaumont Devin made several groundbreaking discoveries including those in the following list:

1. The linguistic link.  All the internal data structures of language can be modeled using nothing but simple linguistic links, each of which consists of a source, a type, and a destination.

2. All words appearing in grammatically correct phrases can be defined simply as two linguistic links emanating from the same node (having a common source). One is usually a link to another word within the same phrase, called a syntactic link, or synlink, and the other is a link to the intended meaning of the word, called a semantic link, or semlink.

3. Two internal data structures are required for an automated system to process natural language. These are (1) a lexicon/ontology, and (2) a corpus of parsed sentences.

Fundamental Theorem. All of the internal workings of language spring from the definition of "word," as given above, namely that every word in every coherent phrase is simply two linguistic links emanating from the same node. Armed with this fundamental theorem, it is possible to proceed in rigorous fashion through all of the other discoveries made during the development of the theory.

The Innate Weakness Of Previous Approaches. It is suprising to find that although the field of linguistics has been carefully studied since the time of Panini (ancient Sanskrit grammarian), the meaning of "word" has been taken for granted and not rigorously defined. Chaumont Devin began with what he termed "universal atoms of meaning" arranged in dependency trees to represent meaning without texts in computers, but later came to realize that these were nothing but words under a different name. He then began to make a sharp distinction between words and "external symbols," an external symbol being a sound that vibrates the air or a pattern of marks representing a word on a page. But he later realized that the meaning of "word" is quite different in different situations, and that it is necessary to accomodate all of these meanings while realizing that they are distinct from each other. We often say that we "here a word spoken" or "see a word written," etc., when what we are really talking about is an external symbol representing a word. These are not the real words of Panlingua, which are invisible within the human mind.

Understanding and Parsing. Panlingua Theory holds that for every coherently uttered word (every word that is part of a meaningful phrase) there exists a REAL word behind the external symbol, be it a word sound or a written word, and that this real word behind the symbol can be determined by understanding or "parsing" the phrase in which it is found. Now since every REAL word, or "universal atom of meaning," is just a syntactic link and a semantic link emanating from a common source, or source node, then to UNDERSTAND or to PARSE means simply to determine the correct syntactic and semantic links intended for each word. Once this has been done, the external symbol is no longer necessary, and (for the general case) can be safely discarded without losing any of the original meaning of the word. Additional information, such as tone of voice, emphasis, punctuation of written words, whether the word is quoted, etc., can be included as special augmentations in the internal representation of the word--in other words in addition to the two fundamental links determining the essential meaning of the word.

How Traditional Grammar Fits In. Recall that every linguistic link has a source, a type, and a destination. The "parts of speech" of traditional grammar are encoded in semlink type or semantic link type. The "syntactic roles" of traditional grammar are encoded in synlink type or syntactic link type in Panlingua. The dependency structures of dependency grammar are encoded by means of synlink source and destination--synlink source being dependent and synlink destination being regent or head.

Breakthrough Results. Using this theory of language, Chaumont Devin has been able to create systems that can quickly learn to interact at a fundamental level in most natural languages. Working examples can be found and demonstrated on http://panlingua.net.

For more in-depth information on Panlingua Theory, read http://panlingua.net/txt/pan.txt and consult the various other materials found on http://panlingua.net.