Computational linguistics is powerless to help us in word processing.
Computer linguistics today has practically exhausted itself. This is directly indicated by the unsuccessful experience of researchers and developers of “intelligent” information products, who have been working for more than half a century to create such ambitious programs, such as adequate machine translation or semantic search for information in arrays of documents in natural language.
The future of machine processing of natural language texts is certainly seen in the creation and development of supra-linguistic technologies capable of analyzing the content of information at the level of semantic understanding of the context, just as man can do. However, the creation of “thinking machines” (Thinking Machine) for a long time was hampered by two main factors - the lack of the necessary methodology and the proper tools for solving two fundamental tasks - finding the “formula of meaning” and building a “model of knowledge about the universe” in some formalized computer-accessible form, without which, in fact, it is impossible to repeat at the program level the nature of human thinking.
Linguists, together with cybernetics, could not overcome these problems, since the latter lies already beyond the boundaries of their subject specialization, which, in fact, significantly slowed down the development of such long-requested text processing applications as, for example, the creation of “smart” dialogue systems or "semantic Internet search engines." And the same machine translation still leaves much to be desired.
')
The experience of the development of scientific and technological progress suggests that the breakthrough desired result ultimately results, as a rule, at the junction of various technological areas and subject disciplines. Apparently, the problem of “machine thinking” will be solved exactly when we understand exactly how our natural consciousness works in the procedural plan, and when we can reliably know whether these thinking procedures that we manifest in the necessary and sufficient quantity will work. final computer algorithms.
It should be noted that in recent years a new (“smart”) scientific discipline has begun to develop, which deals precisely with the fact that it studies the procedural nature of human mental activity. It can be said that at the moment we have a significant breakthrough in this direction and already quite clearly imagine how the algorithm of human thinking works. If we talk about this generally, then, first of all, it should be noted that a person thinks not in images, as it is usually thought, but “models of the behavior of images” (MPO). Secondly, we think “ontologically motivated”, that is, we continuously ask questions, even without noticing it, and constantly look for answers to them (also automatically). Finally, a meaningful understanding of everything that happens around an individual or in his consciousness during any contemplation is carried out precisely with the help of a certain “model idea” about the surrounding universe. This happens by comparing those IGOs ​​that it receives on an operational basis with those of the Universe stored in human long-term memory. Here, exactly, these three main whales make up all the technology of natural thinking, which now remains only to simply shift to a language understandable to programmers and get a long-awaited result.
When people comprehend any natural-language message, they practically never establish an instant correspondence of the expressed judgment with the concepts and models of the behavior of images stored in their memory. Each time they give the first associative heuristic correspondence that they receive (perceived) MPOs that occurs in their minds, based on the specifics of their experience and knowledge, and only then, during further rethinking of the text, they begin to clarify and specify the information received. Computational linguistics seeks to establish exact correspondences of the meanings of words, as well as their mutual relations, trying to overcome the problem of the ambiguity of verbal tools characteristic of any language, which, in fact, is very different from how our thinking operates. After all, a person achieves an understanding of speech or text not at all by knowing the morphological loads of words or establishing syntactic connections between words, and not even because he recognized specific meanings (semantics) of words, but precisely due to initial associative assumptions and the subsequent “iteration scrolling of the whole context "in order to draw the final picture of the correspondence of the perceived information to its internal content.
The future of machine processing of natural language texts is certainly seen in the creation and development of supra-linguistic technologies capable of analyzing the content of information at the level of semantic understanding of the context, just as man can do. However, the creation of “thinking machines” (Thinking Machine) for a long time was hampered by two main factors - the lack of the necessary methodology and the proper tools for solving two fundamental tasks - finding the “formula of meaning” and building a “model of knowledge about the universe” in some formalized computer-accessible form, without which, in fact, it is impossible to repeat at the program level the nature of human thinking.
Linguists, together with cybernetics, could not overcome these problems, since the latter lies already beyond the boundaries of their subject specialization, which, in fact, significantly slowed down the development of such long-requested text processing applications as, for example, the creation of “smart” dialogue systems or "semantic Internet search engines." And the same machine translation still leaves much to be desired.
')
The experience of the development of scientific and technological progress suggests that the breakthrough desired result ultimately results, as a rule, at the junction of various technological areas and subject disciplines. Apparently, the problem of “machine thinking” will be solved exactly when we understand exactly how our natural consciousness works in the procedural plan, and when we can reliably know whether these thinking procedures that we manifest in the necessary and sufficient quantity will work. final computer algorithms.
It should be noted that in recent years a new (“smart”) scientific discipline has begun to develop, which deals precisely with the fact that it studies the procedural nature of human mental activity. It can be said that at the moment we have a significant breakthrough in this direction and already quite clearly imagine how the algorithm of human thinking works. If we talk about this generally, then, first of all, it should be noted that a person thinks not in images, as it is usually thought, but “models of the behavior of images” (MPO). Secondly, we think “ontologically motivated”, that is, we continuously ask questions, even without noticing it, and constantly look for answers to them (also automatically). Finally, a meaningful understanding of everything that happens around an individual or in his consciousness during any contemplation is carried out precisely with the help of a certain “model idea” about the surrounding universe. This happens by comparing those IGOs ​​that it receives on an operational basis with those of the Universe stored in human long-term memory. Here, exactly, these three main whales make up all the technology of natural thinking, which now remains only to simply shift to a language understandable to programmers and get a long-awaited result.
When people comprehend any natural-language message, they practically never establish an instant correspondence of the expressed judgment with the concepts and models of the behavior of images stored in their memory. Each time they give the first associative heuristic correspondence that they receive (perceived) MPOs that occurs in their minds, based on the specifics of their experience and knowledge, and only then, during further rethinking of the text, they begin to clarify and specify the information received. Computational linguistics seeks to establish exact correspondences of the meanings of words, as well as their mutual relations, trying to overcome the problem of the ambiguity of verbal tools characteristic of any language, which, in fact, is very different from how our thinking operates. After all, a person achieves an understanding of speech or text not at all by knowing the morphological loads of words or establishing syntactic connections between words, and not even because he recognized specific meanings (semantics) of words, but precisely due to initial associative assumptions and the subsequent “iteration scrolling of the whole context "in order to draw the final picture of the correspondence of the perceived information to its internal content.
Source: https://habr.com/ru/post/142424/
All Articles