Lecture
Это продолжение увлекательной статьи про словарь по искусственному интеллекту.
...
binary output; X i -resolving binary inputs; Z i - prohibiting binary inputs; a i and b i are weighting factors; h threshold By varying the values of the weight coefficients and the threshold, it is possible with the help of N.F. implement any boolean function. In N.F. inputs are associated with the neuron synapses, and the output with its axon. N.F. does not function as a biological neuron (the relaxation time is not taken into account, the latent period, which always occurs after the neuron triggers and during which it cannot receive input signals). But it is this model that is used in the design of many devices developed in neurobionics (for example, perceptrons).
The property that is typical for inference in an open formal system and in an open knowledge base, and that the previously derived statements may cease to be deductible when new facts appear.
The property of interpreting expressions when credibility scores are attributed to them, other than absolute truth and falsehood. Dealing with such expressions requires special means of recalculating likelihood estimates. In logical inference, when there is uncertainty, either multi-valued logic or plausible reasoning is used.
Uncertainty arising from the vagueness and ambiguity of verbal expressions. When describing quality knowledge, one has to apply special techniques to eliminate N.L. (See also Fuzzy Set, Affiliation Function, Variable Linguistic.)
Property description of the subject area, namely, that this description can not be transformed into a formal system. When dealing with incomplete information, plausible or default arguments are used.
A situation in which for a set of similar tasks it is impossible to find a general algorithm that solves them, although for subsets of this set it is possible to construct specific algorithms for finding solutions. The existence of such algorithmically unsolvable problems is strictly proved.
The technology of information processing and problem solving using computers, based on the achievements of artificial intelligence. The main idea used in BAT is to automate the procedure of building a program that interests the user, based on the task definition introduced by him into the description system of the problem in his usual professional language. Thus, N.I. provides the ability to communicate with the user's computer, which is not a professional programmer. In order for the main idea of N.I.T. to be realized, it is necessary that the computer has an intelligent interface, a knowledge base and a solver, i.e. would be an intellectual system. Another feature of N.I. is a distributed way of solving a problem, when users engaged in solving a common task communicate with each other through a computer network, e-mail and a common knowledge base. The network also includes databases from which users derive information for solving their problems.
The combination of real or abstract objects (entities), relations and relationships between these objects, as well as procedures for transforming these objects to solve problems arising in O.P.
A subject area whose conceptual model cannot be immersed in a formal system or a set of formal systems. Most subject areas that intellectual systems have to deal with are poorly structured.
The subject area, the conceptual model of which can be immersed in a formal system.
See subject area.
The set of techniques and methods that allow in knowledge bases to introduce new knowledge derived from existing due to clustering, the introduction of hypersites and hypotheses.
The process of developing a hypothesis, with the help of which a general pattern is described, connecting together isolated knowledge, acting as particular cases of this pattern.
A tool for designing and creating expert systems. The structure of O. includes tools for designing knowledge bases with various forms of knowledge representation and selection of the mode of operation of the problem solver. For a specific subject area, the knowledge engineer determines the desired knowledge representation and problem solving strategies, and then, introducing them into the O., creates a specific expert system.
One of the functions of an intelligent system, which is to prove or verify that the solution obtained by the system does not conflict with the knowledge stored in the system’s memory. Thus, O. is relative. When you change the contents of the knowledge base or database. A. can either keep her power or become wrong. O. is usually closely related to explanation. O. close the concept of argumentation.
The totality of the processes of analysis of texts in natural language, their understanding and synthesis of texts. In the process of analysis, morphological, syntactic and semantic text analysis takes place in the most developed systems for processing natural-language messages, as a result of which the deep structure of the text is revealed, which is translated into an internal representation used in the knowledge base of an intellectual system. The correlation of this structure with the knowledge that is stored in the system allows us to understand the meaning of the source text. During the synthesis of texts, the semantic structure of the text is first formed, which is then filled with linguistic units, taking into account the syntax and morphology of the chosen natural language. S O.E.Ya. connected with the task of machine translation, automatic summarization, communication with the user in a limited professional natural language, etc.
The process associated with the processing of visual information (changing the scale, the selection of the contours, the recognition of visible and invisible parts of the image, etc.).
A method for solving a problem in which the subtasks extracted from it are executed simultaneously. Due to the specially organized interaction of the processes of solving subtasks at the end of OP. the solution to the original problem is obtained.
A set of procedures used in image processing, a feature of which is that they operate on data representing the contents of one binary digit.
Image of a typical or generalized representative of a certain class of objects.
Fragment of knowledge, according to which the search is carried out according to the model, or the standard, according to which the classification of images, situations, rules, etc. takes place
The assimilation of knowledge, skills and abilities through or receiving and perceiving information from the teacher or processing the observed information with the subsequent construction on the basis of these observations of new general rules and patterns. Both forms of intelligence are used in intelligent systems to acquire new knowledge.
A type of training in which the individual or the intellectual system is presented with a set of positive and negative examples associated with any previously unknown pattern. In intelligent systems, decision rules are developed, with the help of which the set of examples is divided into positive and negative. The quality of the division is usually checked by examining a sample of examples. If the quality of the separation on the examination sample is satisfactory, then the developed decision rules are accepted by the system as final. If the examination was unsatisfactory, then the examination sample is added to the training and new decision rules are built. After this, the exam process is repeated.
The process of establishing and developing contact between people, generated by the need for joint activities and including the exchange of information, the development of a unified strategy for interaction and the perception and understanding of each other. In the broad sense, the concept of O. extends to the contact of a person with a computer, during which a certain task is solved.
The procedure for combining into a generalized hypothesis a number of hypotheses equipped with their likelihood coefficients.
One of the functions of an intelligent system. A. Provides the user with information about how the intelligent system received the decision issued to the user. In contrast, the basis of O. relies only on the route that has been preserved in the memory of the system from the process of finding a solution. Using this route, the intellectual system forms the user O. in a professional natural language, allowing him to present all the fundamental steps of the solution.
Restrictions imposed on a set of information units stored in databases and knowledge bases. These restrictions should be executed in any states that are determined by the current contents of the knowledge base and database.
See graphics dynamic.
One of the functions of an intelligent system. With the help of O., some solution of the system is justified not by logical reasoning or reference to the knowledge existing in the system, but by reference to the value structure existing in the system. O. convinces that this decision does not contradict this value structure. (See also Explanation, Justification.)
Search for errors in the knowledge base of the intellectual system. Distinguish between syntactic debugging and semantic debugging of the knowledge base.
Being in the program of semantic errors. O.S. is carried out by running a program under test on a computer with such initial data for which the correct solution is known in advance.
Identification of errors in the program in the language of knowledge representation, carried out automatically by the parser.
The assignment on the set M of the Cartesian product M '* M' M. The pairs included in M '* M' 'are elements of O., and the combination of these pairs forms the O. graph or its extension. A law may have a number of internal (reflexivity, symmetry, etc.) and some external semantics associated with its name. All this information forms the semantics of O. or its inference.
The term is explained in the article Relation reflexive.
The term is explained in the article Attitude symmetric.
The term is explained in the article Relationship Transitive.
The relationship is not explicitly present in the knowledge base, but determined from the relationships that are in the database.
Attitude, which describes the relationship of factors, events and phenomena in time. Examples O.V. can serve as "be earlier", "simultaneously", "end at the same time", etc. O.V. used in temporal logics, action logic, and other pseudo-physical logics.
Relationships that describe actions that take place in the real world. Examples OD can serve: "move to", "approach", "influence", etc. O.D. used in action logics, which are a type of pseudo-physical logics.
Attitude on a set of statements relating to the intensional component of knowledge about the problems of the region.
The relationship by which various types of cause-effect relationships between objects are described.
A relation defined by a special commutative diagram connecting the elements and operations of one set with the elements and operations of another set with the help of operators characteristic of fuzzy logic.
The term is explained in the article Relation reflexive.
The term is explained in the article Attitude symmetric.
The term is explained in the article Relationship Transitive.
The relationship by which spatial relationships of objects are described. Examples O.P. can serve: "to be close", "to adjoin", "to be inside", etc. O.P. used in spatial logics, action logic and other pseudo-physical logics.
A link between two or more information units, established on the basis of their semantic proximity.
The relation possessing that property that any its element with itself always is in this respect. Examples of O.R. can serve: "match", "be similar", etc. If the reflexivity property does not hold for at least one element of the relation, then it is called a non-reflexive relation, and if it does not hold for one element, it is an anti-reflexive relation.
The semantics used in knowledge bases are defined by its name.
A relation possessing the property that for any pair (A, B) of elements in this respect, it is true that the pair (B, A) is also in this respect. An example would be the spouse relationship for husbands and wives. The relation for which this property is not satisfied for at least one pair is called asymmetric, and if it is not fulfilled for any pair, it is antisymmetric.
Reflexive, symmetric and non-transitive attitude. Such an attitude can be interpreted as a similarity relationship. Unlike the equivalence relation, which divides the set of elements on which it is defined, into disjoint subsets, O.T. gives coverage to this set. FROM. used when classifying information in knowledge bases.
The relation for which the fact that the pairs (A, B) and (B, C) are in this respect always implies that the pair is in the same relation (A, C). If this property is not satisfied for at least one triple of elements, then the relation is called non-transitive, and if it does not hold for any triple of elements, then it is anti-transitive. An example of a transitive relationship is the "be older" relationship.
The relationship with which the knowledge base defines the relationships between information units. These relationships define the procedures for finding (calculating) one units through others.
Symmetric, reflexive and transitive relation. It is used to classify sets of elements by dividing it into disjoint classes, which collectively cover the entire original set.
Attitude on a set of specific facts stored in the database. (See also the extensional view.)
Single logical operation, denoted by a. For two-digit logic, it is defined as follows: if a is true, then a is false, and if a is false, then a is true. In many-valued logic, there are several types of negation. A direct generalization of two-digit negation for k-valued logic looks like a = (k - a) mod k.
See Negation.
See Actant.
See Actant.
Memory-oriented search for stored information on the content. It is organized using features (tags) that link data (information) according to their content, as opposed to conventional memory, in which information is searched by the number of the cell in which it is stored. In P.A. use pattern matching.
"Unlimited" RAM, which the user has. With the help of special system tools in a computer, a part of PV is projected. on the field of RAM. At the same time, the rest of P.V. stored on external media.
A dedicated memory area for storing icons — conditional graphic images of information objects or operations.
See Pipeline Architecture.
A set of procedures with the help of which the computer translates text from one language to another. These procedures implement the analysis of the source text, its grammatical (morphological and syntactic) analysis, translation of the text into a deep structure reflecting the meaning of the text. According to this deep structure, the corresponding text is constructed in another language. The text synthesis procedures are to some extent repeated in the reverse sequence of the analysis procedure. Currently there are systems PM. translate texts from a fixed subject area.
A variable that uses words and phrases that are characteristics of a phenomenon as its values. For example P.L. with the name "length" can take the following values: "tiny", "very small", "medium", "large", "very large". P.L. used when formalizing high-quality information when it is entered into the knowledge base. Extraflux logic values P.L. can act as special quantifiers.
Variable used in propositional logic.
Variable in logic, standing in the zone of action of a community quantifier or existence quantifier.
One of the components of the speech act, along with locus and illocution. The effect achieved by illokutsii.
The device is built on the neurobionic principle. The simplest three-layer P. consists of a field of photoreceptors, each of which can be in two states, the field of associative formal neurons and solvers. Black and white images are projected on the photoreceptor field. Photoreceptors are randomly connected to the enable and disable inputs of associative formal neurons, the outputs of which are also randomly connected to the inputs of the solvers. In the process of learning to classify input images, weighting coefficients and thresholds of associative neurons are selected so that solvers' input (they add signals coming from associative neurons) can be used as a source for separating a set of input images. In addition to the three-layer P., multilayer P. were also investigated, in which several layers of associative neurons were introduced. As studies have shown, the possibilities of P. in the field of classification are limited. Currently, interest in P. has almost disappeared.
See Perception.
The elementary part of the image on the display screen. P. characterized by brightness and color. P. size is standard. From P. builds an image on the screen.
Conditional graphic representation of information objects or operations.
The process of drawing up a sequence of actions, subtasks, operations, sub-goals, the successive implementation of which should lead to the achievement of the goals set for the system. Intellectual systems P. can be carried out either in the task space or in the state space.
See Planning.
Planning, in which at first an approximate plan is sought, with the help of which a fundamental answer is reached about the attainability of the goals set. Then this plan is gradually refined and brought to a level where it can be unambiguously realized.
Planning, in which the individual parts of the plan are formed in different places and by various means, and then combined by a central body. ETC. It is found in intelligent robots and other intelligent systems when they are included in a system for collecting, transmitting and processing data.
Building an initial action plan in hierarchical planning systems. The strategic plan takes into account only the basic information about the planning environment, but does not take into account the particular situation in which the planning takes place. Used then on the underlying levels.
Building an action plan in hierarchical planning systems. When P.T. The plan received at the level of strategic planning is amended to take into account the specific features of the current situation and the state of the planning system. Levels P.T. maybe a few.
A set of software tools designed to search for action plans.
The method of making optimal statistical decisions based on the assumption that the parameter of the distribution of probabilities of an observed random event that influences the nature of the decisions made is a random variable determined by the prior distribution. P.B. minimizes the average risk, i.e., the expectation is now associated with incorrect or inaccurate decisions. P.B. used in the theory of statistical solutions, game theory, pattern recognition theory and for plausible inference in intelligent systems.
Movement in a structured space from one nodes of this space to another. If P. is purposeful, then a set of initial nodes is given, from which P. can start, and a set of final (target) nodes, upon reaching which P. stops. Movement according to the structure of the search space is determined by strategic P. Among them, the most common are the upward search and the search downward, as well as the search in depth and the search in width. If the space of P. is not structured, then the only possibility for P. is a random search.
Search by pattern in knowledge bases. (See also associative memory).
Search in which the movement in the search structure occurs along one path to the end. If unsuccessful, another way is viewed. The backtracking procedure is used to return to the nearest branch point.
Descending search, when the search structure is set by decomposing large tasks into smaller ones, up to tasks whose solution is known to the system. The task to be solved is considered as a target node, the descending search for it goes to a set of tasks, the solution of which is known to the system. P.P.Z. used in intelligent systems when planning activities.
A descending or ascending search, in which the structure of the search space is given by the set of states of a certain system, between which possible transitions are indicated. P.P.S. used in intelligent systems when planning activities.
A search in which movement along a search structure occurs at a certain depth in all directions possible from a given search point. If all advances were unsuccessful, then either the search depth increases, or the search front narrows and further advancement along this entire front occurs by a specified number of steps. The narrowing of the front leads to the fact that P.Sh. combined with depth searching.
Search in which the movement in the search structure goes from the initial given nodes to the target nodes. In this case, you can use the search in depth, search in width, or a combination of them. In logical systems, an analogue of P.V. is a direct conclusion. Another name is P.V. - direct wave method.
Search for the necessary information in a large array of previously known set of features.
Search, in which the movement in the search structure occurs from the target nodes to the specified nodes. Pn can use depth search, width search, or a combination of them. In logical systems, the analogue of PN is the reverse conclusion. Another name PN - backward wave method.
Search for a piece of knowledge in the knowledge base based on a given sample. A sample can be a fully defined fragment, or it can contain free variables. For example, when presented as a semantic network, a sample of the first type might look like "Ivanov - To be born - 1965" , which means a direct query to the knowledge base: "Is it true that Ivanov was born in 1965?" . A sample of the second type: "X - Born - 1965" . It is interpreted as: "Name those who were born in 1965." Requests of the second type can be interpreted with the help of a special index as issuing any answer concerning one subject X, born in 1965, or as issuing all X characterized by this property. BY. is the main procedure for finding information in knowledge bases.
Search, in which at each node of the search space, the continuation is selected which has the local success rate best for the given node. Thus, P.P.C.L. is a type of depth search.
Search unstructured search space. When PS with the help of a certain probability distribution, space nodes are selected and checked to see if they are target. When reaching the first target node PS. is terminated.
See Search Depth.
See search wide.
In artificial intelligence, a set of models and procedures that help intelligent systems correlate incoming text in natural language with pieces of knowledge from the knowledge base, as well as procedures that allow others to be derived from existing knowledge necessary for correct interpretation of the contents of the entered text.
A name assigned to an entity class that is joined by a community of attribute structures. In the logic of P. are strictly defined and unchanging entities, characterizing only the significant structure, immanent inherent in all concepts. In artificial intelligence, as well as in the everyday practice of people, P. is understood more widely. Not only the relevant structures, but also the results of using P. in the activities of people or in the functioning of an intellectual system can take part in the formation of a law. It is in this sense that P. are used in reasoning about the activities of people and about the functioning of intellectual systems. For the formation of P. in intelligent systems, various methods of generalization are used.
The process of obtaining from the facts stored in the database, new information units in the knowledge base.
See Text Generation.
A sequence of data continuously supplied with a pipeline architecture to an operating device that performs the same type operations on it.
Rules establishing the relationship of conjunction and disjunction.
Typical examples:
a & b = ¬ (¬a V ¬b),
a V b = ¬ (¬a & ¬b).
See Products.
The rule is with the help of which in formal systems from a set of axioms correctly constructed formulas are generated, which are interpreted as true.
Fuzzy inference rule based on composition operation. This operation converts the likelihood coefficient of the original premise (by multiplying by a specially selected matrix) into the likelihood coefficient of the conclusion. There are such matrices that PVK turn into. modus ponens and modus tollens.
In the formal system, a rule that determines the method of forming syntactically correct expressions. In linguistics PS. allow you to separate the syntactically correct sentences of this language from those that are not.
In the predicate calculus, a special sign that reflects a certain relationship between a finite set of entities - arguments. In the usual variant of the predicate calculus, two values are taken as the value of P. on the set of marked arguments: truth and false.
The location of the data in the physical environment is fixed by a special database schema. Using this scheme, the database management system correlates the request for finding the necessary information with the physical location of the data. Different principles for constructing a database schema lead to different types of PD. (relational, hierarchical and network).
Formalization of knowledge for their entry into the knowledge base. At the conceptual level of P.Z. knowledge models in the form of semantic networks, frames and production systems are most common. In pz as a direction of artificial intelligence, traditionally also includes the task of checking the knowledge base contents for correctness and completeness, knowledge replenishment by inference based on the knowledge available in the knowledge base, generalization of knowledge, and classification of knowledge.
Representation of constant facts that do not contain free variables in databases or knowledge bases.
The statement, the truth of which is a prerequisite of the truth or falsity of another statement. For example, the two statements "Kepler died in poverty" and "Kepler did not die in poverty" have the same presupposition that Kepler existed.
Метод логического вывода, в основе которого лежит приведение доказываемого утверждения к множеству дизъюктов и поиску в этом множестве пар, один дизъюнкт которых содержит некоторую литеру, а другой - отрицание этой литеры, для их последовательного устранения из исходного множества. Если этот процесс через конечное число шагов приводит к пустому дизъюнкту, то вывод успешен. В противном случае формула недоказуема.
Совокупность методов и процедур, которые применяет инженер по знаниям при заполнении им базы знаний. P.Z. предполагает использование источников знаний двух типов: пассивных и активных. К первым относятся официальные документы, инструкции, печатные издания, кино-фото-документы и многие другие источники, в которых содержатся сведения, важные для описания знаний о предметной области. Ко второму типу источников знаний относятся люди - специалисты в данной предметной области. Инженер по знаниям с помощью специальных психологических методик и инструментальных средств в процессе диалога получает от экспертов необходимые сведения. Все приобретенные знания для ввода в базу знаний формализуются в соответствии с требованиями той модели знаний, которая соответствует выбранному проектировщиком системы представлению знаний.
Программа позволяющая использовать ЭВМ в качестве одного из участников игры. P.I. составляются как для игр типа шахмат, шашек и т. п., так и для реализации на ЭВМ развлекательных игр (погоня, рулетка и др.).
Программа, в основу которой положены соображения о том, как данную проблему решает человек.
Процесс представления алгоритма решения задачи в виде, "воспринимаемом" ЭВМ. Программирование включает детализацию алгоритма уровня элементарных операторов; запись алгоритма на выбранном языке программирования и описание процессов управления ходом выполнения программ на ЭВМ.
Программирование, при котором программа представляется в виде процедуры логического вывода в исчислении предикатов первого порядка. Механизм вывода обычно встроен в язык П.Л. Примером может служить распространенный в интеллектуальных системах язык ПРОЛОГ, в который встроен обратный вывод.
Programming, in which a program is treated as a set of objects and messages circulating between these objects. This approach introduces modularity into the program.
Programming, in which the solution of the problem is reduced to the calculation of the values of recursively nested functions.
The process of compiling heuristic programs.
A way of presenting procedural knowledge in the following most general form: (i); Q; P; C; AB; N. Here (i) is the proper name (label) of P .; Q is the sphere of P.'s application, isolating some part of it from the subject area, in which the knowledge contained in P. makes sense; Р is a precondition containing information about the truth of a given P., its priority, etc., used in the output control strategies for selecting this product for execution; C is a condition representing a predicate, the true value of which permits a given P to be applied at some step; A B is the production core (the core interpretation can be different, for example: "If A is true, then B is true", "If A is in the knowledge base, then B must be entered into the knowledge base", "If A is the current situation, then B " , etc.); N - P.'s postcondition, containing information about what changes need to be made to this P. or other P., included in the system of products, after the implementation of this P.
Proposal, judgment, statement. (See also Calculus Propositional.)
A structure reflecting the decomposition of large tasks into smaller ones, up to standard ones, the solution of which is supposed to be known. P.Z. used in intelligent systems for planning activities and in problems of automatic program synthesis. The solution of the necessary task is sought as a composition of the solutions of standard problems.
Used in psychology, a formalized way to identify the semantic proximity of concepts used by people. To build a P.O. binary scales are used, the ends of which are marked by the words-antonyms of the type "safe - dangerous", "wide - narrow", "good - evil", etc. On these scales a certain number of positions are plotted. Subjects must have points on them that correspond to the words uttered by the experimenter. The results are subjected to statistical processing using factor analysis or cluster analysis. Based on numerous experiments of this type, a three-dimensional PO was constructed, the axes of which are interpreted as a generalization of the scale of assessments, strength and activity. This space uses the usual metric of vector spaces. In P.O. concepts related to one another by a common situation are grouped in some thickenings, which confirms the idea that the principle of situationality is the basis for the classification of knowledge in people.
The structure on knowledge, in which the concept of "semantic distance" is introduced. An example of P.S. can serve space Osgood.
The set of states in which a technical system or process may be located. In PS a metric can be set, and possible trajectories of the change of states under the influence of various reasons are indicated. P.S. used in intelligent systems in the automatic synthesis of programs and in planning activities.
A set of goals with an indication of possible trajectories of their achievement. P.TS. used in intelligent systems when planning activities and in the automatic synthesis of programs.
The possibility of simultaneous withdrawal in the formal system of approval and its negation. In such a formal system, it is possible to derive any statements according to the rule of modus ponens.
The contradiction found in the model. In closed models of P.M. is an analogue of absolute inconsistency, in open models of P.M. is relative. (See also Kripke Model).
A frame in which the filling of some (or all) slots is such that it allows various instantiations of these values.
The establishment of inconsistency (impracticability) of the formula consisting of a conjunction of premises and a denial of the conclusion.
The procedure to which the reference is possible by name used in some information unit. On this behalf, P.P. can be simply called and attached to the description of the information unit or updated and executed.
A complex process consisting of a set of separate subprocesses whose interaction is not synchronized in time.
The processor is adapted to work with associative memory.
See Database Machine.
A device or set of programs focused on the implementation of user communication with the system in a limited natural language.
See Inference Processor.
A specialized processor (system of processors) that implements a set of procedures necessary for organizing a logical inference or extracting consequences from knowledge of some subject area.
A specialized processor that provides parallel execution of operations on arrays of numbers: vectors or matrices. Usually consists of a set of arithmetic processors that perform the same operations on various elements of an array, with a common control device.
Specialized processor focused on processing symbol information.
Direction in joint psychology, in which the central place is occupied by the questions of displaying knowledge in the cognitive structures of memory, the study of these structures and their influence on decision-making and the behavior of subjects. In P.K. Great attention is paid to the relationship between verbal and figurative components in the processes of memorization and thinking. PC. uses analogs between the representation and processing of knowledge by man and the intellectual system. In particular, using the methods of PK cognitive structures such as semantic networks, frames, production systems were analyzed, and it was shown that the organization of knowledge in humans is based on other structures.
The presence of the algorithm for solving the problem.
The scientific direction, the main task of which is the creation of models, methods and means related to the solution of problems of classification, taxonomy, the formation of concepts, etc.
One of the types of perception in intelligent systems. In the processor R.R. the input acoustic signal is analyzed, phonemes, words, lexemes, standard pieces of text that are correlated with the information stored in the system knowledge base allow the system to understand the entered text. There are R.R. at the level of individual words pronounced by the standard speaker, as well as with the adjustment of the system to the peculiarities of the pronunciation of a particular speaker, and RR, relating to the fused text. At present, the RRD systems have been implemented, which allow one to confidently determine the value of 1–2 thousand words, as well as analyze the merged text, in which not too large dictionaries are used.
Evaluation of the "semantic" proximity of information units stored in the memory of an intellectual system or person. The concept of "semantic" proximity is ambiguous. Experiments with people show that, depending on the goal, R.S. can be interpreted as situational proximity (meeting in some typical situations), associative proximity (such as hammer-nail), taxonomic proximity (such as table-bed), etc. R.S. used in knowledge bases to speed up the selection of relevant information for a given concept or situation. To date, there are no satisfactory models within which RS could be measured. (See also Osgood Space.)
The method of obtaining the conclusion on the basis of parcels and auxiliary considerations. The extreme case of R. is a logical conclusion in which the will of the subject does not play a role. In other cases, R. reflects the personal motives and interests of the person who conducts R.
The kind of non-monotonous output when the deducibility of an assertion depends on the context in which it exists. Examples of R.A. There may be numerous exceptions in general human practice from general rules in selected special contexts (situations).
Reasoning based on the schemes of obtaining conclusions adopted in hermeneutics, taking into account the structure of the text, on the basis of which a conclusion is drawn. For example, in the phrase "All the seas kissed his ships" (K. Balmont) there is a clear ambiguity. But with R.G. this ambiguity is removed, because one of the hermeneutics schemes for texts written in Russian says that the subject must be mentioned before the object to which its action is directed. And therefore, the subject of approval is the "sea", not the "ships". R.G. not strictly logical, but based on tradition.
One type of plausible reasoning is not based on grounds that are true in some formal system and not considerations that appeal to human experience, intuition, and faith.
Reasoning in an open model. Due to openness, it is possible to add new information from the outside in the process of conducting reasoning. This leads to the fact that some of the steps of reasoning, correct before the appearance of this information, may become incorrect. When reasoning is a strict conclusion, there is a non-monotonic conclusion.
Transferring the conclusions obtained on the basis of a number of parcels to another set of parcels, which is read by some criteria similar to the first one. In the particular case of R.A. there is a way to get a conclusion based on the Leibniz A, A 1 chart; B, B 1 ; T, T 1 ; G, G 1 ), where A 1 = T (A); where T is some transformative operator; G - homoformizm between A and B; G 1 - homoformism between A 1 and B 1 ; T 1 - the transforming operator B 1 = T 1 (B). R.A. there is a finding In 1 on the known remaining elements of the diagram.
The reasoning is based on the fact that the conclusion regarding one object is transferred to another object that has an associative connection with the first one. This relationship may have a different character (for example, an association of similarity, simultaneity, occurrence in similar situations, etc.). R.A. is a plausible reasoning, the degree of its likelihood is determined by the materiality of the used associative connection.
One type of plausible reasoning where the result is not derived from the premises that are clearly present, but based on "tradition", past experience, internal moral or value beliefs, etc. RU. arise when part of the information is missing in the input information and the intellectual system replenishes it on the basis of special information stored in its memory, intended for cases of incompleteness of the input information. For example, in the frames there may be special slots to which the system applies for information when something is missing to conduct reasoning.
The reasoning, which is based either on knowledge that does not have an absolutely true character, or on methods of reasoning, which are not absolutely true. Usually the result of R.P. provided with an assessment of its likelihood. Examples of R.P. reasoning by analogy or association, or hermeneutical reasoning.
The formation of a clause (ct - lt) V (dt - lt), where c and d are clauses that do not have common variables; l and l are a counter pair of letters, each of which belongs to its disjunct; t is the most common unifier of the counter pair of letters.
A technique used in deductive inference, which consists in finding two clauses, one of which contains a letter, and the other - its denial. On the basis of this comparison, a new disjunct is formed, called the resolvent. The generation of new clauses is the basis of the resolution method, widely used in intelligent systems.
A system capable, thanks to the general strategy of finding a solution embedded in it (for example, by searching in the space of alternatives or by logical inference) to find solutions to problems. R. enters as the main unit in intelligent systems. Sometimes this unit is called complete: Task Solver.
The method of studying the structure of individual consciousness by filling in a specially composed matrix. The rows of the matrix correspond to Kelly constructs. The columns of the matrix correspond to objects arranged similarly to the roles in the play. The method allows us to estimate the strength and direction of relations between the constructs for the person who filled the matrix.
A collection of generalized information on the structure and functioning of the main units and on information and control links between them for a processor with a reduced instruction set implemented on VLSI. The composition of the processor commands is determined by the specific area of its application, and this structure includes the most frequently used short commands that are executed in one machine cycle. In the structure of the processor, a large number of general-purpose registers are used (the data is read from the register and written back into the register), which significantly reduces the access time to the RAM.
A technical device capable of planning expedient behavior in a dynamic, not fully known environment in advance. R.A. should have a knowledge base about the environment and its features, a problem solver with the means to analyze situations and the consequences of their actions in the environment in order to accumulate information about how to act in certain situations. R.A. is a representative of intelligent systems.
Technical device in which there is a developed system of "eye-hand", which allows the observed to coordinate the situation with the movement of manipulators and means of movement. This distinguishes R.I. from robot manipulators, in which there is no feedback from the environment, if the situations differ from the standard ones previously recorded in the design of the robot manipulator. R.I. should be able to analyze visual scenes and be able to make decisions based on this analysis. R.I. is a representative of intelligent systems.
Autonomous robot, in which there are all the basic units characteristic of an intelligent system. With their help, the functions of RI communication are realized. with external partners, behavioral programs are built, knowledge of the external environment and actions in it is accumulated, behavior plans are built to achieve the desired goals.
The process of cleaning the memory associated with the detection of unused program memory blocks and attaching them to the free memory space for reuse. Garbage collection is a necessary process in any system that works with dynamic memory allocation.
A fact used to increase or decrease the likelihood estimate of some hypothesis. C. used in production systems, in which products are hypotheses.
In the narrow sense - the rule of the logical transition AB, which is interpreted as follows: if A is true, then B is also true, if A is false, then nothing can be said about B. In a broad sense, S. coincides with the concept of the core products.
1. One aspect of semiotics. Examines the relationship of signs to signified (content of characters) regardless of who serves as the addressee of the sign.
2. The value of individual units of the mark.
3. The study of individual units of the language - linguistic semantics, the elementary object of study of which is the unity of three objects: the signifier, the signified and the denotation. The signifier is the external element (a sequence of sounds or signs), the denotation is the signified object of reality and the signified sign is a reflection of this object in the consciousness of man.
Attributing to certain objects stored in the knowledge base of certain characteristics depending on the situation in which these objects are observed or used. In systems of understanding texts in the natural language of S.S. associated with the assignment of different values to the tokens, depending on the context in which they are used.
The science that studies the property of signs and sign systems (mainly natural and artificial languages.) Three main aspects of the study are distinguished: syntactics, which studies the internal properties of sign systems without regard to interpretation; semantics considering the relation of signs to their signified, irrespective of the peculiarities of the interpreter of signs, pragmatics studying the problems of the interpretation of signs.
Five H = 1 , C>, where A is the set of vertices, B is the set of names (weights) of the vertices; P is the set of arcs connecting pairs of vertices; P 1 is the set of marked input and output arcs; С - set of names (weights) of arcs.
A semantic network in which relationships point to associative connections between vertices characterizing objects, facts, and situations for the described domain.
A structure that displays the sequence of applying output rules to source assumptions. Due to the ambiguity of the choice of rules, at each step a variety of paths arise, forming S.V.
A semantic network in which arcs characterize relationships used in causal logic.
Expansion of context-free grammar by introducing context into special named registers with a stack structure and introducing special procedures that control the analysis by checking the context by the progress of the analysis is managed by the programmer). See also. The grammar is formal.
A model for describing asynchronous parallel and non-deterministic processes, as well as production-type systems. Statically, a model is defined by a double-sided orthograph with two types of vertices — positions and transitions (usually depicted by circles and shelves, respectively), and transitions (positions) can be connected by arcs only to positions (transitions). The initial state of S.P. set by initial marking of some of its positions. Markers are represented by dots inside positions. The dynamics is entered by an agreement on the rule of triggering an excited transition (containing at least one marker in each of its input positions), which can be triggered after an unknown time, after which all (in all) of its input positions (output positions) are removed (added) one marker. The process of functioning consists in the transition from one marking to another through the operation of excited transitions.
A semantic network in which all relations between vertices are treated as a cause-effect relationship, that is, as a non-reflexive, antisymmetric, and transitive relation.
The network, at the top of which information units are located, and arcs characterize the relations and connections between them. S.S. is the most common model of knowledge representation.
Semantic network, which reflects intensional knowledge about the subject area. This knowledge refers to the general laws of the region and operates not with constant facts, but with statements about the region containing variables that can be signified in specific situations. With the frame representation of the SSI. Corresponds to the frame prototype.
Semantic network, which reflects the extensional knowledge of the specific situation in the subject area. In SS.E. all the vertices of the network correspond to specific objects, and the connections between them to specific connections that are observed in the description of the situation. When frame representation SS.E. corresponds to frame instance.
A network that allows you to connect the processor in an arbitrarily specified manner, including each with each.
A special form of reasoning from the general to the particular. C. is a conclusion that follows from two assumptions containing statements about the ratio of the volumes of two classes or about the belonging of a certain element to a certain class. The following inference can serve as an example of S.: All predators feed on meat. Wolf is a predator. Wolf feeds on meat.
The architecture of a computing system with several identical processors
продолжение следует...
Часть 1 Dictionary of Artificial Intelligence
Часть 2 DENOTAT - Dictionary of Artificial Intelligence
Часть 3 Nemonotonnost with the withdrawal - Dictionary of Artificial Intelligence
Часть 4 SYNTAX - Dictionary of Artificial Intelligence
Comments
To leave a comment
Artificial Intelligence. Basics and history. Goals.
Terms: Artificial Intelligence. Basics and history. Goals.