Intelligent information system (IIS)

An intelligent information system (IIS) is a complex of software, linguistic and logico-mathematical tools designed to accomplish one principal task: supporting human activity and retrieving information through an advanced dialog in natural language. An IIS is a kind of intelligent system and, at the same time, one of the types of information systems.

Classification of IISs

• Expert systems
  • Expert systems proper (ES)
  • Interactive banners (web + ES)
• Question-answering systems (called "communication systems" in some sources)
  • Intelligent search engines (for example, the Start system)
  • Virtual conversational agents
  • Virtual digital assistants

An IIS can be hosted on a website, where the user either asks the system questions in natural language (if it is a question-answering system) or finds the required information by answering the system's questions (if it is an expert system). As a rule, however, expert systems on the internet perform advertising and information functions (interactive banners), whereas serious systems (such as an expert system for equipment diagnostics) are used locally, because they carry out specific, narrowly defined tasks.
Intelligent search engines differ from virtual conversational agents in that they are fairly impersonal and, in response to a question, return a digest of knowledge sources (sometimes a rather large one), while conversational agents have a "character" and a distinctive manner of communication (they may use slang or profanity), and their replies must be extremely concise (sometimes reduced to nothing more than emoticons, where that suits the context).

IISs used to be developed in logic programming languages (Prolog, Lisp and so on), whereas today various procedural languages are used. The logico-mathematical support is developed both for the individual modules of a system and for interfacing those modules with one another. Today, however, there is no universal logico-mathematical system capable of meeting the needs of every IIS developer, so one has to either combine the accumulated body of experience or design the system's logic oneself. In linguistics there are also many problems: for instance, in order for a system to work in dialog with the user in natural language, it must be equipped with algorithms for formalizing natural language, and that task has proved far more difficult than was assumed in the early days of intelligent systems. Another problem is the constant mutability of language, which must necessarily be reflected in artificial intelligence systems.

Support required for IIS operation

• Mathematical
• Linguistic
• Informational
• Semantic
• Software
• Hardware
• Technological
• Staffing

Classification of the tasks solved by IISs

• Data interpretation. This is one of the traditional tasks for expert systems. Interpretation is understood as the process of determining the meaning of data, the results of which must be consistent and correct. Multivariant analysis of the data is usually provided for.
• Diagnostics. Diagnostics is understood as the process of assigning an object to some class of objects and/or detecting a fault in some system. A fault is a deviation from the norm. Such an interpretation makes it possible to treat equipment failures in technical systems, diseases of living organisms and all kinds of natural anomalies from a single theoretical standpoint. An important specific feature here is the need to understand the functional structure (the "anatomy") of the system being diagnosed.
• Monitoring. The main task of monitoring is the continuous interpretation of data in real time and the signaling of any parameters that go beyond their permissible limits. The chief difficulties are "missing" an alarm situation and the inverse problem of a "false" alarm. These problems are hard because the symptoms of alarm situations are fuzzy and because the temporal context has to be taken into account.
• Design. Design consists in preparing specifications for the creation of "objects" with predefined properties. A specification is understood as the entire set of required documents: drawings, an explanatory note and so on. The main problems here are obtaining a clear structural description of the knowledge about the object and the "trace" problem. To organize design efficiently — and redesign even more so — it is necessary to produce not only the design decisions themselves but also the reasons for making them. Design tasks therefore tightly couple the two main processes carried out within the corresponding expert system: the process of inferring a solution and the process of explaining it.
• Forecasting. Forecasting makes it possible to predict the consequences of certain events or phenomena by analyzing the available data. Forecasting systems logically infer the probable consequences of given situations. A forecasting system usually employs a parametric dynamic model whose parameter values are "fitted" to the given situation. The consequences derived from this model form the basis for forecasts with probabilistic estimates.
• Planning. Planning is understood as finding plans of action for objects capable of performing certain functions. Such expert systems use behavioral models of real objects in order to infer the consequences of the planned activity.
• Instruction. Instruction is understood as the use of a computer to teach some discipline or subject. Instructional systems use the computer to diagnose the mistakes made while studying a discipline and suggest the correct solutions. They accumulate knowledge about a hypothetical "learner" and that learner's typical mistakes; in operation they are then able to diagnose weaknesses in a student's knowledge and to find suitable means of eliminating them. In addition, they plan each act of communication with the student according to the student's progress, with the aim of transferring knowledge.

Neural networks are not programd in the usual sense of the word; they are trained. The ability to learn is one of the main advantages of neural networks over traditional algorithms. Technically, training consists in finding the weights of the connections between neurons. In the course of training, a neural network is able to reveal complex relationships between input and output data and also to generalize. This means that, if training is successful, the network will be able to return a correct result for data that was absent from the training set.

• Control. Control is understood as the function of an organized system that maintains a particular mode of activity. Expert systems of this kind control the behavior of complex systems in accordance with given specifications.
• Decision support. Decision support is a set of procedures that provide the decision maker with the information and recommendations needed to make the decision process easier. These expert systems help specialists select and/or formulate the required alternative among a multitude of choices when responsible decisions are being made.

In general, all knowledge-based systems can be divided into systems that solve analysis tasks and systems that solve synthesis tasks. The main difference between analysis tasks and synthesis tasks is that in analysis tasks the set of solutions can be enumerated and built into the system, whereas in synthesis tasks the set of solutions is potentially unbounded and is constructed from the solutions of components or sub-problems. The analysis tasks are data interpretation, diagnostics and decision support; the synthesis tasks are design, planning and control. The combined ones are instruction, monitoring and forecasting.

Intelligent automatic control systems

Where information is incomplete or fuzzy, where external influences cannot be determined and where the operating environment is unknown, systems are built using unconventional approaches to control. They rely on the methods and technologies of artificial intelligence. Four basic intelligent technologies are distinguished:

• Expert system technology.
• Fuzzy logic technology.
• The technology of neural network structures with an implicit form of knowledge representation.
• Associative memory technology.

Principles on which intelligent automatic control systems are organized:

• Close information interaction between the system and the real world over information communication channels.
• Allowance for probable changes in external influences from the real world and in the system's behavior in response to them.
• A hierarchical multilevel structure following the principle that intelligence increases and accuracy requirements decrease as the rank in the hierarchy rises.
• Mandatory retention of operability when links with the higher levels are broken.
• Growth of intelligence and improvement of the system's behavior.

See also

• Expert system
• Hybrid intelligent system
• Business rules engine
• Logic programming
• Artificial intelligence
• Virtual digital assistant
• information system