What is so interesting semantic network
And their problems.
Two days my fingers itched, I really wanted to give a post on a topic that surfaced in Habré with the help of yzh44yzh . What makes yzh44 in your project? It implements a specific method of navigation using such a concept as a network (graph). While his network with the semantic has not very much in common. The most important thing that is still lacking is qualified connections, without which the network cannot be called semantic. It is the qualified compounds that give the network value, and make the semantic network (hereinafter CC) as an interesting design.
')
I will not describe how the SS is determined, anyone (who does not know) can spy on Wikipedia (although I personally do not like what is written there). SS is excellent for describing anything (mostly static structures of course) and is commonly used in a reduced form in computer science. What is UML? This is a meta-meta-language for modeling software products, consisting exclusively of CC elements. What is ERD? Same. What is MindMap? Guess yourself.
Since any domain can be described with the help of SS and ontology, the SS has found its main application in the field of knowledge engineering and the presentation of so-called “knowledge”, which is nothing but a dictionary or encyclopedia. But at the same time, the SS-description of the subject area can be used as a basis for a specific “software” product for the same area. Examples: idea- & innovation management, documentation mgnt., Customer relationhip, task- & todo mgnt. etc. etc. Scope of coverage is immense. Ideally, it is only necessary to define the ontology of the domain in order not to go far, the project yzh44yzh, so we will get a system for managing tasks without additional “coding” .
It is for this reason that SS research has been carried out all over the world since the 1960s, but unfortunately without much success. The goals of the utopian connection of the SS with the “inference engine” (inference engine) for obtaining “artificial intelligence” I do not even closely consider - this goal is currently not attainable.
But another reason for the failure of the SS-based software idea lies, oddly enough and not surprisingly, in a completely different place. Namely: in the presentation of the structure of the SS to the end user. All “researchers” think that the network should be presented as it is: namely, as a network, as a graph. This path works until the network is small and when the context of a specific node is not complex, it has a small complexity, up to 8-10 nodes. But as soon as the network becomes larger (and in fact even not large subject areas, such as task management, can grow to incredible sizes), the user is gone, frustration and rejection of the product begins. Of course, this problem occurs not only here. For example, if you look at the “overgrown” UML diagrams, then the same thing, looking at the context, you need to torment yourself with interconnections for a long time.
There can be only one solution: divide and conquer. The only way (in my opinion) is to create an interface that is not very different from today's programs and use common navigation tools, for example, like on web sites and having only navigation value. This will give the user a “sense of security” in the use of the product, at the same time prevent its loss in the network and create a so-called “perspective”, a certain look at the network.
On this occasion, I propose a game in the conceptual solution of the problem for real computer scientists (following the example of the great exbitionist in LJ): who decides that he will become famous and maybe I (or someone else, for example yzh44yzh, ngupigop ili padm ) use your idea as intended :)
I have already mentioned the framework, if this idea finds a response from you, I will show the way in the mid-end of the next week, which I once did, but somehow I didn’t get to the end;)
Examples of existing implementations can be found in my comments on the post yzh44yzh.
PS
I have absolutely no “scientific burden” and claims for “scientific nature”.
The post was written quickly, the formals were not respected, it turned out quite vague. I apologize. I promise to fix it :)
Pps
Please write about errors in PM, I will fix it.
Two days my fingers itched, I really wanted to give a post on a topic that surfaced in Habré with the help of yzh44yzh . What makes yzh44 in your project? It implements a specific method of navigation using such a concept as a network (graph). While his network with the semantic has not very much in common. The most important thing that is still lacking is qualified connections, without which the network cannot be called semantic. It is the qualified compounds that give the network value, and make the semantic network (hereinafter CC) as an interesting design.
')
I will not describe how the SS is determined, anyone (who does not know) can spy on Wikipedia (although I personally do not like what is written there). SS is excellent for describing anything (mostly static structures of course) and is commonly used in a reduced form in computer science. What is UML? This is a meta-meta-language for modeling software products, consisting exclusively of CC elements. What is ERD? Same. What is MindMap? Guess yourself.
Since any domain can be described with the help of SS and ontology, the SS has found its main application in the field of knowledge engineering and the presentation of so-called “knowledge”, which is nothing but a dictionary or encyclopedia. But at the same time, the SS-description of the subject area can be used as a basis for a specific “software” product for the same area. Examples: idea- & innovation management, documentation mgnt., Customer relationhip, task- & todo mgnt. etc. etc. Scope of coverage is immense. Ideally, it is only necessary to define the ontology of the domain in order not to go far, the project yzh44yzh, so we will get a system for managing tasks without additional “coding” .
It is for this reason that SS research has been carried out all over the world since the 1960s, but unfortunately without much success. The goals of the utopian connection of the SS with the “inference engine” (inference engine) for obtaining “artificial intelligence” I do not even closely consider - this goal is currently not attainable.
But another reason for the failure of the SS-based software idea lies, oddly enough and not surprisingly, in a completely different place. Namely: in the presentation of the structure of the SS to the end user. All “researchers” think that the network should be presented as it is: namely, as a network, as a graph. This path works until the network is small and when the context of a specific node is not complex, it has a small complexity, up to 8-10 nodes. But as soon as the network becomes larger (and in fact even not large subject areas, such as task management, can grow to incredible sizes), the user is gone, frustration and rejection of the product begins. Of course, this problem occurs not only here. For example, if you look at the “overgrown” UML diagrams, then the same thing, looking at the context, you need to torment yourself with interconnections for a long time.
There can be only one solution: divide and conquer. The only way (in my opinion) is to create an interface that is not very different from today's programs and use common navigation tools, for example, like on web sites and having only navigation value. This will give the user a “sense of security” in the use of the product, at the same time prevent its loss in the network and create a so-called “perspective”, a certain look at the network.
On this occasion, I propose a game in the conceptual solution of the problem for real computer scientists (following the example of the great exbitionist in LJ): who decides that he will become famous and maybe I (or someone else, for example yzh44yzh, ngupigop ili padm ) use your idea as intended :)
I have already mentioned the framework, if this idea finds a response from you, I will show the way in the mid-end of the next week, which I once did, but somehow I didn’t get to the end;)
Examples of existing implementations can be found in my comments on the post yzh44yzh.
PS
I have absolutely no “scientific burden” and claims for “scientific nature”.
The post was written quickly, the formals were not respected, it turned out quite vague. I apologize. I promise to fix it :)
Pps
Please write about errors in PM, I will fix it.
Source: https://habr.com/ru/post/29336/
All Articles