Search lies at the heart of the web, with over thirty trillion websites. The web provides us with a diverse and ever-increasing amount of data. The search, therefore, is how we navigate the web. We currently search for information can mean we jump from one website to another to gather all the data we need; this is because answers provided by these searches continue to .direct us to individual and isolated websites.
We often have to access information stored in our brains to solve problems and make decisions. But how is that information stored?
There are a lot of possibilities, but in this article, we will focus on the Semantic networks approach, here we are going to discuss the comparison of semantic networks in AI and partitioned semantic networks in AI.
Semantic Networks in AI
- Semantic networks are alternatives to the predicate logic for the knowledge representation technique.
- In semantic networks, we can represent our knowledge through graphical networks.
- This network consists of different nodes representing objects and arcs to describe the relationship between those objects.
- Semantic networks can categorize objects in different forms and can also link those objects.
- Semantic networks are easy to implement and understand and can easily extend.
- Knowledge can be stored in a graph, with the different nodes representing objects in the world and arcs representing relationships between those objects.
- Mathematically a semantic network can define as a labeled directed graph.
- Its representation consists of mainly two types of relations:
- IS-A relation (Inheritance)
- For nodes – Labels indicate the name – Nodes can be instances (individual objects) or classes (generic nodes).
Semantic networks were first introduced by Quillian back in the late-1960s. A semantic network is a simple knowledge representation scheme consisting of graphs of labeled nodes and labeled arrows to encode knowledge.
- Nodes: Objects, concepts, and events.
- Arcs: Relationships between nodes.
Semantic networks have graphical representation, which is a big reason for their popularity.
This representation consists of primarily two types of relations such as:
This representation consists of mainly two types of relations such as:
- IS-A relation (Inheritance): This relationship is mostly used in inheritance.
- Kind-of-relation: This kind of relationship mostly uses in general graphical representations.
Advantages of Semantic Networks in AI
- Easy to visualize
- Related knowledge is easily clustered
- This semantic, i.e. real-world meanings, are identifiable.
- Flexible and general: relationships can be arbitrarily defined by the knowledge engineer.
Challenges of Semantic Networks in AI
- General representation can be a problem if we are clear about the syntax and all the semantics in each case.
- No standards about nodes and arc values (no semantics in semantic networks)
- Limited expressiveness: May require several special codes and procedures
- Way to express non-monotonous knowledge (like FOL)
- No easy way to express n-ary relationships (reification needed)
- Several AI workers studied the limitations of conventional semantic networks extensively.
- Many believe that the basic notion is powerful and has to be complemented by, for example, logic to improve the notion’s expressive power and robustness.
- Semantic networks take more time to perform computations at runtime because we need to traverse the complete network tree to allow some questions.
- They try to model human-like memory to store information. However, it is practically impossible to build such a complex and vast semantic network.
- These semantic representations are inadequate as they do not have equivalent quantifiers, e.g., for all, for some, none, etc.
- These networks are not intelligent and depend on the creator of the system.
- Inheritance can cause problems:
- particularly from multiple sources, and when expectations in inheritance are required.
- Facts placed inappropriately cause a problem.
The above problems make it difficult to:
- Verify and validate the system.
- Share knowledge
- Reuse knowledge
- Acquire knowledge methodically
Partitioned Semantic Networks
Semantic Networks are simple representation technique that provides graphical representation. The knowledge we use can be more complex. It may contain complex facts and statements here, and semantic networks can only provide partially accurate results.
Partitioned semantic networks overcome the limitations of semantic networks. It is a type of semantic network.
Author: Huma Tariq
Copyright The Cloudflare.