Artificial Intelligence Frame: Fundamentals and Applications

By Fouad Sabry

What Is Artificial Intelligence Frame

Frames are a form of data structure used in artificial intelligence that reflect "stereotyped situations" in order to facilitate the division of knowledge into substructures. In his article "A Framework for Representing Knowledge" published in 1974, Marvin Minsky made the initial suggestion for them. The fundamental data structure that is utilized in artificial intelligence frame languages is known as a frame, and frames are kept as ontologies of sets.

How You Will Benefit

(I) Insights, and validations about the following topics:

Chapter 1: Frame (artificial intelligence)

Chapter 2: Knowledge representation and reasoning

Chapter 3: Ontology (computer science)

Chapter 4: Semantic Web

Chapter 5: Web Ontology Language

Chapter 6: Symbolic artificial intelligence

Chapter 7: Logic in computer science

Chapter 8: Knowledge-based systems

Chapter 9: Reasoning system

Chapter 10: Glossary of artificial intelligence

(II) Answering the public top questions about artificial intelligence frame.

(III) Real world examples for the usage of artificial intelligence frame in many fields.

(IV) 17 appendices to explain, briefly, 266 emerging technologies in each industry to have 360-degree full understanding of artificial intelligence frame' technologies.

Who This Book Is For

Professionals, undergraduate and graduate students, enthusiasts, hobbyists, and those who want to go beyond basic knowledge or information for any kind of artificial intelligence frame.

• Language: English
• Publisher: One Billion Knowledgeable
• Release date: Jun 26, 2023

Book preview

Chapter 1: Frame (artificial intelligence)

By describing stereotyped circumstances, frames are an artificial intelligence data structure that divides knowledge into substructures. In his 1974 article A Framework for Representing Knowledge, Marvin Minsky made the suggestions. In artificial intelligence frame languages, frames serve as the main data structure and are kept as ontologies of sets.

In both knowledge representation and reasoning techniques, frames play a significant role. They belong to the class of knowledge representations known as structure-based since they were originally created from semantic networks. A vast taxonomic hierarchy, similar to a biological taxonomy, is created by grouping together data about specific object and event kinds in structural representations, according to Russell and Norvig's Artificial Intelligence: A Modern Approach..

The frame includes instructions on how to use it, what to anticipate after that, and what to do if those predictions are not satisfied. While some of the data stored in the frame is typically static, data saved in terminals typically changes. Terminals are a type of variable. Terminals do not need to be true; top-level frames include information that is always accurate regarding the issue at hand. With the discovery of new facts, their worth could change. The same terminals may be used by different frames.

Each piece of data pertaining to a specific frame is stored in a slot. The data can include:

Facts or Data

Values (called facets)

Procedures (also called procedural attachments)

Deferred evaluation, IF-NEEDED

IF-ADDED: revises related data

Default Values

For Data

For Procedures

Subframes or additional frames

The default settings in a frame's terminals are based on how the human mind operates. For instance, most people will see a specific ball (such as a well-known soccer ball) when told a boy kicks a ball as opposed to visualizing an abstract ball with no characteristics.

Contrary to semantic networks, frame-based knowledge representations have the advantage of allowing for exceptions in specific circumstances. This offers frames a degree of flexibility that enables more accurate reflections of representations of real-world occurrences.

Spreading activation is a method for querying frames, just like semantic networks. Any value added to a slot that is inherited by subframes will, in accordance with the inheritance rules, be updated (IF-ADDED) to the appropriate slots in the subframes, and any subsequent occurrences of a certain frame will use that new value as the default.

Despite the lack of explicit arcs, a semantic network can be generated from a set of frames because frames are founded on structures. Generally speaking, Noam Chomsky and his generative grammar from 1950 are not mentioned in Minsky's work.

Frames' streamlined architecture make it simple to use analogous reasoning, which is a desirable ability in any intelligent entity. Additionally, the procedural attachments offered by frames provide a level of flexibility that enhances representational realism and provides a natural affordance for programming applications.

The simple analogy (comparison) that can be drawn between a boy and a monkey only by having slots with similar names is noteworthy in this case.

Another thing to note is that although Alex, an instance of a boy, inherits default values like Sex from the more inclusive parent object Boy, the boy may also have unique instance data in the form of exceptions like the number of legs.

Artificial intelligence uses a technology called a frame language for knowledge representation. Despite the fact that their core design objectives are different, they are comparable to class hierarchies in object-oriented languages. While objects concentrate on information encapsulation and information concealing, frames concentrate on the explicit and intuitive representation of knowledge. Objects and frames both have their roots in software engineering. However, in actual use, frame and object-oriented languages' features and approaches heavily overlap.

The Friend of A Friend (FOAF) ontology, which is a component of the Semantic Web and serves as the basis for social networking and calendar systems, is a straightforward illustration of a notion modelled in a frame language. This straightforward example uses a Person as the main frame. The individual's phone, home page, email, etc. are some examples of slots. Additional frames describing the space of commercial and entertainment domains can be used to represent each person's interests. Each individual has connections to others that the slot is aware of. The web of people a person is friends with can provide default values for their interests.

The first Frame-based languages were created from scratch for particular research projects, not as tools that could be used by other academics. Researchers quickly saw the advantages of removing a portion of the fundamental infrastructure and creating general purpose frame languages that weren't connected to particular applications, much like with expert system inference engines. KRL was one of the original general-purpose frame languages.

Automatic categorization and frame languages have seen a resurgence of attention as a result of the Semantic Web research agenda. The Web Ontology Language (OWL) standard for representing information on the Internet serves as one illustration. On top of the Internet, OWL is a standard that offers a semantic layer. Instead of categorizing websites using keywords as most apps do today (such as Google), the idea is to categorize websites using concepts arranged in an ontology.

A nice illustration of the benefits of a Semantic Web may be found in the name of the OWL language itself. Nowadays, the majority of pages returned from an Internet search for OWL would be on the bird Owl rather than the generic OWL. The user wouldn't have to worry about the different potential acronyms or synonyms as part of the search with a Semantic Web because it would be easy to express the concept Web Ontology Language in that way. Similarly, the user wouldn't have to be concerned about homonyms clogging up the search results with unnecessary content like details about raptors, as in this straightforward example.

In addition to OWL, other Semantic Web-related standards and technologies that were impacted by Frame languages include OIL and DAML. Ontology editing is offered by the Stanford University Protege Open Source software tool, which is based on OWL and has all the features of a classifier. However, as of version 3.5 (which is still supported for individuals who want frame orientation), it no longer explicitly supported frames; the version in use in 2017 is 5. Moving away from explicit frames is justified by the fact that OWL DL is more expressive and industry standard..

There is a large overlap between frame languages and object-oriented languages. The two communities had different terminologies and objectives, but as they transitioned from the academic and research worlds to the business world, developers tended to be less concerned with philosophical questions and more interested in specific capabilities, combining the best elements from both groups regardless of where the idea originated. The goal of both paradigms is to shorten the gap between ideas in the real world and how they are implemented in software. As a result, both paradigms came to the conclusion that the fundamental software objects should be represented in taxonomies, starting with very generic kinds and moving toward more particular types.

The relationship between common terms used by the object-oriented and frame language communities is shown in the following table:

The main distinction between the two paradigms was how much encapsulation was valued as a crucial element. Encapsulation was one of, if not the most, crucial requirements for the object-oriented paradigm. A major force behind the development of object-oriented technology was the aim to minimize potential interactions between software components and, as a result, manage huge complicated systems. This criterion was less important to the frame language camp than the goal to offer a wide range of potential tools to convey rules, restrictions, and programming logic. Everything in the object-oriented world is governed by methods and their visibility. So, for instance, using an accessor method is required to access the data value of an object property. The data type and other restrictions on the value being fetched or set on the property are among the things that this method regulates. The same kinds of constraints could be addressed in a variety of ways in Frame languages. The timing of triggers can be configured to occur before or after a value is set or retrieved. Rules could be created to handle the same kinds of restrictions. With the same kind of constraint information, the slots themselves might be enhanced with additional data (referred to as facets in some languages).

Multiple inheritance was the other key distinction between frame and OO languages (allowing a frame or class to have two or more superclasses). Multiple inheritance was necessary for frame languages. Human conceptualizations of the universe rarely fit into tightly defined, non-overlapping taxonomies, which is a result of the desire to represent the world in a similar manner to how humans do it. Single inheritance was either highly sought or necessary for many OO languages, particularly in the later years of OO. In order to maintain encapsulation and modularity, multiple inheritance was considered as a step that may be taken in the analysis phase of modeling a domain but that should be discarded in the design and implementation phases.

Psychological studies from the 1930s that suggested humans used stored stereotyped knowledge to evaluate and act in novel cognitive contexts served as inspiration for early work on Frames. Even the tiniest difficulty might have a sizable potential solution area in difficulties like these and many others. For instance, separating the phonemes from a raw audio stream or spotting an object's boundaries. Humans tend to oversimplify things when they are actually highly complex. In fact, it's likely that their true difficulty was not fully appreciated until researchers in artificial intelligence started looking into how tough it was to get machines to solve them.

The original idea behind frames, or scripts as they were originally known, was that they would create the context for a problem and thereby drastically decrease the potential search space. Schank and Abelson utilized the concept to demonstrate how an AI system could handle regular human interactions like placing a restaurant order. These exchanges were codified as Frames, each of which had spaces for storing pertinent data. In object-oriented modeling and entity-relational modeling, slots are comparable to object properties and relations, respectively. Slots frequently had default values, but they also needed to be refined more during the execution of each scenario instance. To put it another way, the process of carrying out a task, like placing an order at a restaurant, was managed by starting with a simple instance of the Frame and then instantiating and refining different variables as necessary. In essence, the frame instances served as an object instance while the abstract Frame served as an object class. The static data descriptions of the Frame were the main focus of this early study. Different procedures were created to specify a slot's range, default values, etc. However, procedural capabilities were present in these pioneering systems as well. One typical tactic was to connect triggers to slots, which are comparable to triggers in databases. Simply said, a trigger is procedural code that has been connected to a slot. The trigger could go off before or after accessing or changing a slot value.

Frames were arranged in subsumption hierarchies, just like object classes. A simple frame might be placing an order at a restaurant, for instance. Joe visiting McDonald's would be an illustration of that. A frame for ordering at a nice restaurant would be a specialty (basically a subclass) of the restaurant frame. In addition to adding more slots or changing one or more default values (such as expected price range) for the specialized frame, the fancy restaurant frame would inherit all the default settings from the restaurant frame.

Findings from experimental psychology and attempts to develop knowledge representation tools that matched the patterns people were assumed to employ to function in daily activities were major influences on much of the early Frame language research (e.g. Schank and Abelson). Mathematical formality wasn't as interesting to these academics since they thought it wasn't always a good representation of how the typical person thinks about the world. For instance, human language use is frequently not at all logical.

Similar to this, Charles J. Fillmore began developing his theory of frame semantics in the field of linguistics in the middle of the 1970s. This idea later gave rise to computational tools like FrameNet. The inspiration for frame semantics came from considerations of human language and cognition.

On the other hand, academics like Ron Brachman intended to provide AI researchers with the mathematical formalalism and computational power associated with Logic. Their goal was to translate set theory and logic to the Frame classes, slots, constraints, and rules in a Frame language. The use of theorem provers and other automated reasoning