Unit 6: Expert system design

Activity 1: Review

Work in pairs. Without referring to the course website or any notes, take it in turns to answer the following questions. After each, answer agree or disagree with your partner. If you disagree, explain why.

1. What are the necessary features of an expert system?
2. Can you explain why ChatGPT is not an expert system?
3. Can you provide a specific example that shows the different between the two types of chaining?
4. What is contained in the knowledge base?
5. Is implicit knowledge the same as tacit knowledge?
6. What are the five phrases in knowledge engineering?
7. What is a frame?
8. What is the difference between first order logic and fuzzy logic?

Activity 2: Rule-based expert system explanation

Listen to this clear and concise explanation of expert systems.

The video is 14 mins 15 secs long. The presenter uses many of the target vocabulary items that are the focus of this course.

Activity 3: Evaluation of expert systems to detect AI-generated texts

The images below are representations of expert systems created by course participants. Each system has both pros and cons. Discuss each system focussing on the benefits and drawbacks. Where possible, suggest ways to improve each system. You can open the images in a new tab to enlarge them.

Evaluate each of the expert system designs by considering the following:

1. Does the system provide expert answers for each instance?
2. Can the system be converted into a program without further analysis?
3. If necessary, what additional information or details are needed to create a working prototype?

Activity 4: Terminology specific to authorship

Read.

Use ChatGPT and/or Google to understand the following terms when used in the field of authorship analysis. Share your findings with your partner/team.

1. linguistic features
2. shared linguistic features
3. questioned document/text/corpus/dataset
4. known document/text/corpus/dataset
5. reference corpus/dataset
6. word frequency
7. keyword frequency
8. part-of-speech (POS)
9. permutation

If you cannot understand any terms even though you checked ChatGPT and/or Google, ask your tutor.

Activity 5: Authorship analysis expert system

Read.

This is an extension to the scenario presented in the slide in the Problem Breakdown. Three use cases are presented which should help clarify what the expert system needs to do.

1. Use Case 1: To identify the shared linguistic features of an individual author across a collect of short messages, emails, reports and blog entries.
2. Use Case 2: To identify the author of a questioned document by comparing the the linguistic features in the questionned document to the writing of three different known authors. The system should compare the top twenty features and then rule out one candidate. The system should then compare features twenty-one to forty, and then rule out the next candidate.
3. Use Case 3: To verify whether the same author wrote different messages on the dark web using pseudonyms. Using a keyword trigram approach, authorship can be established
4. Use Case 3: To attribute authorship based on patterns of parts of speech following keywords in the questioned text and a known text.

Activity 5b: Additional information

Listen to your tutor explain this slideshow. Work with your parter/team when prompted.

Activity 6: Design a detailed expert system

Work in teams to create a detailed design of the expert system that can be used to attribute or verify authorship in the Use Cases using word frequency, keyword frequency and keyword and pattern frequency (Here pattern means the POS permutations followign the keyword).

Submit your work on ELMS.