Projects

This is a list of possible seminar projects.

1. The default project is to take a passage containing an "unknown" word, represent it in SNePS, determine what prior knowledge is needed to compute a meaning for the word from context, represent that knowledge, and run the appropriate definition algorithm on it, modifying either the representation or the algorithm, as needed, to get it to work. For details, see the webpage "Programming Project"

2. Revise the noun-definition algorithm.

3. Revise the verb-definition algorithm.

4. Possibly merge the N and V algorithms?

5. Revise the adjective-definition algorithm.

6. Create or modify a computational grammar for one of the demos.

7. Re-implement CVA belief revision (i.e., the "smite" and "dress" examples) using Johnson & Shapiro's AutoBR and/or Bhushan's theory of "Preferential Ordering of Beliefs for Default Reasoning" [PDF] instead of Cravo & Martins's SNePSwD.

8. Link SNePS with the CYC commonsense knowledge base.

9. Link SNePS with the WordNet knowledge base.

10. Implement a theory of "internal" context, i.e., using morphological or etymological information to compute a meaning for an unknown word.

11. Analyze the kinds of background knowledge that seems to be needed, especially very generic rules.

12. Explore other machine-learning techniques
    • Especially Version Spaces?.
    • Examine the relevance of Valiant's "Theory of the Learnable"

13. Rewrite the definition algorithms using SNeRE, so that Cassie will "consciously" create defs, rather than "unconsiously" use her Lisp definition-"calculator".

14. Create a "master" demo: Combine all PK files and all words into a single demo, to see what happens on a large scale.

15. Experiment with using passages/words from the MICASE on-line corpus of spoken text.

16. Re-implement using SNePSLOG [.ps].

17. If you have ideas for other projects, please discuss them with me.