experiment round 1: identify optimal length for Summariser S-1

Several experiments until optimal S-1 length found.

How: * create a collection of notes: C-n (url) (C-1, C-2...) * we already have "short summaries" created using Tim's simple prompt (Summariser 0) to chatgpt right now * but I cant send their notes * manually create file of their notes: "C-n-AllNotes" * using rss to clean notes prompt * manually create summaries using Summariser S-n of length X1 of their notes: "C-n-S-n-X1" * use Prompt-P-x (P-1, P-2) * x = 1 * in myhub: use short summaries (S-0) to create response to Prompt-P-x: C-n-S-0-response-P-x * in chatgpt:
* send "C-n-AllNotes" with Prompt P-x to create C-n-AllNotes-response-P-x * send "C-n-S-n-X1" with Prompt P-x to create C-n-S-n-X1-response-P-x * compare quality of reply to P-x operating on Collection-C-n: which one does best? * Responses (evaluation input files): * C-n-P-x-S-0-response * C-n-AllNotes-response-P-x * C-n-S-n-X1-response-P-x * Evaluation file: C-n-P-x-evaluation * if AllNotes is best * then * X2 = X1 + 20% * repeat with X2 * better than AllNotes? * y: exit * n: X3=X1+40%, repeat * Once having found the optimal length for Prompt 1, use that in round 2

experiment round 2

We have the optimal length for Summariser 1 (S-1-X) using Prompt 1. * Test it against the other Prompts (P-x = P-2, P-3, ...): is C-1-S-n-X-response-P-x as good as C-1-AllNotes-response-P-x? * Test it using other Collections (C-n = C-2, C-3, ...)

Take stock, design round 3

experiment process 2

• As above but add variation by Summary-X-N (Summary-X-1, Summary-X-2...)
  • Summary created of X length using summariser N

---

Backup

Comparisons

Collection C-1

Description: total notes: - short summaries: URL: * Collection-C-1-AllNotes - Collection-C-1-Summaries-X

!Prompt 1 title

Responses

• in myhub: using short summaries - Collection-C-1-P-1-response-shortsummaries
• in chatgpt: using
  • AllNotes: Collection-C-1-P-1-response-AllNotes
  • Summaries of length X: Collection-C-1-P-1-response-Summaries-X

evaluation

Comparisons

Collection C-1

Description: total notes: - short summaries: URL: * Collection-C-1-AllNotes - Collection-C-1-Summaries-X

!Prompt 1 title

Responses

• in myhub: using short summaries - Collection-C-1-P-1-response-shortsummaries
• in chatgpt: using
  • AllNotes: Collection-C-1-P-1-response-AllNotes
  • Summaries of length X: Collection-C-1-P-1-response-Summaries-X

evaluation

---

Archive

Collection C-1

prompt	Goal	Short Summaries	Full note	250Note	Comparison
Prompt 1 title	goal	Collection 1 - Prompt1-ShortSummaries	Collection 1 - Prompt 1-FullNotes	Collection 1 - Prompt 1-XNotes	Collection 1 - Prompt 1 - Comparison
Prompt 2 title	goal	Collection 1 - Prompt2-ShortSummaries	Collection 1 - Prompt 2-FullNotes	Collection 2 - Prompt 2-XNotes	Collection 1 - Prompt 2 - Comparison

Prompt 1 & Collection 1

!Prompt 1 title

Short Summaries

Input - Collection 1: URL Collection 1 - Prompt1-ShortSummaries

Full Notes

Input: Collection 1 - Full Notes Collection 1 - Prompt 1-FullNotes

X Notes

Input: Collection 1 - X Notes Collection 1 - Prompt 1-XNotes

Prompt 2 & Collection 1

!Prompt 2 title

Short Summaries

Input - Collection 1: URL Collection 1 - Prompt2-ShortSummaries

Full Notes

Input: Collection 1 - Full Notes Collection 1 - Prompt 2-FullNotes

X Notes

Input: Collection 1 - X Notes Collection 1 - Prompt 2-XNotes