Second Query …
Here I’m not interested in the UUID per se.
Rather I got interested in using the Crockford with the 3 word lists to generate random phrases.
This would be as part of a “Cut-Up” machine I’ve mentioned before and have partly finished.
Q: Is it possible to do the following with your tool (just wondering) …
Best wishes
TT
So you want to create “random-like” titles that make sense. IMO to create meaningful titles, we need to follow NLP rules - right? That’s not as simple as it seems. – or are we able to split at spaces and randomise words from many different titles.
Most of those full random titles will not make any sense. They will be random, but that’s it.
Yes. They don’t need to make sense. Just be serendipitously evocative.
This thread is about UUIDs and I really don’t want to spoil it with side questions!
I was pleasantly surprised by your comments.
They have helped me.
FYI, I just thought the Crockford mnemonic would be useful in my Cut-Up experiments. They look very useful!
I’d like to know how to get Crockford mnemonics as Titles as my cut-up approach is doing everything else through Titles.
Right.
But with limited word lists within a Domain Of Interest they may get better?
The Crockford lists require …
32 adjectives, 64 nouns, and 32 verbs respectively
I will likely use the Aarne–Thompson–Uther index of Folktales to derive sets of word indices that will work okay together. For example …
huge giant rampages, small kitten bleets … etc.
I have a few technical questions (just about the word indices) …
I will understand if you split this post to a different thread.
My interest is not about UUIDs per.se.
It is about uses of Crockford Mnemonics.
Best, TT
It seems the ATU indices seem to be massive.
I had a look at: https://vocabularyserver.com/atu/en/index.php?tema=163&/supernatural-adversaries but do not really understand the format. How do you intend to use it and how many “word lists” would you like to have?
I did choose those word lists for what they can encode. 5bits 6bits 5bits, which is 1 triplet.
These word lists are a good fit for unique words that may be rememberable as a tripple, and are easy to maintain. Also the size is important for me. 1 bit more will double the size of the word-list, which can make them hard to maintain.
The length of the words is short and matches some rules that fit the UUID usecase (only)
In your case you could probably have more word lists or a data tiddler that contains whole titles that can be split by space.
The main problem in TW are random numbers. Every refresh, when a tiddler is changed a random word list would be refreshed too.
So the real challenge will be to solve that problem. I do have some ideas here, but did never implement them. We would need a random number generator that allows “seeds” so they can be reliably replicated …
Interesting area or research, I wont pretend to yet understand the algorithms etc… being discussed but some thoughts arise from the concept of creating random titles or random word sets.
Decades ago a password suggestion tool would do this, generating a list of potential passwords, or more what we would call pass phrases. It would display the list and the user would choose the one they like and use it and the above list will not be displayed again. One can add steps to include more complexity like uppercase in random locations, numbers and even symbols to such phrases.
Having word lists that are held externally and can be retrieved on demand may make sense so as to allow minimal byes on a wiki could make sense even if it is another TiddlyWiki. If a query is made and a random title returned using SSL or some other method may be essential for security solutions, but not for titles. Perhaps a method like the “Random Dog” could be used?
Perhaps you could even use a resource like word of the day from a dictionary site, then pair/triplet it with another source of words.
If we are considering external resources of course it may make sense asking an LLM because in many ways it is a big keyword list but may even make a reasonable brave attempt to reflect the content in the tiddler.
We could play funny games by using common misspellings as the words may not be as important. eg vowel switches, “sound alike” constants eg c / k
I was also thinking one could use an external solution with access to large word lists and generate [[three word phrase]] list of links in a source tiddler. you could then source new titles from this list, only if the title is missing (ie not used yet) then once used ie has[title] field then consider it used if so.
I asked the following of ChatGPT;
Please consider the following words and return a set of three word summaries of the text even if it is insufficient to fully describe the text;
I got the following;
Here are concise three-word summaries capturing different aspects of your text:
- Password phrase generation
- User selects phrase
- One-time display list
- Memorability over randomness
- Incremental complexity options
- Uppercase random insertion
- Numbers symbols inclusion
- Early security tooling
- Passphrase usability focus
- Human-friendly passwords
- Regenerate excluded options
- Choice-driven authentication
If you’d like, I can bias these toward security terminology, UX design language, or TiddlyWiki-style tags/fields.
But you can also see the above is somewhat acting like an aid to brainstorming.
Q: How use it?
A: I will extract using regular expressions mainly nouns and adjectives from tale titles. For the purpose you don’t need so many.
Q: How many “word lists”?
A: I don’t know yet. I have to try two or three in the context of the other tools I’m using to find out whether a few or many might be needed.
TT
Verily a …
TT