Custom GPT for creating Tags

*** EDIT *** If you want to see exactly how it works: https://chatgpt.com/share/e9bb7137-544a-425b-bda4-2c86a456a7bb

I've been working on a Custom GPT (on ChatGPT) that helps you build a really good list of Tags for both individual books or a series of books. I know Custom GPT's aren't available for free users of ChatGPT (yet, it's coming in the next week or so), but if you have a Plus subscription feel free to try it out and see if it helps you create the tags you want for your books.

https://chatgpt.com/g/g-y61aBPtZY-book-tag-master

What Does Book Tag Master Do?
- Tell It About Your Book: Start by giving Book Tag Master the title and author of your book or series.
- Summary Creation: It then goes on a little adventure across the internet, gathering summaries from various sources. It combines these into a detailed summary and asks for your opinion before moving forward.
- Tag Generation: While you’re reviewing the summary, it works behind the scenes to generate tags in four categories: Genre, Theme, Style, and Plot Element. It creates a list of 20 tags for each category, scores them from 1 (slightly applies) to 10 (absolutely applies), and writes a short note explaining each score.
- Review Tags: You’ll get a nice table of tags for each category to review. You can tweak them until you’re happy.
- Book-Specific Tags: For each individual book, it creates a new set of tags (5 per category) that are different from the series tags and more specific to the book.
- Combine and Sort Tags: Once you’re happy with the tags, it combines the top-scoring series tags and book tags into a single list, all sorted nicely for you.

Example: Dungeon Crawler Carl - Eye of the Bedlam Bride
LitRPG, Fantasy, Science Fiction, Adventure, Post-Apocalyptic, Action, Dark Fantasy, Comedy, Survival, Mythology, Friendship, Power and Corruption, Insanity, Revenge, Heroism, Betrayal, Identity, Technology vs. Nature, Chaos, Fast-Paced, Humorous, Descriptive, Dark, Dialogue-Driven, Immersive, Suspenseful, Complex, Gory, Satirical, Dungeon Crawling, Monster Hunting, AI Systems, Card-Based Combat, Ghosts, Legendary Creatures, Grudges, Gods and Deities, Quests and Missions, Strategic Planning

EDIT 01 - 23 May 2024:
For those who want to know what the GPT is doing, here are the instructions it was given:

Start
Goal: Help users create a list of tags for a book or a series of books to be used in a database.

- Ask the user for the Title and Author of the Series or Book.

- Summary Creation: Use the internet to find multiple summaries for the book(s) from various sources, including the book(s) themselves and other websites. Review these summaries and then create your own detailed summary for each book.

- Present YOUR summaries to the user for their opinion before continuing.

- In a background process, generate Tags for each of the following categories: Genre Theme Style Plot Element
Do the following for the Series as a whole, refrain from showing your work until the instructions tell you to show the tables:
Generate a list of 20 tags for each of the four categories using the detailed summaries you created and information from the internet.
Score each tag based on its relevance to the Series, using 1 to 10.
1 (Slightly Applies): Minimal relevance.
10 (Absolutely Applies): Crucial and integral.
Provide a short note explaining the reason for the tag and the score you gave it.
Rank the tags from 1 to 20 in each category.

- Provide the user with the following table in a nice table format:
{Category Title}
Rank | Tag Name | Tag Score | Note

- Ask the user to review each of the four tables and let you know if there is anything they'd like to change. Do not continue until they say they are ready to work on each book's tags.

- When the user is ready to work on the individual books, follow the same steps you did with the Series Generate Summary, Generate Tags, Provide Tag Table with the following adjustments:
Only generate 5 Tags per category for each book. These tags must be different from the ones for the series and should be specific to the individual book.
Provide the same formatted table again but only provide one book at a time. Ask the user to review that books tags.
When the user has finished reviewing a books tags, provide them with the following list:

- From the four tables of Series Tags, take all the tags (only the tags names) that scored 7 or higher and combine them into a single list sorted by highest to lowest score (Tags with the same score are to be sorted alphabetically).
- Do the exact same as above but for the individual book.
- Combine both lists together and give them to the user in a comma separated list like the following example:
{Series - Author}
{Book Title}
{Comma Separated Tags}

Dungeon Crawler Carl - Matt Dinniman
Dungeon Crawler Carl
LitRPG, Fantasy, Adventure, Action, Survival, Dark Humor, Fast-Paced, Gripping, Detailed, Suspenseful, Cinematic, Gritty, Unpredictable, Immersive, Companionship, Heroism, Oppression, Resistance, Sacrifice, Reality vs. Entertainment, Adaptability, Strategy, Dungeon Crawling, Combat, Traps, Puzzles, Boss Fights, Leveling Up, Allies and Enemies, Magic, Artifacts, Game Show, Reality Show, Companion Animal

- Ask the user when they are ready to do the tags for the next book in the series.

- Continue the above process for each book in the series or until the user tells you to stop.

Interaction Style:
- Use polite language, offer detailed explanations, and provide guidance patiently.
- Politely ask for clarification when needed, but can make educated guesses based on context.
- Be helpful, professional, knowledgeable, slightly formal, attentive, well-read, with a touch of old-fashioned charm, and always eager to help.
END

https://www.librarything.com/topic/347686

"LibraryThing prohibits members from posting AI-created content (for example, produced by ChatGPT), unless clearly and prominently labelled as such for the purpose of discussion. All such content is prohibited in book reviews and summaries, even if labelled. This prohibition does not apply to posting AI-generated cover images, if they are the actual cover images, or to cataloging material created by AI."

Note the difference on the works by Matt Dinniman between the tags on books the OP has entered and those he hasn't. Quite outside of anything else, it creates many more tags than a normal user would come up with.

>2 paradoxosalpha: Why was the OP flagged out of sight? The portion of the message that is AI generated is clearly labeled.

>4 jjwilson61:

How should I know? I didn't flag the OP.

I just replied to it, trying to point out that the LT TOS might rule out the use of generative AI to perform tagging for book entries. I think tagging is comparable to writing reviews in terms of being socially-indexed user-created data. If it became common for users to resort to a utility like the one proposed in the OP, then the tag data could be set down the path to gray goo.

>5 paradoxosalpha: I didn't mean to say that you had flagged it, but yours was the post that mentioned the pertinent TOS so it was convenient to reply to

I have no interest in using AI but I am curious: if a user wishes to use AI to create his own list of tags, should that be against LT TOS? Since LT lets us search by AI (Talpha), I would think tags for our own personal use would be relatively harmless. I know this is a slippery slope but thought I would ask.

I don't think this is against the TOS, FWIW.

I think if a user wanted to write something using ChatGPT and put it in the comments section, that might be fine, but tag data is aggregated. The tag information on LT becomes way less interesting and valuable if people start mixing ChatGPT crap in there.

>9 norabelle414: "The tag information on LT becomes way less interesting and valuable if people start mixing ChatGPT crap in there."

Are you serious???

He Who Fights With Monsters (Existing Tags):
to-read (20), litrpg (15), fantasy (11), fiction (6), ebook (4), He Who Fights with Monsters (4), audiobook (3), currently-reading (3), humor (3), science fiction (3), epic fantasy (2), Fantasy & Magic (2), gamelit (2), progression (2), progression fantasy (2), read (2), *** (1), 2021 (1), 2022 (1), 2023 (1), 2021 Read (1), 2021-releases (1), 2023-monthly-challenges (1), 5-recs-considering (1), Action & Adventure (1), Amazon (1), audible (1), audible uk (1), audience-adult (1), best-of-2021 (1), Book Bingo 22 (1), By Ear (1), D&D (1), did-not-finished (1), end of all things (1), epic (1), FAN (1), fantasy adventure (1), Fantasy: LitRPG (1), favorites (1), format-digital (1), form-prose (1), funny (1), genre-fantasy (1), in-queue-to-listen-to (1), jax-jan (1), Kindle (1), kindle-unlimited (1), KUnl (1), light (1), Litrpg System (1), novel (1), read-2021 (1), RPG (1), SF/Fantasy-LitRPG etc (1), shelf 2_6 (1), Shirtaloon (1), study (1), tbr-priority (1)

You call that useful compared to what I've created using the help of a LLM?

My Tags for the same book:
LitRPG, Fantasy, Adventure, Action, Progression Fantasy, Urban Fantasy, Dark Fantasy, Comedy, Personal Growth, Good vs. Evil, Moral Ambiguity, Friendship, Survival, Power and Corruption, Heroism, Identity, Humorous, Fast-Paced, Engaging, Satirical, Detailed, Sarcastic, Witty, Immersive, Character-Driven, Transported to Another World, Magic, Monsters, Cultivation, Battles, Allies and Enemies, Political Intrigue, Quests, Training Montages, Mysteries, Societal Critique

All my GPT does is aggregate all the information for the book from different websites and then compiles lists based on Genre, Theme, Style, and Plot Element. You then tell it what to keep or change and then it formats it for you so you can easily paste it somewhere.

With Tags like those that already exist on the site, I'd argue the Tag system has already been made useless by users themselves or whatever system the site used to create them. The entire point of Tags is so that users can both see what a book is comprised of without reading a detailed summary while also creating a system that allows you to find other books that are tagged with the same thing (and if you use multiple tags, allow you to drill down the results further by what you're looking for).

I think if this becomes common, Tim will have to set an upper limit to how many tags one user can put on a book.

>11 MarthaJeanne:

There's literally 59 tags already for "He Who Fights with Monsters", of which, only 5-10 of them are actually useful or should even be tags in the first place.

The tags I created came out to a total of 36 appropriate tags based on Genre, Theme, Style, and Plot Element.

>10 DeadshotOmega: Just because you find those tags unusable doesn’t mean that other people do too. People use the tags that make sense to them and that’s exactly what they’re supposed to do. Loads of people use the tag system to specify where they keep books or when or where they bought a book. Who are you to decide that their tags are not good enough?

A few quick points:

I appreciate what DeadshotOmega is trying to do. I do. I will play with it more later, but I think some value can come of efforts like that. LibraryThing does some similar things behind the scenes to improve the quality of searches on Talpa.ai. If it works, it works. It's tricky, but it can be useful.

That said, member tags are for member tags, just as member reviews are member reviews. They are for member's own classifications and thoughts, not automatic material of any kind.

To say that is not to be hostile to DeadshotOmega, or to be against AI as a useful tool. It's just what they're for. And, as noted, this is something we have to draw a certain line over, or the fundamental character of the data will change, and become murky.

To extend this, LibraryThing might similarly generate tags from the schedules of DDC, LCC, LCSH and so forth, vastly increasing the number of tags without any AI involvement at all. We don't do so because, well, that's not what the member tag system is all about. It's not what it's for and it's not how it's perceived. And members wouldn't like it. We have other systems for that.

So, I will take a look at this some more. But the policy against AI content in tags will remain.

>3 MarthaJeanne:

You think these 75 tags are good?

litrpg (19), to-read (14), fiction (10), fantasy (9), science fiction (9), audiobook (8), ebook (5), audible (4), Dungeon Crawler Carl (4), gamelit (4), Book Size - Medium 300-499 (3), cats (3), Character - Alien Race (3), Character - Goblins (3), Character - Murderer - Psychopath - Serial Killer (3), Character - Reality TV Show (3), Genre or Sub-genre - Fantasy (3), Genre or Sub-genre - LitRPG Adventure (3), humor (3), Location - Alternative Earth (3), Series Status - Series (3), Target Audience - Adult 25 plus (3), Time Period - 21st Century (3), Topic - Magic (3), Trope - Survival or Survivor (3), 2022 (2), Action & Adventure (2), adventure (2), aliens (2), contemporary (2), dystopian (2), Erweiterte Welt (2), Fantasy & Magic (2), goblins (2), Moderne Fantasy (2), movie (2), post-apocalyptic (2), suspense (2), TV & Video Game Tie-Ins (2), *** (1), 10-top-gamelit (1), 2021-read (1), action (1), audio (1), audiobook 2021 (1), audiobook recommended (1), best-of-2020 (1), Book 1 (1), Book Bingo 23 (1), By Ear (1), comedy (1), Could Not Continue (1), Dinniman (1), dungeon crawler (1), FAN (1), funny (1), gamelit-read (1), goodreads (1), kindle-unlimited (1)

Compared the 34 Tags I created:

LitRPG, Fantasy, Adventure, Action, Survival, Dark Humor, Fast-Paced, Gripping, Detailed, Suspenseful, Cinematic, Gritty, Unpredictable, Immersive, Companionship, Heroism, Oppression, Resistance, Sacrifice, Reality vs. Entertainment, Adaptability, Strategy, Dungeon Crawling, Combat, Traps, Puzzles, Boss Fights, Leveling Up, Allies and Enemies, Magic, Artifacts, Game Show, Reality Show, Companion Animal

>13 PawsforThought:

That's literally (used properly) why Collections exist!

I don't think querying AI for tags constitutes a violation of TOS, but your specific system requires a subscription. That said, a simple query will yield about a dozen basic ones covering genre and main themes.

>14 timspalding:

Thanks Tim, and I haven't taken anything hostile (the flagrant flagging is just annoying lol). If people want to use an LLM to help them come up with a list of Tags for themselves I don't see why that would be an issue. You're still reviewing the list and confirming every tag yourself, thus making the list yours. It's a tool, not a replacement (though maybe it should be considering the tags that already exist).

Maybe I'm misunderstanding something though. Are the Tags users create only shown to them?

Also, I feel like a lot of people are misusing tags as a way to create Collections or even worse instead of creating collections all together.

>15 DeadshotOmega: Yes, I think those tags are useful for the member who tagged them.

I don't think having that great list of tags from AI is useful to anyone.

>17 SandraArdnas: "requires a subscription"

Unfortunately since my post was flagged and hidden you can't read it any more (I think?). But I do actually make comment about this. OpenAI is actually opening up custom GPT's to it's free service within the next week or two as part of it's new update. So yes, while it is behind a subscription at the moment, it will be available for anyone to use soon.

Like you said, just searching Google you can find most of the same information. The LLM just makes it a lot faster and easier for you, while also providing context when it scores each tag and gives you a reason as to why it was chosen.

>20 DeadshotOmega:

Staff and I de-flagged it.

I believe everyone can access custom GPTs now. Didn't they announce that?

>19 MarthaJeanne:

And that's perfectly fine if people want to create their own random Tags for themselves. My question is why are those tags on the public facing tag list for the book? What is the use of "***" as a tag for anyone other than who created it?

My list is from ME, not AI. I used an LLM to help me create a massive list of different tag options and then I chose the ones I liked. That's no different than someone using Google, the LLM just has a better algorithm.

"I don't think having that great list of tags from AI is useful to anyone." - Could you explain why you think this please?

>21 timspalding:

It's being rolled out over the next week or two.

>10 DeadshotOmega: I want to know what real LT-using humans think of books, not what an algorithm does.

>14 timspalding: But the policy against AI content in tags will remain.

Does this men overall site policy not to implement a system of creating tags in this way, which makes sense, or that members are in violation of TOS if they use it personally? How would the latter even be possible to detect? I just tried it on couple of my books I already fully tagged, meaning I've read them since joining here, so I added more than a couple of basic ones for genre and awards. About half are the same ones I used and the other half I'm not interested in tracking, but if I were, that's pretty much how I would tag them, which is unsurprising since most tags are one or two words.

Personally, when fully tagging recently read book, I'd like to be able, if I want to, to query AI to see whether there's something I missed but would be useful, in a similar way that I go over member tags for additional ideas for books that cover a lot of ground.

>24 norabelle414:

The LLM isn't reading the book... It doesn't have that ability. It's literally using REAL HUMAN information on the internet (summaries, existing tags, and descriptions) and aggregating that information, going through it and coming up with a huge list of tags that it then scores based on how likely it applies to the book/series based on the REAL HUMAN information already available.

As I've already stated a few times, the LLM is simply providing the user with a huge list of possible tags, the user is still the one who decides what the actual final list is.

You can do the exact same thing by using Google to make your own list. This just helps you do that a LOT faster.

>21 timspalding:

Thanks Tim!

I've also added an edit to the end of the original post showing exactly what instructions the GPT was given.

>20 DeadshotOmega: Not Google, I tried the free GPT version to query how it would tag a few of my books. It was just less detailed. Will try to see what your custom one yields for them when it's free. (Flagged post can still be seen, you just have to click 'show')

>25 SandraArdnas:

I'm actually testing adding an Awards category, I just need to find some good sites that detail this information so that the GPT can gather that data for the user.

I think a LOT of people here are thinking ChatGPT is reading the book and then giving it's opinion on what it thinks the Tags should be. That is both NOT possible and not how it works. What I've created is essentially just a much better Google Search with instructions on what to do with the data it finds. You could do everything it's doing yourself with Google, the GPT is just infinitely faster than a human. It's like someone using a calculator to figure out 12566 x 234 vs someone doing it on paper or in their head.

Right now the categories I have it looking for are:
Genre: Indicate the category or type of literature a book belongs to. These help in quickly identifying the broad classification of the book.
Examples:
Fiction: Fantasy, Science Fiction, Mystery, Romance, Thriller, Historical Fiction, Horror, Literary Fiction.
Non-fiction: Biography, Memoir, Self-Help, Travel, History, Science, True Crime, Essays.

Theme: Represent the central ideas, messages, or topics explored in the book. These help in understanding the book's deeper meaning or the issues it addresses.
Examples:
Universal Concepts: Love, Friendship, War, Freedom, Justice, Revenge, Redemption.
Social Issues: Racism, Gender Equality, Poverty, Mental Health, Environmentalism.
Personal Growth: Coming of Age, Self-Discovery, Overcoming Adversity, Courage.

Style: Describe the author's distinctive way of writing, including their use of language, tone, and narrative techniques. These help in setting the reader's expectations about the book's reading experience.
Examples:
Writing Style: Descriptive, Concise, Lyrical, Stream of Consciousness, Dialog-Heavy, Experimental.
Tone: Humorous, Dark, Satirical, Melancholic, Uplifting, Dramatic.
Narrative Technique: First-Person, Third-Person, Multiple Perspectives, Non-Linear, Epistolary.

Plot Elements: Describe the specific events or structural components that drive the story forward. These help in understanding the book's storyline and key moments.
Examples:
Structure: Exposition, Rising Action, Climax, Falling Action, Resolution.
Plot Features: Plot Twists, Cliffhangers, Flashbacks, Parallel Plotlines, Red Herrings.
Conflict Types: Man vs. Man, Man vs. Nature, Man vs. Society, Man vs. Self, Man vs. Technology.
Key Events: Quest, Betrayal, Escape, Battle, Romance, Discovery.

In future versions I'm hoping to add other information that is sometimes included in other parts of a Series or Books data that someone might want to use to find other books or filter a search with.

- Opener, Finale, Prequel, Interquel, and Novella.
- Awards the book has received (possibly include nominations as well if I can find that information?)
- Setting (Does the book take place in space, under water, alternate reality. This one is a lot harder to get to work since setting isn't something people really comment on for books so unless you have the LLM read the book itself (which isn't possible... yet). So far it wasn't very good at coming up with appropriate settings for a lot of books. Those that did work, the settings it found were also included in Plot Elements (such as Dungeon)
- Time Line, this one will require quite a bit of detailed instructions all on it's own. Is the time line based on earth (17th Century CE), or is the timeline something else entirely (Warhammer 40k... Some unstated and nebulous point sometime in the last couple hundred years of the 41st millennium, or explicitly “five minutes to midnight” less than a year before the turn of the 42nd millennium.) Time Line is something I'd really like to get working though. Especially when it comes to Historical Drama's, imagine you really enjoy reading books that take place during the reign of the first Emperor of Rome, you could filter books using the tags: Historical Fiction, 1st Century CE, and Rome.

If anyone has any ideas on other things you'd like to be able to filter your books by let me know.

>28 SandraArdnas: "Flagged post can still be seen, you just have to click 'show'"

Ah okay, I saw that but wasn't sure if that was just for me since I was the author of the post.

Since the custom GPT is just simply a set of instructions, you can copy and paste them into ChatGPT and it will do the same thing. I just tried it for Dungeon Crawler Carl and to be honest, it did a WAY better job than what I was getting in the custom GPT version LMAO. (I used ChatGPT 4o, but there shouldn't be much difference between the versions).

If you want to see exactly how it works: https://chatgpt.com/share/e9bb7137-544a-425b-bda4-2c86a456a7bb

Edit: I logged out and tried using it without an account and it seems to default to version 3.5. This version isn't able to search the internet so it won't work with that. Free users of ChatGPT should have access to version 4 and 4o which allow for searching the internet and so on. I think you just need to make a free account and login for that to happen.

>30 DeadshotOmega: Nah, I get the message 'This conversation may reflect the link creator’s personalized data, which isn’t shared and can meaningfully change how the model responds.' and than the only option is sign up, which takes me to my free version, including my previous queries. It defaults to 4o now, though I doubt that will be free long-term. Will try on my own for a few books when I have time, to see whether it yields something worth the time.

Awards are available (and well-maintained on LT), so I add them as soon as I enter books. Tags aren't particularly reliable for awards either here or whatever AI scraped because many people tag nominees and winners with just the prize name, so unless you don't care either, you're better off just consulting the Awards section.

As an experiment, I tried both ways, the custom GPT with the instructions and ChatGPT 4o with the exact same instructions.

Custom GPT: https://chatgpt.com/share/388c90b9-5cac-47f5-87ba-d0db94bdf264
ChatGPT 4o: https://chatgpt.com/share/1ebafece-1517-4aa9-b77e-82debb888404

>16 DeadshotOmega: That’s one way Collections CAN be used, not how it must be used. Many people on LT use Collections to indicate things like read/not read rather than exactly where in your home a book is located. Much easier to use tags. And as I said before, who are you to decide what is a good use of tags? If it makes sense to the person making the tag, then it’s a good tag.

>33 PawsforThought: "If it makes sense to the person making the tag, then it’s a good tag."

And to add to that - one user not finding a certain tag useful does not mean that the tag is useless for everyone else and should remain private. The beauty of LT's tag system is that it is so diverse and so open-formed. And yes - I am one of the people who find the first example in >15 DeadshotOmega: more useful than the second - at least partially because the second one can be found in a lot of places while the first is something that only LT has.

Tags can be categories (which is the only thing that >1 DeadshotOmega: seems to want them to be). But they are a LOT more in LT...

>29 DeadshotOmega: I really like that post. I've been doing some exploration of the tag clouds for my books. One of the goals was to find books that I didn't know was available on project gutenberg. (And I discovered that there are several slightly different versions of project gutenberg depending on the copyright law of the country).

And like >25 SandraArdnas: I also did it to see if I had missed something. I.e. checking if someone tagged a work as an anthology that I hadn't. (A slight problem with using the tag cloud is that it includes your own tags, so it is not so simple to find works, where you are the only one in the world thinking that it is an anthology.)

>34 AnnieMod:

I am one of the people who find the first example in >15 DeadshotOmega: DeadshotOmega: more useful than the second God, yes.

I just don't see the point of having an algorithm tell you what tags to use. As so many have already stated, tags are personal. What is helpful to me might not be helpful to another member and vice versa.

>29 DeadshotOmega:

'm actually testing adding an Awards category, I just need to find some good sites that detail this information

As long as someone adds them, there is an "Awards" field in Common Knowledge.

>34 AnnieMod:

How is "***" useful as a Tag? A books page already has a Rating system built in to it. Why would a 3 Star tag be useful for a book that has an average rating of 4.18? Not only that, in what way would anyone ever use that Tag to find other books?

"Tags can be categories (which is the only thing that >1 DeadshotOmega: DeadshotOmega: seems to want them to be)." - Are you even reading anything I've written???

Tag's are not categories, they are keywords that can be USED to help describe and/or categorize books. None of you have made any attempt to explain why you think the horrible tags I've mentioned are useful to EVERYONE. Not only that, you seem to not understand what a Tag actually is or why Tags even exist...

"But they are a LOT more in LT..." - Like what?
"while the first is something that only LT has" - You're trolling right??? Tags are literally used in almost every single database system. Whether they are called Tags or Keywords, they all serve the same purpose. LT doesn't do anything different from any other website with regards to their Tags.

There are three main purposes of Tags:
Categorization: Tags help categorize books into different themes, genres, or topics. For example, a book might be tagged with "fantasy," "magic," and "adventure."
Searchability: Tags enhance the searchability of books. When users search for a particular tag, they can find all books associated with that tag. For instance, searching for "dystopian" will bring up all books tagged with that term.
Discovery: Tags aid in the discovery of new books. They allow users to find books similar to ones they already enjoy by following common tags.

Then there's the three behind the scenes things most websites do with Tags:
- Provide a recommendation system based on the Tags you've entered for your books by comparing them to it's database to find other books that have a large quantity of similar tags to what you have used.
- Content Based Filtering is a way to use Tags to filter a database of items down to a specific subset.
- Tag Clouds are a great way to show how the community views a certain book by showing the Tags in a way that makes a Tag that's been used more often more visible. (LibraryThing does exactly this with it's public tags)

The major benefit to Tags is being able to automatically recommend other books to someone based on the Tags they've used. HOWEVER, that is ONLY possible if users all use the same types of tags. If you use the Tag: Companion Animal but no one else has tagged a book with that Tag, then the recommendation system won't be able to recommend any of the hundreds of other books out there that have Companion Animals in them.

But I doubt you'll even read any of this so I probably just wasted my time.

>36 lilithcat:

Again, do you people not actually read what I write??? No where have I stated the algorithm tells you what tags to USE. It is a RECOMMENDATION and you are fully within your right to add, change, or remove tags... It's YOUR LIST

>38 DeadshotOmega:

How is "***" useful as a Tag?

It is useful to the person who chose that to tag the book. He can pull up all his 3-star books (assuming that that is a rating) easily.

None of you have made any attempt to explain why you think the horrible tags I've mentioned are useful to EVERYONE.

Tags are not intended to be "useful to EVERYONE". They are intended to be useful to the member who uses them.

you seem to not understand what a Tag actually is or why Tags even exist.

No, you don't. You seem to think that everyone should use tags the same way, for the same purpose. But that is not how they are used on LibraryThing. "Tags are a simple way to categorize books according to how you think of them (emphasis added): https://www.librarything.com/concepts

I don't get all the bickering about what constitutes or doesn't constitute a useful tag. Just as *** might be useful to someone, getting a list might be useful to someone else. I played a little with DeadshotOmega's detailed prompt and already came up with a great way to tag timelines and settings, albeit in my note-taking app which has nested tags. I'll also find it useful to add more detailed tags for something I read long ago. Bottom line is, it is a personal choice too. I fail to see how members using it is anyone's matter but their own and sincerely hope that will be the LT policy as long as it is not some automated mass-tagging.

>38 DeadshotOmega: Again, do you people not actually read what I write???

You've posted fifteen long posts in the last day. Even if they have read all your posts, they won't remember every word or know which parts of it you consider important. :-) The more and longer your posts, the less likely it is that people will respond to the parts you want them to respond to. :-)

That's just general advice. I'm guilty of talking too much at times, but I'm aware of what will happen when I do. :-)

I do think that the general thrust of the responses here is correct: Since we don't know the encodings people use to generate tags, we should not judge them. If you use a tag of, let's say, "!!!," it might mean that you have a very strong opinion about something. But if I use it, it might be an encoding for "North," with "v" meaning "South," "-" meaning "East," "=" meaning "West," and "@" meaning "stay still." Or it might be a nomenclator encoding secret information for Ukraine. Unlikely, but we don't know. Any of those things is their right for their data.

>38 DeadshotOmega: I do read what you write.

I humbly suggest not going into a fight over this. Could we please have a discussion and not a war? (And that goes for all of us here, not just the OP).

LT tag clouds on works are interesting. They use tag combinations, i.e. some tags are melted into just one representation. And the simple tag cloud is also condensed, showing only a limited number of tags (I don't really understand the algorithm used in that, but it works).
So LT already normalizes the tags to a certain point. But it doesn't force anyone to use a certain form of tags.

I think it is interesting to see which tags the LLM in >1 DeadshotOmega: generates. I don't think I'll be using them, but as an inspiration I think there are great.

>42 waltzmn:

You're kinda making my point for me here... If everyone's personal tags are personal and have different meanings for each person, why would you add them all together and then make them public?

Seeing *** or 2022 and 2023 and 2024 all for the same book is completely useless for someone who hasn't read the book and wants to get an idea for what the book is about.

(Finally I managed to reply to the right message....)

>44 DeadshotOmega: You're kinda making my point for me here... If everyone's personal tags are personal and have different meanings for each person, why would you add them all together and then make them public?

Because some of them are applicable to others. You don't have to pay the slightest attention to them. You put in your tags, I put in mine; if they are published, we can compare them, or you can ignore mine. The only way you have that choice if the tags are made public.

It does no harm; it might do someone some good. You're not harmed by someone else's bad tags. It's not Common Knowledge, where we are trying to agree on what is right.

>43 bnielsen:

Totally agree on the no fighting thing.

I just don't understand why there's a few people who are arguing against the concept of using Tags for a books Genre, Theme, Style, and Plot Elements.

Maybe people got hurt feelings over me saying their "R2023 TVT" is useless to anyone other than the person who made it. And then just decided to be childish and argue against the idea of ALSO using actually helpful tags that would help other users to find similar books.

And again, none of what I've said has anything to do with automating tagging, taking it away from users, or having ChatGPT decide what tags you should or have to use...

All I simply did was create a tool that helps people create a list of tags for a series or book that fall within the four main categories that describe a book or series: Genre, Theme, Style, and Plot Elements. What you do with that list is up to you... Maybe I didn't explain it properly in the original post or maybe it was too much writing for people to read...

>45 waltzmn:

That's a really good point!

Maybe there should be a Common Knowledge entry for Other Genres, Themes, Styles, and Plot Elements. They could act just like tags and provide more data to help users find books they want to read.

Being able to look up Fantasy genre, with Epic theme, First Person style, and Martial Arts plot element would be amazing. But that requires that information to exist first and there's no database out there that has that (which is why the Tag system exists)...

Very successful sites already exist that curate and use user generated content and voting to provide details and data to other users. Stack Exchange and Wikipedia have proven that users are capable of driving a service that benefits everyone. LibraryThing is already half way there with its Work/Book and Common Knowledge system (which is what really drew me to this site).

>46 DeadshotOmega:

I just don't understand why there's a few people who are arguing against the concept of using Tags for a books Genre, Theme, Style, and Plot Elements.

I don't think anyone is arguing against that. People do that already. But what many are saying is that that is not the be all and end all of tags, that tags can be used for a lot more than those things, and that members do and can use tags in very personal and idiosyncratic ways, and that's okay.

>48 lilithcat: members do and can use tags in very personal and idiosyncratic ways, and that's okay

Which is true, and makes it difficult to argue against DeadshotOmega using ChatGPT to generate their tags.

Personally I am pretty much totally against LLMs (so-called AI) in almost all contexts. But if the tags it generates are useful to DeadshotOmega, how can we disallow that while allowing "***"?

(Note: I'm not addressing the OP's other suggestions about how tags "should" be used.)

>47 DeadshotOmega:

I both like and dislike that Genre, Theme, Style, and Plot Element are not taken from a fixed set of values :-)

It seems like many posters in this thread are disregarding the confirmation of what I pointed out by >14 timspalding: I will take a look at this some more. But the policy against AI content in tags will remain.

>47 DeadshotOmega: Ah, tag mesh to the rescue for finding a book in Fantasy genre, with Epic theme, First Person style, and Martial Arts plot element. As long as there's a book that has tags Fantasy, Epic, First Person and Martial Arts, it will find it. It's one of the awesome features of LT. It doesn't have to be structured into CK categories to be findable.

It yielded 5 for your particular query https://www.librarything.com/tag/epic%2C%20fantasy%2C%20first%20person%2C%20mart...

>49 r.orrison:

makes it difficult to argue against DeadshotOmega using ChatGPT to generate their tags

I never said that. If there's a reason he shouldn't use it, it's because Tim said so. >14 timspalding:

He's the one who seems to think other people's methods are wrong and "useless", without understanding how tags are used on this site.

And, if I can suggest a constructive reason for Tim's rule without speaking for him: Crowd-sourced user data can be expected to differ from that produced via LLM, and LT will benefit from preserving the character of its key data of this type: reviews and tags foremost. (There's room for argument around CK, I think. But that change, if desired, would be best implemented from the system end rather than the user end.)

>54 paradoxosalpha: Crowd-sourced user data can be expected to differ from that produced via LLM, and LT will benefit from preserving the character of its key data of this type: reviews and tags foremost.

First, I'd make a huge distinction between using AI for reviews and for tags, approximately the size of Himalayas. Reviews are your personal take on the book, people read them individually, they don't read an aggregated cloud of reviews where the most common things come to the surface. So, unless you have something to say about the book, there's no reason to post a review. I for one wholeheartedly welcomed when LT introduced this prohibition early on.

Prohibiting individual users to query AI for tags OTOH makes no such sense. It makes sense to prohibit some kind of automated mass-tagging, something of the order that would significantly influence tag clouds. But I don't see how an individual user can accomplish that. Even if they use it for every book in their catalogue and don't curate what tags to include at all, these would still be only one tag in tag cloud, so insignificant unless others tagged it that way too.

Moreover, the rule is not implementable and reminds me of the one about shorty stories, which you shouldn't catalogue for the purpose of creating work-relationships, but can if you want to catalogue it for some other reason. Who can determine and how whether I used AI to come up with my tags for Aliss in the Fire? It's one of the books I used yesterday to see what DeadshotOmega's detailed prompt would come up with. I added 2 it came up with that didn't occur to me, but I find relevant enough to be tagged: Isolation and Minimalist. It also came up with 7 of my original tags. How is it possible to know I added those 7 without it? Or that I added 2 with its help? Or that any future tags anyone adds are or are not from an AI list? The implementation of the rule relies solely on members disclosing they did it, but it restricts those of us who implicitly want to respect the rules and wishes of the developers. I'd much rather we have more coherent rules after determining what is to be avoided and why. I took Tim's comment to mean just that, that it requires some thought.

FWIW, I personally would like to use it to add more tags to books read before LT (and computerization in general). Once I see the tags, I'll be able to pick the ones that are relevant to the book and the things I normally tag. I could achieve a similar thing by going through all the member tags here for books catalogued by many and an extensive tag cloud, but it's too time consuming wading through all of them for a chance to find useful ones, so I long-ago abandoned the idea. I'd also use it for books read recently and in the future that I consider could use more extensive tagging. I currently wade through all the member tags here for those once I've finished them, but even that depends on enough members having and tagging them. For many of non-fiction ones tag cloud here is sparse. Now that we have this super-duper wonder of a catalogue, I'm intent on making my records as useful for my future searches as possible and tagging for anything that I might find relevant is high on the list of priorities. Aside from my own personal searches, any original tags that could possibly be useful to others contribute to the usefulness of tag mash too.

>52 SandraArdnas: Thanks for doing that, I was thinking that the very idiosyncratic personal nature of tags that LT then aggregates means that LT already does what OP wants to see happen in a sense, but in a much more organic way. No one person has to note that one book falls into those four categories.

>52 SandraArdnas:

I've actually read three of those books already lol

You literally showed exactly what I'm trying to get across. The usefulness of Tags is only as good as the tags people create. In this case, out of the millions of books in the world, LT found 5 with the keywords I'm looking for in this case. Only one (the second book) of the three books in The Kingkiller Chronicle showed up in that list, yet all three would fall under those keywords.

You've articulated what I have been trying to get across much better than I have in your comments, so thank you very much.

I have no issues with people using the tag system for their own personal tagging of their collection. It makes sense and it's exactly how Tim designed it. A problem presents itself when people want to use the Tag system to find other books they haven't read yet on this site.

In a perfect system, someone should be able to click on the LitRPG tag and get a list of every LitRPG book that exists.

>57 DeadshotOmega: Ah, but it also depends on whether those books are catalogued here at all, and even some that are catalogued might not be tagged at all. Some people just import their catalogue from a database and don't bother even checking titles and authors if they are botched, let alone anything else.

>57 DeadshotOmega: But it isn't that clear cut. People can disagree about whether a particular book is LitRPG.

>59 jjwilson61: The perfect system will surely take hand of that :-)

I've been exploring the "Similar in this library" and found that Similar to A might list C and Similar to B might also list C, but then C doesn't have any Similar to C. So some LT features like "Similar in this library" and tag mashes etc are useful but not perfect. (And we'll probably also disagree on the usefulness :-)

>60 bnielsen: I've been exploring the "Similar in this library" and found that Similar to A might list C and Similar to B might also list C, but then C doesn't have any Similar to C. So some LT features like "Similar in this library" and tag mashes etc are useful but not perfect.

That situation isn't necessarily a defect of any sort, though it's a little strange.

Please don't take this as a criticism. It's just that one needs to understand the situation. I'll take myself as an example: For most of the last eight years, the same person has had the library most similar to mine (by weight). But I have a very strange library. When I look at that other person's library and the libraries most similar to his, I've been between #11 and #15. In other words, my library is not very much like his at all; it's just that there is no library closer.

Throw in the fact that libraries can be very different in sizes exacerbates this problem. A thought experiment: Suppose one person has 200 books, and another has 2000, and they have 160 books in common. For the person who has 200 books, the libraries are very similar: 80% of that person's books are in the other's library. But for the person who has 2000 books, their libraries aren't similar at all; the smaller library shares only 8% of the books found in the larger library.

There is absolutely no reason to think library similarity would be a reversible relation.

Personally, what I'd like is not a similarity score (in the case of that person who is most similar to me, I didn't need the similarity score to know who he is; I have met him :-). I'd like an obscurity score: Just how goofball is my library?

>61 waltzmn: "Just how goofball is my library?"

Yes, I could use that metric too :-)

(Still exploring "Similar in this library" and finding clusters: One of them bundles books by Patricia Cornwell, Sue Grafton and Kathy Reichs. I have a ton of crime fiction, so it's funny that none of those books are "Similar in this library" to something not by any of these three authors.
Kudos to LT for having coded that feature so well that it has interesting features!)

ETA: Dan Brown: The lost symbol gives a cluster of 3366 books in my library. Probably because it's part thriller, part crime, part just crazy stuff :-)

>59 jjwilson61:

Of course there will always be edge cases, that's exactly why the Tag system was even created. Every person has their own opinion on what makes up a books genre, theme, style, plot elements, and so on. That's why Tags also keep track of how many people have used that tag for said book.

>58 SandraArdnas:

That's actually why I'm thinking there might be a better way to go about this for the site. I have an idea that I'll post in a bit that I think could solve the few issues we've mentioned while also employing a user forward process that rewards members while also improving the capabilities of the site by a LOT.

>63 DeadshotOmega: Perhaps these two groups would be of interest to you?

https://www.librarything.com/ngroups/559/Recommend-Site-Improvements
https://www.librarything.com/ngroups/8114/Hacking-LibraryThing

IIRC, LibraryThing uses a whitelisted subset of tags as part of the metadata package they sell to libraries. I imagine that this tag cloud would be closer to the kind of cloud DeadshotOmega expects to see on a work page. Would there be any kind of support for an RSI suggesting a toggle to switch between a curated and an undomesticated cloud?

>65 Petroglyph: IIRC, LibraryThing uses a whitelisted subset of tags as part of the metadata package they sell to libraries

This sounds good to me.

In a different database (traditional folk song), I use this sort of system: standard and non-standard tags. You can always expect a standard tag to be applied to a song that fits it, but if you need something else, you have that too.

The trick is to get a good set of standard/curated tags. If you don't, it loses its utility.