I know little about generating a novel... or writing one... I'm pretty sure it involves words... hopefully not spelling because I'm bad at that. I fully expect to create the greatest work of cliched nonsense ever created within a month, with zero experience and little free time.

This is an open issue where you can comment and add resources that might come in handy for NaNoGenMo.

There are already a ton of resources on the old resources threads for the 2013 edition, the 2014 edition, the 2015 edition, and the 2016 edition.

... [N]ew directors often work with low-budget horror movies. Witness Sam Raimi - he did Evil Dead not for any particular love of the genre, but for most-likely return on investment (time+money) (a source). The audience tends to eat it up no matter how low the quality. Witness the large numbers of self-published zombie books on Amazon. Or romance novels of any of the vast, arcane genotypes of romance. ---Micheal Paulukonis, Nov. 4th, 2015, GitHub comment

Generally, as the corpus grows larger, the generated output becomes more varied and interesting. This means that the best way to improve a text generator is to simply gather more data. Of course, the process of gathering data is itself an expensive endeavour, forcing programmers to be more creative in how they use the corpus they already have.---Tariq Ali, Dec. 23rd 2016, "NaNoGenMo 2016 and My Predictions About Text Generation

In the past, I have attempted to handwrite several science-fiction stories (most of them being glorified Terminator fanfic). Unfortunately, they are all half-finished. Fortunately, I still own copyright to them. Therefore, I plan on combining all my stories under an overarching "pulp sci-fi" storyline about Skynet. If Michael is right, referring to such an iconic pulp icon would immediately attract attention to the generated novel. In addition, since I had handwritten most of the corpus, I don't need to worry about the text being seen as "too ancient". I may have to worry about the coherence of the text though.

I do not have 50,000 words of corpus lying around for me to reuse to my heart's content. So what I plan on doing is to extend the corpus by adding █ to the text I already have. The fluff justification here is that you're reading censored data about the world that Skynet had gathered. I'm not sure how much █ I can get away with using though, since having entire pages filled with █ will be boring to consume. Still, it's worth a shot.

EDIT: While "probable obsolesce and extinction at the hands of a potentially super-intelligent supercomputer" can be interpreted as a type of psychological horror story, it may be better to take Michael's words literally and ramp up the "pulp" nature of this sci-fi novel. So in addition to fighting Resistance agents, Skynet will also battle against aliens and zombies.

Recently, I've been thinking about bouldering a lot, and I'd like to generate a book about how it feels to be finding your way up the wall.

First, I got my foot up on a small blue hold. Then I had to dyno up to a green jug hold with my right hand. My left hand grabbed a white sloper above my head. ... ... ...

I was inspired by #14 and dariusk/NaNoGenMo-2014#76, because I love these stories that are sort-of-monotonous-but-also-moving-forward. If I succeed in my novel generation, I may try to use the generated novel as warmup in my next workout. (Not all 50,000 words, probably.)

Me and @spurll are going to try to generate our own monomyth in the vein of Hero with a Thousand Faces by Joseph Campbell. We're hoping to generate our monomyth arc, our own myths and legends, and our own surrounding Freudian psychoanalysis. This is definitely biting off more than we can chew

I'm planning to write something to generate something very much like a narrative, using the Harvard sentences as input. Because these sentences are intended to test speech-based systems rather than text, there will be a generative audio component as well, resulting in an audiobook version of the generated novel.

Code will be available on Github and I'll provide a writeup upon completion.

Instead of distracting myself from work by doing a silly NaNoGenMo project, this year I want to aid my work with the silly NaNoGenMo project I'll attempt.

I'm working on verb frames, so I hope to find out who does what to whom... like... in general. Turning my initial findings into a 50,000 word book should be a piece of cake/fun impossible mess.

In another installment of things that Doug wants to do, but we'll see if they actually happen, today's idea (I think I'm going to submit three) is to work with air traffic control (ATC) conversations to create a novel from the perspective of one ATC tower along a fictitious or real airline route. The thing that fascinated me most about flight simulators was always the ways in which pilots talk to each other and to ground. Whenever I used to travel frequently, I'd plug in my headphones and listen to the squawk between parties. Somehow, it relaxed me to fall asleep and wake up to the various hand-offs and brief conversations about weather.

This has a kind of literary precedent; I'm thinking of Seascape, by Heimrad Bäcker, which is one of the most riveting data-as-book works I've ever read. When crisis appears, it seems benign, but it's really troubling. I'll save the specifics in case anyone ever wants to read it. The physical book object is a fantastic work as well. Plus, there's a great essay by Charles Bernstein in it where he asserts that it can't possibly be a narrative or poetry, but I digress.

I think I'm just trying to find ways to use Markov chain/LNN/RNN skeletons that I have lying around. That seems to be one of the solutions here. However, I would like to make each page a daybook-as-map in which various planes and their messages are spatially located on the page to represent the origination/position of a given message. That seems like the complicating piece here that interests me the most. The unresolved question that I just haven't yet researched is where to find either current or historic logs of these conversations. I may also rely on flight simulators to get stems of what these conversations look like.

Another thing I thought of while writing the above is the complication that I'll need to store what other planes are doing. This makes me think of "Talk Town". So that'll be another wrinkle I need to smooth out for which I have some guide.

The title comes from the last message usually transmitted by ground to a plane, ending communication.

Let's see how crazy I can drive myself this November.

So. I have another idea that's 3 months early. I'm probably not going to end up being able to finish both this submission and my other one in only November, so I may end up being a little cheeky and working on them separately before November.

To keep in the spirit of NaNoGenMo, I do want to keep the development time to a month.

This is another simulation based approach that I have in mind. To sum it up briefly, a narrated game of Illuminopoly.

Illuminopoly is a set of rules that modify Monopoly, changing the narrative of Monopoly to one of conspiracies trying to run the world through mind control systems installed in real estate. Now, I'm no lawyer, but as far as I know, despite the rules being trademarked, the mechanics are not trademarked, and I should be in the clear legally speaking.

I'm going to have a set of AIs playing Illuminopoly, narrate it, and frame the story as a narration of competing factions vying for control of a specific city. I'm going with London, because why not.

If this makes it to 50k words, great! If not, I'll either insert more boilerplate into the system, or just run it multiple times.

I plan on using Project Gutenberg, as well as a Markov library and Python, to create a novel based on as many public domain novels as I can find.

I may also throw in some XKCD transcripts for flavor.

The plan: take some existing input as a stream, isolate the nouns, take their hypernyms and choose a particular hyponym at random of that hypernym, then memoize that choice so that all later substitutions are consistent.

Lots of people have used wordnet's synonyms, hypernyms, and hyponyms to mutate existing texts, but I don't recall prior entries doing exactly this.

The idea I'm working with is to create an engine for a character to go on an adventure, with short diary entries logging the adventure they're going on. I'd like to try to implement a system like the roguelike Unexplored does with dungeon generation with looping paths, pre-seeding important items, etc.

My plan would be to pre-generate a 50k word text for NaNoGenMo, but also have an ongoing twitter bot based off of the same engine and possibly corpus.

Some kind of new Sherlock Holmes story by stitching pieces from the original stories

The goal here is to create a novel that follows a party of D&D characters. The characters, objects, and settings are generated based on character and item creation in D&D. Events in the story will come in 3 types: dialogue between party members, dialogue between the party and NPCs, and combat encounters with randomly generated monsters pulled from the D&D bestiary.
The story will consist solely of those encounters until the party reaches some end location and a final encounter occurs.
Simulations of the dialogue and combat will be used to create the actual sentences and paragraphs in each chapter.

Working together with @benmcmorran.
Here is where the code will be.

Hi everyone, I'm going to try to make something this year! I haven't planned out anything yet but it will probably have to do with word embeddings somehow. Also the WikiPlots corpus.

I dunno. I've written a sentence generation program already, but it's nonsense. Hopefully planning to incorporate corpora and make it sliiightly less nonsense. (i.e. get words from each chapter of a novel using POS tagger and then creating mini-corpora... I don't even know how to explain this, but something.) I'd like to do something more complicated, but I doubt I'll have time. We'll see how it goes.

Going to maybe keep a record of my progress here, if other participants don't mind daily edit-updates?

Daily Updates
11/1: Downloaded CoreNLP. Figured out mostly how to call CoreNLP from Python. Wednesdays are busy.
11/2: Wow I suck at Python. Modules are hard. Even Stack Overflow holds no hope for me. :|

A couple years ago I did a demo that generated an alphabet & vocabulary for asemic writing: https://github.com/enkiv2/misc/blob/master/asemic.py

I've modified it to use PIL instead of turtle, use markov chains (rather than picking random words), and write to an A1-sized page image.

Code: https://github.com/enkiv2/misc/blob/master/nanogenmo-2017/asemic.py
50k 'book': https://github.com/enkiv2/misc/blob/master/nanogenmo-2017/asemic.pdf

I don’t really have a plan for how to scaffold my generator’s output, but I’m going to use Node to generate it! I have some experience generating sentences, so I’ll be pulling from that background for this.

I think what I’m going to do is set up a database of words to pull from and figure out how to generate characters and story arcs procedurally.

Should be fun!

Here's my repo where I'm working: https://github.com/Alamantus/NaNoGenMo2017

The intent is to produce a book that describes a computer's thought process as it makes one chess move.

The book opens with a chess position, followed by an in-depth analysis of the board and potential follow-up moves. The whole thing is presented of a player narrating to themselves, as in "Harry considered moving his bishop from a3 to c5. If he did that, Maude might respond by moving her pawn from f4 to f3. Then..."

At the end of the story, the player will announce their move.

Think of it as a verbose version of the alpha-beta or minimax algorithm. Prior work would include other noisy algorithm output, like this one:
dariusk/NaNoGenMo-2014#76

Still brainstorming theme ideas/more specifics, but I'd like to participate this year and I'd like to create a generator which generates my novel deterministically. Thus distributing the source code and a seed will be equivalent to distributing the source code and the generated novel.

Development will not be happening on GitHub: I use Fossil so the majority of my notes, etc. will be there. However, once I have a title I will update this issue with a link to the Fossil repository and a new title. And of course will do a post-November write-up at the end.

I'm going to attempt to generate a play using the speeches of President Obama as source material for an RNN. Other characters' lines will likely also come from speech transcripts, although I may use other playwrights works too.

I work in web operations at a [LARGE GOVERNMENT ACCOUNTABILITY OFFICE]. Often, what we do revolves around companies and contractors appealing contracting decisions made by various arms of the federal government. Typically, these are service or manufacturing contracts whose arbitration falls to the teams of lawyers here. This seemed somewhat benign and mundane to me until I thought about what I was going to do this NaNoGenMo.

Science fiction loves the idea of the "corporation." Big, bad, evil corporations exist in most sci-fi worlds. While their degree of control over economies and daily life varies from world-to-world, I got to thinking about ways to use corporate constructs and the form of the legal arbitration documents I process (they're public documents) as a kind of storytelling. Similar in spirit to the Marvel comic, Wrecking Crew, I am injecting a little joy into the hapless legalese that I uncomprehendingly stare at every day, the storyline of what happens behind the scenes in a "Freaknomics"-ish kind-of way.

"Blade Runner" takes place in November of 2019. I seem to like doing things that revolve around the experiential time period of NaNoGenMo, so one possibility I'm considering is acting from a world government perspective, creating a book of various bid protest successes and failures of corporations in that world. I could also take it to a fictitious place and create dummy corps. (probably what I'll do), but the genesis of the project lies in thinking about mapping the "Blade Runner" universe onto this one -- one in which Tyrell Corp. and others bid (sometimes successfully, sometimes not) for government contracts to manufacture or service "future stuff."

These would take the form of the decisions that some faceless "accountability office" issues, using the form and text from real-world documents that have been processed during the month of November for as far back as I can go. Since these documents are public (some redacted, others not) the corpus is accessible and usable (even if I need to do a little scraping or batch document processing). This seems ideal for Markov chaining or neural networking, but the final method is somewhat undecided at this time.

I'll post my various updates here, though I'll probably be submitting other projects.

I've wanted to try NaNoGenMo for a couple years but this time, I'm going to work really hard to build up a buffer so I can actually focus on it for a month. My current idea is Girl Kills the Dragon. It will probably be written in C# because that's the language I'm most comfortable with.

Actors

The basic element of the program will be an Actor, basically a character. Actors have various aspects (I'm planning on using Fate RPG for a framework), motivations, and relationships with other characters. All actors are going to start in a neutral state.

Actors will have names and be managed by a ActorManager which is universal for the entire story even across rewrites (later).

Plots

An actor is changed by acyclic directed graph of Plot objects. These take in one or more actors, make changes, and then send them along their merry way. Since a plot is something happening, basically the novel is going to be two more plots rendered out until it makes 50k words.

Plots are going have a type based on Dwight V. Swain's Scene/Sequel. Basically, the system is going to alternate between those types (more later).

The basic plot item is the DefineActorPlot which basically creates a new actor, give them personality, attributes, and the like. There might be back story plots (ProloguePlot) to build up that character and give references (basically a character generator but those won't be written in the novel, but may be referenced). This will be something like "Mary is a girl" and basically establishes the where, what, when they begin the story. The items will also create Location objects which will eventually have a relationship with each other.

The end point is a ClimaxPlot, basically the end point. This is the "Mary kills The Dragon." There might be some plot items to tie up loose ends (EpiloguePlot). Both climaxes and defines are going to be Scene plots as opposed to Sequel plots.

The rest of the plot objects basically have a set of initial requirements (including actors), perform various operations on the actors, and then connect to another plot. A good example is the TravelPlot which basically takes one or more actors and moves them to a new location (controlled by a LocationManager).

The Novel

The initial graph is simply two DefineActorPlot (good guy and the bad guy), one travel plot(good guy goes to bad guy's location as a sequel), one bad guy does something as a sequel, and one ClimaxPlot (good guy kills bad guy).

Every plot will eventually have a time that it happens, so they can be sorted in temporal order. This makes it easy to actually write out the novel: grab all the plots, sort them chronologically, and then write them out. The version I'm aiming for this year is basically the summary but eventually each plot could be expanded into a proper prose and dialog.

Writing

The basic 5-plot chain is rather boring. So, I need to add more complexity to the story. This will effectively be an infinite loop that breaks out with certain conditions. The initial was is going to be a minimum number of non-ProloguePlot in the chain since prologues won't be written. Say, 15k plot points for a long rambling story. If I can expand plots into actual narrative, that number would go down sharply since I'll have 500 word plots instead of 5 word plots.

For every iteration, a new plot is inserted between two adjacent Plot objects. Which plot is going to be "random" (there will be a seed to produce consistent results) but there will be some rules. First, I'm going to resolve all the characters in the preceding plot (since they start neutral and have operations by the plots they are going through, I can figure out the state at any point consistently). Then, I will find all the known plot objects that have their requirements fulfilled by the preceding plot, then can fulfill the succeeding plot after the operations are done. Once I get that list, I pick a random one and insert it into the chain.

Once I do that, I resolve the scene/sequel by picking a plot to make sure it goes A/B/A/B. I also play through the rest of the plots to make sure each of those is still valid. If they aren't, then I either correct them with more plots or remove them (the rewrite phase).

Example

Using a starting example:

1. There is a actor1, they are at location1. (Scene to 2)
2. actor1 does something bad to actor2. (Sequel to 5)
3. There is a actor2, they are at location2. (Scene to 4)
4. The actor2 goes to location1. (Sequel to 5)
5. The actor2 kills actor1

Say we have a random point that introduces a new character and insert it between 4 and 5.

• There is actor3, they are at location2, they join actor2.

Now we have three characters involved in the story. However, we have two scenes together (the new 5 and 6), so we find a sequel to put between them.

• actor3 talks about backstory with actor2 (sequel)

Other Notes

That's the basic idea. I figured the complexity comes from the plot items. The above example has an ActorSet which is traveling together so actor2 and actor3 can attack actor1. I can have scenes where characters build up relationships with each other which would then let me have secondary plots. I can also backfill introduced characters so I could add some plots for actor3 to get them into the story.

Romance plots might have a "split the party for one-on-one time", "some event", "rejoin next day". This probably means I would be randomly selecting PlotFactory instead of Plot directly for the insert operations.

Plot objects will have a time to happen and time after them which would also help me figure out where to insert new plots. Also, if everything is relative, I can reference future times (you must be at the castle in {3 days}) which will change based on the intervening plot so they are always arriving at the last second.

1 part drink mixing book and 2 parts journal.

I'm sorry I'll let the generator do the rest of the puns.

Since I'm not the same person as I was before the eclipse, I might try some new things, but I still want to get "Beatles Bible" working converting song lyrics to a chapter & verse reference (maybe update to py3, maybe buy into a better lyrics API)

If I can get my Glitchy3bitDither project updated in time, I want to try to use that to generate a comic-book-style AESTHETIC look book.

Maybe something with tracery? I need to improve my tracery chops. What I'd really like to do is take a written work, run it through a parts of speech tagger of some sort, and turn those tags into tracery lists so that I can randomize the content of the book without changing the structure.

When I was a kid I really enjoyed those Choose Your Own Adventure books. Ever since learning to program, I've thought about how I could use software to make books, movies, and even music more interactive. For NaNoGenMo, I want to write a program that generates 50k-word Choose Your Own Adventure novels.

Edit:
All done! Here's a link to the repo: https://github.com/CameronEdmond/NaNoGenMo2017

Direct link to the output (Warning: big ol' PDF).

Direct link to the script (I am not good at coding)

Original:
Thinking of generating a text that builds up the necessary credentials to live in a fictional city over time, and judges its inhabitants as it goes.

Hoping to clear enough Uni work off my plate this month and next to allow myself to really give it a go. I'm currently writing a thesis on computer generated literature, and figured I should get in on the action.

We (re)watched Gossip Girl a few months ago; maybe procedurally generating e.g. 'A. and B. were seen at Hotel Fake-Name' would be fun?

Still thinking about what exactly I want to make, but excited to participate for the first time.

Full text: 🌱 [bean32_by_kasey_and_mark.rtfd.zip] 🌱(https://github.com/NaNoGenMo/2017/files/1425954/bean32_by_kasey_and_mark.rtfd.zip)

Warning: Not for those who do not want to be enlightened,
or as we call it, beanlightened.

The larger goal for November:

When one opens the .zip file, it first replays the process of its own creation. Then it posts to the repo by itself up to 5 times (up to 32 copies of itself). After that, it goes back and deletes itself from the repos in pairs until 2 becomes 1; the original Bean. Yet is the original bean, the original bean? Then it repeats the process for... (see Anti-Life Equation below).

I don't have a solid plan of how to go about this, but I'm thinking of generating a novel using a minute-by-minute technique similar to Nick Montfort's World Clock, but also using an api to grab real-world data. An early thought is to use the themeparks api to generate a novel about a day at Disney World, complete with real wait times (and possibly real travel times between attractions). After "riding" each ride, the computer will determine which ride to visit next based on travel distance and wait time, "wait" in line, and ride the attraction. To make the narrative a bit more interesting, I'm thinking of using Tracery to generate the narrative while the bot is waiting in line and walking between attractions.

Train up an embedding with some kind of older writing; attempt some kind of (LSTM?) based style transfer onto a newer story of appropriate length. This will probably be a complete disaster, but at least I'll have an excuse to play with some modern computational linguistics tools in the process.

Maybe some kind of journal format, based on whats happening at the geodetic coordinates of the international space station?

I have a few vague ideas about what to do, but only because I just heard about this five minutes ago on Twitter. I've been wanting to play around with machine learning and text summarization (robot reporting), and I've also been wanting to do #NaNoWriMo, and so I think that this will be a good combination of the two. I'll update this issue later as my idea develops.

Named after the scene from American Psycho... because this is probably going to be random, weird, and a little violent.

My program will generate a novel based on source material fed into it, using next-word prediction and style analysis with some adjustable settings for randomness and arc structure. The goal is to get it to produce something that is (at a minimum) readable; reasonably enjoyable; and to see how often it actually surprises me.

This will involve a hybrid approach to dialogue generation incorporating sentiment analysis on multiple emotional vectors as a selection mechanism for Markov models.

Will try and use an existing z-machine implementation to generate the text, but am not adverse to brute force extraction if I have to to.

I'm new to NaNoGenMo; this will be my first year! I want to make a picture book of 50 or so images that tell a story somehow. I'm inspired by @lowpolybot and @softlandscapes on Twitter. I expect this will be a bit abstract, but I think it will work out.

So. This is indeed an early set of ideas that I've submitted, by what, 3 months?

Well. I want to try my hand at some kind of soap opera simulation. I suspect that rather than generating a novel per se, this particular attempt will manifest itself best as a script. One that people could theoretically use to act. I wouldn't recommend it, but still. So I'm going to be simulating a soap opera, and that means simulating characters! Lots and lots of characters. That means emotional systems for said characters.

I think I'm going to use Robert Plutchik's system of classification of emotions, typically known as "Plutchik's wheel of emotions". There's 8 basic emotions:

• joy
• trust
• fear
• surprise
• sadness
• disgust
• anger
• anticipation

And each of these "primary emotions" per se have 5 states attached to them (I'm using this diagram as a simple basis) and these are:

• Stimulus event, which is what causes the character to feel a particular emotion
• Cognite appraisal, which is a linguistic summation of the emotion
• Feeling state, i.e. the actual emotion itself
• Behavioural reaction, which is what the character does as a result of the emotion
• Function, which is based on the utility of the character's action (e.g. anger is for destroying obstacles)

These primary emotions then combine into dyads, which are feelings made up of two separate emotions.

And that's only for my emotions. I also want to set up my characters to have personalities and moods, such that these three components (emotion, mood, personality) affect each other in a constant feedback loop.

And not only do I want to have my characters act like this, I also need to create some form of drama. Which is, in true soap opera fashion, likely going to revolve around romance, secret relationships, extramarital affairs, hatred, adultery, and so forth.

Wow. I've gone and done it this time, haven't I. At least I have another 3 months to dig my hole deeper.

I have a bunch of vague ideas for this.

Genre-wise, it's about a person in some vaguely define far future setting traveling the galaxy, exploring planets and the weird stuff that's on them.

As far as the generation goes, I have this idea of combining some template/grammar-based method of text generation (Probably Tracery.) with dynamically changing corpuses based on planets, situation, time and so on. I want to try and code some "decision making" into the character, so that things like "Which site to check out" or "when to leave a planet" depend on some internal or external state, like a "boredom" stat or an exploration percentage. (If I have the time, I'm also thinking random events that can influence that state, like the protagonist getting sick and thus missing couple of log entries or spending them at base recivering and watching old 21st century holonovels.

Locations like planets or the various exploratory sites on them would be randomly generated, in part using Tracery, while also serve as part of the dynamic corpus feeding the narration. For example, every location could be associated with specfic mood words, adjectives or colors which could be used to describe the things the character finds there.

(I also have this vague idea of the character potentially dying and, should the novel not be finished, the story continuing from the perspective of one the characters the earlier protagonist met during their travels.)

...

I'm pretty sure I won't get all that done, but I figure if I can do more than something that prints out "Meow" 50000 times I'm ok with whatever. :)

Links to previous years:

• 2016
• 2015
• 2014

I'm going to try to create a set of Shakespearean sonnets that are acrostics using every one of the letters of every sonnet to start each line. Each line of the result will be selected from elsewhere in the Shakespeare cannon, and they will use the standard 'ababcdcdefefgg' rhyming form.

The first 2 lines of the first sonnet are:

From fairest creatures we desire increase,
That thereby beauty's rose might never die,

So, removing non-alphas and splitting into 14-letter chunks, the first five output poems will be acrostic sonnets on:

fromfairestcre
atureswedesire
increasethatth
erebybeautysro
semightneverdi

There are 154 total sonnets, so the output will definitely be at least 50,000 words.

I think this should be possible with my poefy gem and I hopefully will be able to use the Open Source Shakespeare database as input.

There may well be 14-letter chunks that have no corresponding rhyming lines starting with the correct letters, but I'll cross that bridge when I come to it, in two months time :-)

So I write mostly short fiction (see here) - I thought it might be fun to generate a novel using Markov chains based on my stuff (published and all the stuff in the trunk).
Not sure what I'll write it in yet.

Retrying an idea from last year.

An Alien-style spaceship-based horror novel where all the narrative is told by way of computer logs from the various onboard systems. I want to generate a spaceship with various computer systems (e.g., security doors, air supply, research equipment), generate a bunch of characters with goals they are trying to accomplish within that ship, then just have the computerized systems log what happens as people move about and are attacked.

Inspired by "The Prophet's Paradise" from The King in Yellow. I'm interested in creating a machine that can output an enigmatic, repetitious prose poem in response to a prompt. At the end I'll compile 50k words of whatever it spits out.

Heroditus uses a structure called Ring Composition, which is (as far as I can tell) a series of recursive digressions. Injecting digressions seems like a good way to expand upon a short story.

So, I'm thinking that, given an ontology, it should be possible to generate an arbitrary statement and then apply a filter that, choosing a noun at random, injects a digression that describes that noun. Such a filter can be run on an arbitrary statement or sequence of statements until we reach 50k words.

(Alternately, we can have two categories of expansion: events and nouns. We can inject an event after another event with some indication of whether it happened before or after, or inject a digression about a noun in that event description, which could itself contain an event that the noun was associated with. I'd probably implement this by having specially formatted event IDs at the end of an event description and stripping them for display.)

It's very early to be starting on this, so I hope it still sounds like a good idea in November.

I'm going to generate an illustrated book recounting the adventures of a team of archaeologists as they uncover and postulate about a series of mysterious ruins. It is very much inspired by Emily Short's NaNoGenMo entry, "Annals of the Parrigues".

Lives of the Azar. Written by Idrizink Erebzidy (1215 - 1297), presented to Queen Maki Denkage (1248 - 1275). A collection of historic texts, covering some 1000 years of history from the land of Azar. Accompanied by maps and appendices.

One place where random content shines, I think, is as inspiration for human authors. Things like the random NPCs from https://donjon.bin.sh or the villages in Annals of the Parrigues make excellent resources for worldbuilding. I want to experiment with creating a 'novel' usable as a starting point for hand-crafting a fictional setting.

I plan to simulate a couple thousand software agents. These will interact semi-randomly, creating a history with tens of thousands of events. Agents can then be selected, and events relating to them collected into individual chapters. This will produce a collection of intertwined narratives - the novel - padded out with appendices listing notable agents and places.

Note: Description updated to match procedurally generated output.

I could use some feedback on this idea.

I'm thinking of creating a novel generator that generates text based off my geolocation. Could use suggestions for a good way to do this. I'm on my feet walking, biking and taking the train all day. I love the idea of using the location data to generate stories of kingdoms or other alternative stories.

One potential form is that it generates text over one day of my life, kind of like Ulysses. Maybe it generates text based on my geolocation data and then uses Wikipedia or other corpus to fill out information? Another idea would be that it creates almost like a fake guidebook, describing cities and towns

I'm particularly interested in roguelike text generation, like Dwarf Fortress's Legends Mode. I see there are some python hooks. Maybe I can connect this. Open to any/all ideas!

Basically it's just going to be a compendium of every Tracery trick I know, with some semblance of a plot attached.