Welcome to the (fairly maturing) NLP Playground
Select an App from the menu to get started or read on for a summary of the applications offered in this Natural Language Text Generation Playground.
Get started with BLOOM, the largest language model on the planet (as of January 2023...) and see what it can do!

Play the Choose Your Own Fate app that uses BLOOM to generate a story with some interesting flavor.

{" "} This app uses a tolerable set of ranges to give you an interesting set of options to choose from, guaranteed to be unique and different each time you press the magic "Generate" button. Since each option has its own parameters and "seeds" per run you're likely to find a great continuation to the text - be entirely on topic or taking a dramatic turn!

The app will start your story writing with a random prompt from our{" "} GitHub Gist , a collection of prompts that might lead you down an interesting path -{" "} or you may have a random chance at pulling today's top news, a headline, summary or abstract!

I personally find the stories generated from the headlines the most interesting as they have a chance at containing "Named Entities" (proper nouns, places, people) and it's also fun to see the model generate text in context of the latest events it actually has no knowledge of!

As mentioned each story-step generation is unique and backed by a randomized set of parameters. If one of the three options isn't suitable, that's okay, simply generate more options to choose from until you find the right one to fit your story.

As you make your selections the story will grow and saved for sharing or just to read later. You can also save your story as a PDF file for printing or sharing (coming soon!). See what stories already have been written or search for stories by their ID for later lookup.

Visit the Playground and play with{" "} BLOOM, the world's largest language model .

Built by the{" "} BigScience team (at{" "} HuggingFace), it is a 1.5 billion parameter model that can generate text of various flavors (be warned!) though you get to set your own parameters.

Let it loose on a prompt and see what it comes up with! Run it for 5 seconds on long prompts or for up to 2 minutes on short prompts. Set the temperature to get a varied response or consider the "top_k" values that define the vocabulary the model should consider when searching for the next "best word". Build as you go, but know your results aren't saved - you'll want to copy and paste the interesting results and share with friends. We're working on a few more apps to make this a more interesting and there is a growing backlog to improve Choose Your Own Fate.

Recently added was the ability to not only save the options a user creates but the various attributes about the generated selections. Training is ongoing to define a model that may further enhance the story generation. We would like to refine and provide the "best" options for the user to select from by understanding the language people select for (tone, coherency), but also seek and experiment with the best fit model parameters.

As our dataset to perform this analysis is still quite small feel free to generate many stories, with many steps. Our models will only improve with more user feedback and interactions.

There is zero usage of cookies or other tracking mechanisms on this site, however, feedback you provide is saved in a database for later analysis. Even the generated unique identifiers are randomized and not connected to a user. While there are plans for future apps to be added to our NLP playground, some may requiring authentication or cookies - those terms will be made clear for the application sections/apps in question.

Enjoy, and try to have fun!

Joe{" "} Henzi


Note: All stories and text generated by BLOOM are covered by the following License:{" "} BLOOM License (Big Science RAIL v1.0) . Using any generated text/stories for harm, spam or other illegal purposes is strictly prohibited by the License. Please read the License carefully before considering distributing any generated text/stories. The work you create is yours, but the text generated by the model comes from a variety of inputs and careful consideration should be made before sharing or taking the output of the model as any sort of guidance. The intent is to provide the public with a playground for experimentation and exploration of the capabilities of the model. The model is not intended to be taken seriously (like life itself at times).