₦∃⋁∃ℝ²ᒪΔͲ∃ ▓░▒ d⍜t⍜ unicode poetry ▒▓▓░

"the day she picked a p̕͜r̀o̸tec̀͝t̨ed̷͞ f͟͡l̀o̵͢w͜͝e̕r̀͝ the sky ͡҉͜҉͢f̡͘e҉͠ļ̷̵̴̵͘͟͟ļ̷̵̴̵͘͟͟"

ᓚᘏᗢ A B̶r̴o̴k̶e̶n̷ I̴ and A B̶r̴o̴k̶e̶n̷ Y̵o̷u̶ ᓚᘏᗢ

Never2Late 𝓲𝓼 an ██████ poetry-██████ making machine ████████ from what is ████████ lost. In the spaces between ███ █████ on WordReference.com, ███████ oblivion, we find the ██████████ of language from ███████ our collective ████.

Developed for la Fête des Fleurs, 2025 edition.

---

"𝔻𝕖𝕒𝕥𝕙 𝕚𝕤 𝕟𝕠𝕥 𝕥𝕙𝕖 𝕖𝕟𝕕 𝕠𝕗 𝕓𝕖𝕒𝕦𝕥𝕪. ℙ𝕠𝕖𝕥𝕣𝕪 𝕚𝕤 𝕥𝕙𝕖 𝕣𝕖𝕤𝕦𝕣𝕣𝕖𝕔𝕥𝕚𝕠𝕟 𝕠𝕗 𝕞𝕖𝕒𝕟𝕚𝕟𝕘."

▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓

ℂ𝕣𝕖𝕕𝕚𝕥𝕤 𝕒𝕟𝕕 𝔸𝕔𝕜𝕟𝕠𝕨𝕝𝕖𝕕𝕘𝕖𝕞𝕖𝕟𝕥𝕤 ⌬⎰◊⠬⠃↤☘

• Profuse thanks to everest pipkin, whose work "i've never picked a protected flower" inspired this project.
  Their 400-page generative glitch poetry PDF is IMO a masterpiece of digital art and concrete poetry <3
• The original forum thread which inspired the projects' names

"i've never picked a protected flower" See the source at github.com/everestpipkin/never

🅼🅰🅽🅸🅵🅴🆂🆃🅾

Instead of relying on external APIs for conceptual relationships, we use local NLP models with spaCy + vector similarity computations to retrieve semantically proximate words directly from the corpus, based on English NLP pre-trained word models.

The project is composed of four independent yet interconnected parts:

1. scraper.py - ɨצɬཞąƈɬʂ raw language from the collective unconscious
2. clean.py - ꎇꀤ꒒꓄ꏂꋪꌗ the noise from the signal
3. generator.py - ੮гคຖŞ৲๏гຖŞ random words into semantic poetry
4. wallpaper.py - єη¢α∂єѕ the poems into unicode patterns

▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓

🄸🄽🅂🅃🄰🄻🄻🄰🅃🄸🄾🄽 ╜◊⎼▲▅⠌◇╯◆

# Clone the repository
git clone https://github.com/PLNech/never2.git
cd never2

# Optional but recommended: create a virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install spacy beautifulsoup4 requests numpy
python -m spacy download en_core_web_lg

🆄🆂🅰🅶🅴

1. 𝕊𝕔𝕣𝕒𝕡𝕚𝕟𝕘 ℙ𝕠𝕖𝕥𝕣𝕪 𝕀𝕟𝕘𝕣𝕖𝕕𝕚𝕖𝕟𝕥𝕤

python scraper.py --start 1 --end 100 --delay 2 --batch 20

Options:

• --start START - Starting page number
• --end END - Ending page number
• --delay DELAY - Delay between requests (seconds)
• --batch BATCH - Save to disk after this many pages

"In the gaps between forum words, I found sentences that never existed."

2. ᶜˡᵉᵃⁿⁱⁿᵍ ᵗʰᵉ ᴰᵃᵗᵃ

python clean.py -i english4.csv -o clean4.csv

Options:

• -i INPUT, --input INPUT - Input CSV file (default: english4.csv)
• -o OUTPUT, --output OUTPUT - Output CSV file (default: derived from input)

"Language is a virus 𝒇𝒓𝒐𝒎 𝒐𝒖𝒕𝒆𝒓 𝒔𝒑𝒂𝒄𝒆."

3. 𝔾𝕖𝕟𝕖𝕣𝕒𝕥𝕚𝕟𝕘 ℙ𝕠𝕖𝕞𝕤

python generator.py -i clean4.csv -n 20 -l 5 -f html --feet 575 --cache vectors_cache.pkl

Options:

• -i INPUT, --input INPUT - Input CSV file with cleaned sentences
• -n NUM_POEMS, --num-poems NUM_POEMS - Number of poems to generate
• -l LENGTH, --length LENGTH - Maximum number of lines per poem
• -m MODEL, --model MODEL - spaCy model to use (default: en_core_web_lg)
• -o OUTPUT_DIR, --output-dir OUTPUT_DIR - Directory to save generated poems
• -f {txt,html,json}, --format {txt,html,json} - Output format
• -s SEED, --seed SEED - Initial seed word for poem generation
• -p PORT, --port PORT - Run as HTTP server on specified port
• -b BATCH, --batch BATCH - Generate a large batch of poems (specify count)
• -w WORKERS, --workers WORKERS - Number of worker processes for batch generation
• --cache CACHE - Cache file for word vectors and similarity
• -r RELATED, --related RELATED - Test related words
• --feet FEET - Pattern for syllable counts (e.g., "575" for haiku, "12x4" for alexandrines)
• --test - Run tests for word similarity

Example - python generator.py -i clean.csv -b 100 -f txt --feet "575" --cache .poem_cache.bin

Loading spaCy model: en_core_web_lg...
Loading cache from .poem_cache.bin...
Loaded 1545 word vectors, 310284 similarity pairs, and 216 related words sets
Loading data from clean.csv...
Loaded 10268 sentences from 2166 users
Found 15 words related to 'thought': ['think', 'believed', 'probably', 'because', 'something', 'understanding', 'knowing', 'deliberately', 'remember', 'imperceptibly']
Using cached related words for 'blush'
Found 15 words related to 'trail': ['road', 'mountainside', 'roadway', 'mountain', 'route', 'canyon', 'path', 'ridge', 'mountains', 'outcropping']
Using cached related words for 'flame'
Using cached related words for 'decay'
Using cached related words for 'sky'
[...]
Using cached related words for 'brick'
Using cached related words for 'boundary'
Saved 100 poems to poems
Saving cache to .poem_cache.bin...
Saved 1545 word vectors, 322200 similarity pairs, and 225 related words sets

"Every poem is an epitaph. ／人◕ ‿‿ ◕人＼"

4. 💮 Creating Unicode Wallpaper 💮

python wallpaper.py -W 80 -H 24 -g pmm --poem poems/poem_1.txt -f html -o pattern.html

Options:

• -W WIDTH, --width WIDTH - Width of the pattern in characters
• -H HEIGHT, --height HEIGHT - Height of the pattern in characters
• -s SEED, --seed SEED - Random seed for pattern generation
• -g {p1,pm,pmm,...}, --group {p1,pm,pmm,...} - Specific wallpaper group to use
• -p PORT, --port PORT - Run as web server on specified port
• -o OUTPUT, --output OUTPUT - Output file path
• -f {txt,html}, --format {txt,html} - Output format
• --poem POEM - Text file containing poem to embed
• -D DENSITY, --density DENSITY - Character density (higher = more characters)
• -b {white,black}, --background {white,black} - Background color for HTML output
• -i [INTERACTIVE], --interactive [INTERACTIVE] - Run in interactive mode with auto-updates
• --chaos - Run in chaos mode with varying density and update intervals

Examples ⌯◓▀⌂┣⎺◊

The -D parameter controls the density of Unicode characters in the output. Here are examples with increasing density values:

Command	Results
-D 50 -H 10 -W 40
-D 100 -H 12 -W 40
-D 500 -H 28 -W 80
-D 1000 -H 10 -W 40

🅣🅞🅓🅞 ▒▓▒▓▒▓▒▓▒▓▒▓▒

• Rework wallpaper.py: Current implementation is far from the unicode magic of everest's original project. Needs significant upgrades to achieve bolder aesthetic impact.
• Add support for more output formats and embedding options
• Improve interactive visualization modes
• Rework web interface
• Enhance thematic poem generation
• Fine-tune feet/syllable matching logic

▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓▓

ℓιƈєηѕє ╛╝╜┶⌧┊

This project is free software. Take it, use it, modify it, reshare it. Just add/download your own dataset!

🅛🅘🅒🅔🅝🅢🅔🅓 🅤🅝🅓🅔🅡 🅣🅗🅔 🅖🅝🅤 🅖🅟🅛-𝟹.𝟶