regexpoetics-lang

This document sketches out a basic domain-specific language (DSL) for converting text into values that can be used for music synthesis, etc.

This idea is connected to a project called regexpoetics developed by Jonathan Wohl and Andy Dayton, a simple tool for writing text live and using it to generate audio.

Video screen capture of a Regular Expressions performance.

The language itself would be implemented using JavaScript and either peg.js or ohm.

Semantics

The semantics of this language are based on the final goal — using words to generate sounds. There are three main steps in that process.

1. Finding patterns within a larger text
2. Filtering those patterns or extracting information from them
3. Sending the resulting information out to be interpretated

Syntax

This is a very early draft of potential syntax, subject to quite a bit of change.

Expressions can be written in LISP-y syntax:

(/channel (~notes (pattern all /(.*)/)))

Or in arrow notation

-> all ~notes /channel

Or a combo

-> (pattern /(.*)/) ~notes /channel

Define a pattern

(pattern all /(.*)/)

Run a filter

-> all ~notes

Send to OSC channel

-> all ~notes /channel

Core Functionality

• Define Patterns - use regular expressions to define patterns that can include groupings
• Output - output data to various channels. most likely will be sent to WebSocket which can then be sent via OSC
• Define Filters - define JavaScript functions that will filter incoming data and return modified data. filters can be synchronous or asynchronous.

Potential Addons

• Notes Filter - parses data for "note" letters (A-G) and returns only the note data
• Word Distance Filter - given a set of two words, uses NLP to generate a numeric value based on word distance.
• Sentiment Analysis Filter - given a word or a set of words, run sentiment analysis and return a numeric value.