Normalizers are created with the particulars for any given document format captured in on-page Code items that are collected and installed as a module.
digraph { rankdir=LR node [shape=box style=filled fillcolor=bisque] html -> normalize -> import docs -> normalize -> download { rank=same page -> normalize normalize -> norm normalize -> script -> norm } }
A page is configured to run normalize-edits.html in a Frame plugin. This application finds and loads script fragments from Code items as a module to customize the document normalization process. github 
The scripts export two functions used to traverse the chosen document which will have been already loaded into off-screen DOM memory.
export const titles = () => { ... } export const items = (title) => { ... }
The scripts import the norm.js module which manages the off-screen DOM and offers helper functions proven useful for finding and formatting titles and items. github
div = { ... } setdiv = (newdiv) => { ... }
Read and update the off-screen DOM.
select = (selector) => { ... } selectAll = (selector) => [ ... ]
Query the off-screen DOM with querySelector and querySelectorAll returning a element or an array.
frequent = (selector, nextselector) => { ... }
Find the most frequently used class for selected elements, optionally further selected with the selected element.
more = (elem) => () => { ... } repeat = (cursor) => [ ... ]
Create a cursor over element siblings returning <p> items. Repeatedly apply and collect a cursor's non-null returns.
item = (elem) => { ... }
Create a wiki story item to represent an element, typically a <p>. Attach an id constructed from a few long words that are likely to match across document types.
Create new normalizers from the Normalize Template.