Skip to the text
Parseh
پParseh guide Lookup and languages
On this page

Lookup and languages

This section has two halves, and they meet in the same place: the language of what you are reading.

The first half is reading help. A book or a video can be read long before its glosses are written, and for those weeks most chunks are phrases you cannot get past. One page, Reading help in Settings, sets up everything that helps with that: a dictionary for each language, the components of a Japanese or Chinese character, sentences somebody has already translated, a translation model that runs inside the page, and a small table of English synonyms that sharpens what the model shows. All of it is optional, all of it is off until you turn it on, and none of it is ever written into a book or a video as if it were a gloss.

The second half is the languages themselves. Parseh teaches eleven: Persian, Arabic, Italian, Japanese, French, German, Turkish, English, Hindi, Spanish and Chinese. They share every door, button and file format; what differs is what each language needs in order to be read — its direction, its script, its fonts, its passes, its transliteration.

1. Reading help#

  1. The reading-help page — how to reach it, what each language can have and what it costs, what every row and button does, where the files go, whose work they are, and how it all works.
  2. Dictionaries — the dictionary switch in every reader and player, and the panel it opens under a chunk nobody has glossed.
  3. Kanji and Hanzi components — Decompose Kanji and Decompose Hanzi: a character taken apart into the pieces it is built from.
  4. Sentences somebody translated — a real sentence and its real translation, chosen because it shares the rare words of your chunk.
  5. A translation model in the page — the whole sentence read by a machine, inside your browser, with your chunk’s share of it marked.
  6. Synonyms for the machine’s reading — how the mark finds begin where the dictionary said start.

2. Languages#

  1. The eleven languages — codes, scripts, directions, fonts, digits and what each one calls its transliteration.
  2. Target language and gloss language — what you are learning, and what the meanings are written in.
  3. The language chips — the row of languages at the top of the hub and every library.
  4. What differs, language by language — passes, readings, words, vertical text, verbs, and what each door does differently.
  5. Adding a language — for whoever maintains Parseh: the one command that adds a twelfth, and what is left to write by hand.

3. Speech to text#

  1. Speech to text — an optional program and two Whisper models that make a transcript on this computer while you add a video: the CPU that always works, an NVIDIA graphics card when Parseh can prove it is usable, what each part costs, who may get it, and where the files are. It has a door of its own in Settings, and is not a part of Reading help.

Nothing here is required. A reader with no dictionary, no corpus and no model is exactly the reader it always was: it asks the server once whether anything is installed, and when nothing is, it draws no extra button and makes no further request.