There have been a number of previous constructed language translation relays, in which someone writes a text in their own conlang, then sends it with a mini-grammar and lexicon to the next player, who translates it into their own conlang, and sends it to the next player with mini-grammar and lexicon, etc. In the first Inverse Relay, we started with David J. Peterson writing a text in my conlang, gjâ-zym-byn, which he sent to me (no grammar or lexicon needed, of course; just the bare gzb text). I translated it into Lars Finsen's Urianian, which I had been studying over the two months or so between the time we fixed the relay schedule and the time I got the text from David. And so forth, with each participant translating from their own conlang into another person's conlang; finally Arthaey Angosii translated the text from her own Asha'ille into David's Kamakawi. Because of the greater time commitment required and the greater amount of documentation required for participating conlangs, only five people participated in this relay, as opposed to 20-30 in a typical relay.
Only fairly complete and well-documented conlangs should be used.
We start by posting brief descriptions of the conlang(s) we're offering for use in the relay, with links to online documention, and a note about what unpublished documentation we could send to the translator preceding us.
We'll also hash out a consensus on the general theme of the text or the semantic domains it will involve.
After everyone who seriously thinks their conlang complete and well-documented enough and is ready to commit the time for the relay has posted their language summary, we'll start working out the actual participants and the schedule. Everyone will list the languages they would like to translate into in decreasing order of preference, and specify when they will and won't be available during the planned relay period.
If some languages aren't listed by anybody as one of their preferred languages to translate into, they will not be included in this relay. [This might leave someone out who really would like to participate, but I don't want to make someone else translate into a language they would rather not spend two months studying.]
The relay master will figure out as best they can an ordering and scheduling that gives enough lead time (two months or more), doesn't have anybody translating during a time they will be even partly unavailable, puts the easier languages first and the harder languages later, and fulfills as many first-choice preferences as possible.
In the lead time before the relay starts, translators should not only study the public and unpublished documentation on their target conlang, but write practice sentences, send them to the conlang's creator for correction, and repeat. Conlang creators should coin new words they think might be needed in the semantic domains announced for the relay text.
The person at the beginning of the relay, who may or may not be the relay master, composes an original text in the language of the language of the person they're preceding (or if that seems too hard, composes an original text in English and translates it into their target conlang).
Each person will have one week to work on the translation from their own conlang into their successor's conlang (except for the person at the beginning, who has all the lead time before the relay starts to work on their original text in their target language).
During the relay itself, conlang creators can and should answer grammar and semantics questions from translators and provide new words as requested. They should not ask their predecessors any questions which might reveal details about the relay text. In short, ask questions only of your successor and answer questions only for your predecessor.
You need to commit to checking your email as frequently as possible during the week your predecessor is working on their translation into your conlang, so you can answer their lexical and grammatical questions as quickly as possible.
When you get the translated text from your predecessor in [a more or less mangled version of] your conlang, you then translate it into your successor's conlang based on your best attempt at understanding what your predecessor wrote. You don't ask them questions about what they meant to say in places where the text is ambiguous.
You will send to your predecessor a corrected version of their translation into your conlang; but you need not do this before you send your translation to the next person, or by any particular deadline.
If anyone has to drop out, we won't make their predecessor suddenly learn their successor's conlang and re-translate the text; instead, their predecessor will work up a short grammar summary and text-focused lexicon like we use in normal relays, and send that with their imperfectly translated text to the successor of the person who has dropped out. Or if the person who dropped out is unavailable to answer questions and coin new words for their predecessor when that person would have been translating into their conlang, their predecessor could instead send the corrected version of the text in their own conlang with a normal-relay grammar summary to the successor of the person who has dropped out.
But if someone is just going to require a little more than a week, and their successor's work/school/travel schedule is flexible, we might extend their turn instead.
After the relay is over, send your smooth English translation, your corrected version of your predecessor's text, and your comments on why you made the corrections you did to the relay master.
I posted to the Conlang Relay mailing list about my idea for an inverse relay on 19 March 2007. There were a couple of long threads where we hashed out the rules and the participants; at first about fifteen or twenty people said they would like to participate, but finally only seven people decided they had a language well-documented enough and enough free time during the summer to participate.
I worked up a schedule that included those seven people and their conlangs; two people dropped out long before the relay started, so it was easy to adjust the schedule then. No schedule adjustments were required during the relay itself. Roger Mills was originally going to go first and create the text, but he still hadn't come up with anything by the time he dropped out, so David J. Peterson created the text instead. We did not, in fact, ever get around to hashing out a consensus about the theme or semantic domains of the text, except that it ought to be low-tech and generic-human (not depending on the peculiarities of one fictional culture or another); David's text fit those criteria admirably.
I don't know for sure about the later relay participants, but David didn't correspond with me much about gjâ-zym-byn before he started composing his text (and only a little while he was composing the text), and I didn't correspond much with Lars before I started translating the text into his Urianian. I exchanged a number of emails with Lars during the several days I was working on the translation, however, asking him for words I couldn't find in his lexicon, and clarification about the usage of some that I did find but had ambiguous glosses.
Henrik Theiling used a program running on his company's supercomputer to crunch people's scheduling preferences and languages-to-follow preferences and find an optimal schedule, when he was relay master for Relays #13 and #14. I used a lower-tech method, drawing a graph with circles representing people, dark arrows representing first choices, light arrows representing second choices, and dotted arrows representing third choices; then trying to try to find an optimal path through the graph that followed as many first or second choice arrows as possible, and then trying to match up the best sequences I could find to people's announced dates of availability. I wound up having to have a gap in the schedule, where David sent his gzb text to Alex Fink as proctor, to send to me several weeks later.