Hey guys in gals, I was wondering how I should let all the different species on my story communicate with each other, I don't want to go the cheap route and hand-wave it by saying something like "They all speak galactic basic" or "It's just translated for your convenience". I was wondering if a sort of translation device would be conceivable? A sort of multi-language dictionary that can translate the user's words into the language of the participant/receiver. I don't like to go too "soft" when it comes to Sci-Fi, I try and keep things real and conceivable. But I would love to hear other people's ideas!
If you want a universal interpreter*, you may just have to concede this bit of softness in an otherwise hard Sci-Fi setting. To date real-world interpretation software is extremely basic and has to be fed ultra-crisp, super-precise syntax to produce anything usable. And as we all know, people don't speak in ultra-crisp, super-precise ways. And this is dealing with human languages. Just imagine the dynamics and structures that would come into play for languages evolved on worlds away from Earth. * Remember mentioning your pet peeve over sentient vs sapient? This one is mine: I do this kind of work for a living and in the field interpreting is live speech and translation is written documents. Every time I hear Sci-Fi talk about a "universal translator" I get pissed when I don't see a hard-copy or words on a screen.
Sorry to pet your peeve, I will use "interpreter" for now on! Thank you for the feedback, it really means a lot to me. I guess I can go a tad bit soft here, I know exactly what you mean with things like exact syntax, even programs like Cortana on my computer often mis-interpret what I say. Another area I go soft in is FTL travel, because "they spent 18 months in cryogenic stasis." does not an exciting plot make. I just don't want to hand-wave it, especially for species that would physically be incapable of saying some languages. A Universal Interpreter ( ) should be standard-issue gear on all trans-planetary voyages. They would come especially in handy for those working on the multi-species Capital* *Placeholder name, I assure you. Imagine two gigantic toy tops, connected at the handles, one atop the other. The upper top is the delegations sector, where representatives of all sapient species can gather to discuss political matters. The lower top is internally divided into many sections, each designed to mimic the home world of a representive, to combat homesickness during long political debates (some of which have lasted for months).
A universal translator would likely be used. It's not as hand-wavy as you might think. Google is a universal translator. You know how if you ask Google to translate something for you, it'll do it? Nobody at Google programmed that translation. What they did is built a generic language learning algorithm, then fed it thousands of legal documents in various languages and compared and contrasted them to build its translations. It has limits of course, but it successfully and easily bridge language family gaps and grammar differences. Information is communicated in a predictable density, which can be figured out by a complicated enough neural network. We've also discovered some language structures that we believe are universal to all languages because many different species that have convergently evolved. Things like the second most used word will happen half as much as the most and the third most common word will be used a third as often, and all the way down into the thousands. Dolphins, meerkats, humans, crows, all show this structure. While that's a long way off from having Google be able to translate the language of dolphins, but there is no real reason that we couldn't get there as it stands right now.
Nope. Disagreed in full. You're talking about written text. We don't write in the same manner that we speak. Speech is a confused, disjointed paradigm which the human brain, most miraculous of computers, manages to smooth out and process in a way that makes sense. I do transcriptions for the USDOJ on a regular basis. Legal transcriptions must be ultra-exact. Every uh, um, false start, broken sentence, detour, derail, every single sound exactly as it is spoken. When you do this kind of work you come to realize just how utterly amazing it is that the listener is able to process any of this into something that makes sense. The thing is, we don't notice this train-wreck that we call spoken language. We genuinely don't. We hear it and our brain is able to put the train back on the tracks, the cars all in a nice little row. And what one hears in movies and TV as dialogue, that's not real speech either. That's as smoothed out as written text. A universal interpreter (not translator; translation is written) would have to deal with the way we actually speak, not the way we think we speak. And that's just taking into account English. There are other languages in other cultures that are much more permissive of all the anomalies I mention. My native Spanish is a nightmare for a transcriptionist. The culture is very permissive of conversations that branch off in ever-evolving fractals, which gives rise to the same level of permissiveness at the syntactic level. Human interpreters have a hard enough time, and this with the most complex computer that has ever existed - the human brain - at our disposal.
Yeah but the babel fish .... (which if you recall cause god to explode in a puff of logic by proving he existed and therefore that he didn't)
All I can say is that if the U.S.S. Enterprise's universal interpreter had to deal with the Trump during some kind of treaty negotiations or some shit...
Yes, but even speech to text is heading in the neural network direction. Deep neural networks have the perfect architecture for language recognition because they allow for lots of abstract layers, so handling things like context, accents, slurring, stuttering, end up just being extra normalization nodes in the network, which would be figured out by the network itself. It's not that much harder to pull apart language from an audio clip than it is to pull objects from an image. The deeper you make the network, and the more data you give it to learn from, the more accurate it will become. It will make lots of mistakes at first, but in the time between now and First Contact, there is plenty of time for it to be figured out. An example of things that are very hard to explain, that a NN figured out on it's own was a problem that Google had where it's software misidentified black human faces as gorillas. To a human mind, the difference is obvious, but like language, it's really really hard to describe why it's different. To correct the issue, Google simply showed the network more black people and gorillas and the network learned the difference on it's own. We would likely have to do the same thing with strange accents and phrasing, but the time required for that would be minimal for a cursory understanding and only an order of magnitude more for higher precision.
Unless this piece of technology is core to your plot, rather than just part of the foundation of your world, I think that you may be better off handwaving and doing some kind of technological babelfish.
It's not super-super core to the story, but it does form a small plot point in the early chapters of the story. Mostly in regards to the interaction between two of the main characters, who are of different species.
