This probably seems like a strange question so I'll get to the point first.

Main Point

I essentially want to be able to take two input languages and create sentences by using vocabulary from both based on an algorithm. For example: take English and Spanish and decide that the sentence is going to be 80% English and 20% Spanish, a typical output might be "Hello, my name es Tim".


Is there an algorithm that essentially allows you to "lerp" between two languages?


I acknowledge that such an algorithm wouldn't be perfect at appreciating the nuances between different languages, but to some degree this wouldn't be necessary.

I also acknowledge that this wouldn't necessarily be the correct community to ask this question in, but honestly I can't decide where it would go.

  • I don't see an actual question here. Please edit to clarify. May 3 '18 at 18:41
  • What is the goal? Do you want realistic Spanglish? An artificial language? To bypass some filter? May 4 '18 at 4:47
  • @A.M.Bittlingmayer A procedurally-generated environment within which language differs depending on where you are. May 4 '18 at 6:44
  • 1
    The answer below is on point. In the real world fusion usually occurs asymmetrically, with structure and function words from one, and content words from the other. The less related the languages, the less their structure is 1:1, so it becomes non-trivial to do substitutions programmatically. May 4 '18 at 7:51
  • But if realism is a non-goal, then it is easier. Your system can generate constructions like Hello my name está Tim or How sois you? that are not found in the wild ie invalid. May 4 '18 at 7:54

In general, there is probably no such thing because different languages tend to differ in word order and syntax.

There are some solutions: Use handcrafted sentences which are word-aligned and then choose the word from language A with probability p or from language B with probability p-1.

Less precise: Take sentences from language A and replace the words with probability p-1 with words from language B. In this model, the underlying syntax is always from language A.

You may be interested in the project INCOMSLAV and their publications: They use (among other methods) interpolated mixes of Slavonic languages to test mutual intercomprehension.

EDIT: Instead of the probabilistic algorithm suggested above, one can also use a fixed pattern for the language selection, e.g., for 50% alternately selecting from languages A and B. This will lead to a more uniform distribution of the two languages.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.