Our data is released under various Creative Commons licenses.More information
If you love this content, please consider a donation.
Menu
Another suggestion (because I'm in the process of listing ones that have occurred to me before): it would be nice to have an function that lets you search for existing duplicates or near-duplicates of a sentence that you're about to submit, ideally without disturbing your workflow. (Thus, displaying this information in a separate frame would be helpful.)
Automaticamente, no se como se podria hacer. Pero he pensado que seria util un sistema para relacionar frases, diferente del de las etiquetas. Seria como etiquetas o grupos sin nombre.
Por ejemplo, "a Tom le han despedido" se podria relacionar con "A tom le han contratado"
Se pueden hacer etiquetas, pero a veces no sabes que nombre darle o son unas pocas frases, que no merece ponerles una etiqueta, pero es util que te sugieran cuando ves alguna de las frases relacionadas.
Esto con ejemplos se veria mucho mejor, a ver si busco unos ejemplos.
Con la version actual de tatoeba, no se puede pero en la siguiente se pueden utilizar el sistema de enlaces entre frases para mas cosas que unir tradiciones entre idiomas.
Espero que se me entienda. Tal vez he sido un poco lioso.
I did understand what you said. It makes me realize that my original comment included several requests:
(1) Provide a search that automatically looks for fuzzy matches. The current search engine (Sphinx) doesn't do this.
(2) Provide functionality that allows you to perform a search through the corpus for the content that you're about to submit. If part 1 is implemented, then this search can find similar as well as identical sentences.
Currently, in order to search for content that you're about to submit, you basically need to keep a separate open Tatoeba window, then copy your content (wrapped in double quotes if you want to find an exact match) into the Tatoeba search bar. This takes up extra screen space and of course also requires extra manual effort. As a result, I suspect that few people look for duplicate or related sentences on every submission. (Realistically speaking, it probably only makes sense to do these searches on short sentences with common vocabulary.) There is a script that is run periodically on the corpus to merge exact duplicates, so adding them is not as problematic as it might be. However, it's still a pretty complicated task to merge comments and histories as well as content, so avoiding the insertion of duplicates in the first place is preferable, especially if you're an advanced contributor, which gives you the ability to link to existing sentences.
It sounds like you're talking about an ability to manually create a new kind of metadata link that would aid fuzzy searches by including searches for synonyms, rather than relying on some existing synonym dictionary to do the fuzzy search. Is that right?