Closed GMarzinotto closed 6 years ago
If I recall you cannot really pass an external lexicon file for inference, even to the binary (of if you somehow can, send me how :-)
The UDPipe API documentation is available at http://ufal.mff.cuni.cz/udpipe/api-reference. All bindings have exactly these methods, but renamed to use camelcase and more trivial types -- this simplified API is described at http://ufal.mff.cuni.cz/udpipe/api-reference#cpp_bindings_api.
Also, UDPipe comes with a REST server (udpipe_server
), which can be used to load the model once and then process any number of inputs without a delay.
Hi all, I forked udpipe some months ago to add exactly this functionality. You find it here https://github.com/ioan2/udpipe
Johannes
@ioan2 :-) New version of UDPipe will be released in a couple of months (with much more powerful models); external dictionary for inference will also be supported [optionally specifying only some columns] (and external word embeddings will also be supported).
I apologize, I got confused as I was working with the forked version!
However, thank you for the documentation and the information about the REST server ! I think the REST server is just what I need!
I'll be closing the issue
I have been using UDPipe and so far it works like a charm
But when I want to use the CLI to parse several sentences in different files UDpipe reloads the model for each file. Since I have many small files, this lags the process a lot.
To prevent this from happening I created a python script using the wrapper provided here. So I load the model once and then I use it to parse all the files that I want.
The problem is that the python wrapper is not fully documented I would like to know how to use all the parameters that are available in the CLI.
For instance, how to provide an external lexicon file for inference using the python wrapper.
Thank you very much for your help, Gabriel M