31 lines
1 KiB
Nix
31 lines
1 KiB
Nix
{ lib, stdenv, fetchurl, cmake, protobuf, libiconv }:
|
|
|
|
stdenv.mkDerivation rec {
|
|
pname = "jumanpp";
|
|
version = "2.0.0-rc3";
|
|
|
|
src = fetchurl {
|
|
url = "https://github.com/ku-nlp/${pname}/releases/download/v${version}/${pname}-${version}.tar.xz";
|
|
sha256 = "sha256-ASdr6qbkSe71M7QmuuwidCa4xQhDVoXBJ2XqvSY53pQ=";
|
|
};
|
|
|
|
patches = [ ./0001-Exclude-all-tests-from-the-build.patch ];
|
|
cmakeFlags = [ "-DJPP_ENABLE_TESTS=OFF" ];
|
|
|
|
nativeBuildInputs = [ cmake ];
|
|
buildInputs = [ protobuf ]
|
|
++ lib.optional stdenv.isDarwin libiconv;
|
|
|
|
meta = with lib; {
|
|
description = "A Japanese morphological analyser using a recurrent neural network language model (RNNLM)";
|
|
longDescription = ''
|
|
JUMAN++ is a new morphological analyser that considers semantic
|
|
plausibility of word sequences by using a recurrent neural network
|
|
language model (RNNLM).
|
|
'';
|
|
homepage = "https://nlp.ist.i.kyoto-u.ac.jp/index.php?JUMAN++";
|
|
license = licenses.asl20;
|
|
maintainers = with maintainers; [ mt-caret ];
|
|
platforms = platforms.all;
|
|
};
|
|
}
|