search

IBM / transition-amr-parser

SoTA Abstract Meaning Representation (AMR) parsing with word-node alignments in Pytorch. Includes checkpoints and other tools such as statistical significance Smatch.
Apache License 2.0
237 stars 45 forks source link

Instruction of Successfully Installing JAMR for obtaining Alignment #25

Closed longxudou closed 2 years ago

longxudou commented 2 years ago

The original JAMR was installed in 2015 or 2016 in the server, thus some packages were broken or not updated. The default setup script in JAMR repos is set to a version no longer available. I have pushed a new JAMR repos to deal with this issue. More details on https://github.com/jflanigan/jamr/issues/44.

The most important part is to install the updated version of scala and sbt. To manually install them, you can use SDKMAN! and run the following:

sdk install sbt 1.0.2
sdk install scala 2.11.12

And then you could replace https://github.com/IBM/transition-amr-parser/blob/dfd8352ea2ee3ff153b691edb6cd7ee541d53b2e/preprocess/install_alignment_tools.sh#L13 with https://github.com/DreamerDeo/JAMR.git , where the scala version are clarified in some config files. (of course, you could also clone the original JAMR repos and do this manually)

ramon-astudillo commented 2 years ago

thanks! with https://github.com/IBM/transition-amr-parser/releases/tag/v0.5.2 we are deprecating the JAMR aligner in favour of the ibm-neural-aligner. We can make the JAMR alignments available via the same mechanism as the checkpoints.