From c51fc966af10e93d5935ac01b4d8998a7b33939e Mon Sep 17 00:00:00 2001 From: bjascob Date: Sat, 27 Nov 2021 08:47:10 -0700 Subject: [PATCH] update docs for parse_t5 v2 --- README.md | 2 +- docs/install.md | 2 +- docs/models.md | 2 +- 3 files changed, 3 insertions(+), 3 deletions(-) diff --git a/README.md b/README.md index 4e2e1a5..cbcb3b7 100644 --- a/README.md +++ b/README.md @@ -34,7 +34,7 @@ The system includes different neural-network models for parsing and for generati [SPRING model](https://github.com/SapienzaNLP/spring). The basic process is similar to the parse_t5 model except that the pretrained transformer used is the bart-large model which is larger than the t5-base model used in parse_t5. -* Parse (StoG) model parse_t5 gives an **81 SMATCH score** with LDC2020T02. This model uses the +* Parse (StoG) model parse_t5 gives an **82 SMATCH score** with LDC2020T02. This model uses the pretrained HuggingFace T5 transformer model to convert sentences to graph-encoded sequences which are then deserialized into an AMR graph. diff --git a/docs/install.md b/docs/install.md index d95325b..7bb43ef 100644 --- a/docs/install.md +++ b/docs/install.md @@ -43,7 +43,7 @@ Download the pretrained parse models from: * [model_parse_spring-v0_1_0.tar.gz](https://github.com/bjascob/amrlib-models/releases/download/model_parse_spring-v0_1_0/model_parse_spring-v0_1_0.tar.gz) (or see [GitHub/amrlib-models](https://github.com/bjascob/amrlib-models)) -* [model_parse_t5-v0_1_0.tar.gz](https://github.com/bjascob/amrlib-models/releases/download/model_parse_t5-v0_1_0/model_parse_t5-v0_1_0.tar.gz) +* [model_parse_t5-v0_2_0.tar.gz](https://github.com/bjascob/amrlib-models/releases/download/model_parse_t5-v0_2_0/model_parse_t5-v0_2_0.tar.gz) (or see [GitHub/amrlib-models](https://github.com/bjascob/amrlib-models)) * [model_parse_gsii-v0_1_0.tar.gz](https://u.pcloud.link/publink/show?code=XZD2z0XZOqRtS2mNMHhMG4UhXOCNO4yzeaLk) diff --git a/docs/models.md b/docs/models.md index 29b5d81..cada950 100644 --- a/docs/models.md +++ b/docs/models.md @@ -19,7 +19,7 @@ Additional inference parameters: See amrlib/models/parse_spring/inference.py for implementation details. ## Parse T5 -**81 SMATCH score** with LDC2020T02. +**82 SMATCH score** with LDC2020T02. This model is based on the pretrained [HuggingFace](https://github.com/huggingface/transformers) T5 transformer to convert sentences to graph-encoded sequences which are then deserialized into