A python package of common operations for AMRs
I wrote amr-utils to store operations that I often need when doing research with AMRs.
- Load multiple AMRs from a text file
- iterate through nodes, edges, or named entities
- output AMRs to useful formats: html (AMR string) or latex (AMR graph)
- associates a unique id to each node or edge (can be used for styling a particular element in a webpage or web app)
- use a Rule-Based Aligner to align AMR nodes and edges with words in a sentence.
Python 3.6 or higher
Input should contain AMR strings separated by a blank line. Lines starting with # will be ignored.
Amr-utils allows you to read AMRs from a text file and output them as latex diagrams, such as the following.
The default coloring assigns a different color to each node in a given row. To change a color by hand, just rewrite \node[red] as \node[purple], etc.
Run as follows:
python amr_latex.py [input file] > [output file]
Add these lines to your latex file:
\usepackage{tikz}
\usetikzlibrary{shapes}
Amr-utils allows you to read AMRs from a text file and output them as html. You can look in style.css for an example of styling.
Run as follows:
python amr_html.py [input file] > [output file]
Aligns AMR nodes and edges to words in its reference sentence. The aligner handles named-entity alignment, alignment of frames, and relations. A single alignment looks like:
broaden-01 :ARG0 :ARG1 ~ broadened (readible)
b b_ARG0_i b_ARG1_r ~ 1 (machine-readible)
Run as follows:
python alignment.py [amr file] [sentence file] > [output file]