paraphrase-visualizer

This script visualizes token sequences as a graphml graph in a form of

• a finite state automaton with one node per token; or
• a paraphrase graph.

The output graphml file can be displayed with a graphml viewer like yED. The output file does not contain any node positioning, thus the user can utilize the different layouts provided by the graphml viewer for the final graph representation.

Requirements

• python>=3.7
• yaml

Intro

The base for the output graph(s) is a list of sentences in a simple text file with one tokenized sentence per line.

In the output graph(s), the start and end tokens in the input sentences are marked.

The option -e/--end_points adds an artificial start and end token to the input sequences. Thus, the output graph is guaranteed to be a connected graph. Otherwise, also multiple (sub)graphs can be built according to the input sentences.

Automaton graph generation

src/automaton_graphml_builder.py <input_fn> [-o <output_dir>] [-e]

This script generates (a) "finite state automaton" graph(s) from the input sentences.

Note that from this automaton, you cannot exactly reconstruct the original sentences.

Paraphrase graph generation

src/paraphrase_graphml_builder.py <input_fn> [-o <output_dir>] [-e]

This script generates (a) graph(s) from the input sentences.

Contact

Eva Mujdricza-Maydt, me.levelek@gmx.de