=== reel.py

The python version of reel is better suited for pulling small graphs. Performance degrades rather quick.

Requires Elementtree and igraph (python-igraph, beware there is a pip package that goes for the name igraph; also you need imagemagick/convert to install this package). Pretty much alpha and prone to breaking. Check the parameters by running ./reel -h

For example:

~./reel.py -w banco -l es -f es pt apertium-es-pt.es-pt.dix -f en es apertium-eng-spa.eng-spa.dix -f fr es apertium-fr-es.fr-es.dix -d 5 -m -g

Will return both a list of links via stdout and a graph (in a temporary file opened using imagemagick). Each word is in the form surface|POS|language, where language is the name given as parameter (e.g. es for Spanish in this example). The different POS classes are defined in the "header" of the bilingual dictionary file (<sdefs>).

The d3js mode can be used to generate a d3-like json file. Dump the output into web/data.json, serve the folder (e.g. python -m SimpleHTTPServer) and open index.html to see the representation.

==== reel.cc

The c file was used to generate the dictionaries for the submission. DFS is used to generate all length-4 cycles. Check example for an imput example.

Code for the DFS was copied from https://www.geeksforgeeks.org/depth-first-search-or-dfs-for-a-graph/

==== Apertium bilingual dictionary format simplified reference

Apertium refers to the languages as R for the "right" language and L for the "left" language; e.g. eng-spa L=eng R=spa

• dictionary
• alphabet > defines the alphabet, not usually ysed
• sdefs > the POS classes
• sdef > each POS class
• section > a section of the dictionary. @id=main is always there. Additionally, you might find other sections with automatically generated entries or regular expressions
• e > an entry
• @r="RL" > means that the entry is only valid when translating from right to left language. There is also LR. Default is both directions are fine
• p > a pair
  • l > the left word
  • r > the right word * i > an identity. It is exclusive with p, and, while it is discouraged, still shows in old dicts.

Each p or i have

• Plain text with the word in canonical form, mixed with <g> and <b/> in the case of multiwords, e.g. <r>seguro<g><b/>de<b/>vida</g><s n="n"/>, seguro is a word that can be flexed and de vida is the invariable part
• s POS info
• @n the POS