Graph

its all a graph maaaaaannnnnn

Classes:

Graph(recurser)

Represent and manipulate a Recurser object as a graph!

Functions:

datashader_network(recurser[, root_word, …])
draw_labels(stack, net_layout)

class recurse_words.graph.Graph(recurser: recurse_words.recursers.graph.Graph_Recurser)

Bases: object

Represent and manipulate a Recurser object as a graph!

Variables

filtered_edges (pandas.DataFrame) – Edges after some call to Graph.filter(), initialized as Graph.edges.

Methods:

_translate_edges(edges)
_sort_edges(edges)	Sort edges in order of source, subword, replacement, target
filter(root_word[, depth])	Filter the graph from some root word and depth
make_datashader()

Attributes:

edges	Pandas dataframe of all edges, in columns:
untranslated_edges	Precurser to Graph.edges, gets edges before they’re translated back to human-readable.
filtered_edges	Edges after being filtered by Graph.filter()

_translate_edges(edges: pandas.core.frame.DataFrame) → pandas.core.frame.DataFrame

_sort_edges(edges: pandas.core.frame.DataFrame) → pandas.core.frame.DataFrame

Sort edges in order of source, subword, replacement, target

property edges

Pandas dataframe of all edges, in columns:

('source', 'subword', 'replacement', 'target')

If corpus has a translation layer, edges should be the translated version of the edges.

Returns

pandas.DataFrame

property untranslated_edges

Precurser to Graph.edges, gets edges before they’re translated back to human-readable.

Equivalent to edges if there is no translation layer

Returns

pandas.DataFrame

property filtered_edges

Edges after being filtered by Graph.filter()

Until then, equivalent to Graph.edges

Should be used by all graph generating mechanisms rather than

Graph.edges, which is intended to be the immutable source of edges.

Returns

pandas.DataFrame

filter(root_word: str, depth: int = 0) → pandas.core.frame.DataFrame

Filter the graph from some root word and depth

Note that this words off the original, untranslated edges to avoid any loss of meaning that might happen in the mapping from the network edge space to the human-readable space, eg. in the corpi.CMUDict phonetic corpus, multiple english words map to a single phonetic representation, so when translated back to english one is chosen at random. Filtering in the network space avoids the ambiguity, hopefully.

Parameters

• root_word (str) – Root word to

• depth (int) – number of additional steps to take. 0 returns just the words immediately connected to the root word.

Returns

pandas.DataFrame

make_datashader()

recurse_words.graph.datashader_network(recurser: recurse_words.recursers.graph.Graph_Recurser, root_word: Optional[str] = None, depth: int = 0, res=(10000, 10000), cmap_min=65)

recurse_words.graph.draw_labels(stack, net_layout)