Migrating from 0.0.x¶

Version 0.1 is a rewrite; the camelCase functions of the prototype were replaced by a typed object API. The table maps old calls to their replacements.

0.0.x	0.1
`wn.readDocuments(path)`	`Corpus.from_txt(path)`
`sp.load(...)` + `wn.parseDocuments(D, entity_types, nlp=nlp)`	`SpacyEntityExtractor(model, labels=entity_types).annotate_all(corpus)`
`wn.createCorpMat(D_parsed)`	not needed (annotations are typed `AnnotatedDocument`s)
`wn.buildGraph(D_mat, c)`	`ImplicitNetwork.from_documents(docs, NetworkConfig(window=c))`
`V["entities"]`, `Ep[("e", "e")]`	`network.entities()`, `network.edges()`
`wn.clusterEdges(Ep, D_mat, model=...)`	`ContextualEdgeClusterer(SentenceTransformerEmbedder(...)).cluster_edges(network)`
`wn.convertToNetworkX(V, Ep)`	`network.to_networkx()`
`wn.plotNetwork(G, mode="show")`	`plot_network(network, show=True)` (extra `viz`)

# 0.0.x
D = wn.readDocuments("data.txt")
D_parsed = wn.parseDocuments(D, ["PERSON", "ORG"], nlp=spacy.load("en_core_web_sm"))
V, Ep = wn.buildGraph(wn.createCorpMat(D_parsed), c=2)
G = wn.convertToNetworkX(V, Ep)

# 0.1
from implicit_word_network import Corpus, SpacyEntityExtractor, build_network
network = build_network(
    Corpus.from_txt("data.txt"),
    extractor=SpacyEntityExtractor("en_core_web_sm", labels=["PERSON", "ORG"]),
    window=2,
)
G = network.to_networkx()

Behavioural differences:

Edge weights follow the literature exactly (ω = Σ exp(-δ)); the prototype's NetworkX export additionally halved the weights.
Compound entities come from the extractor's spans instead of merging I-tagged tokens.
The interactive networkx_viewer mode was dropped; use NetworkX/matplotlib or export to GraphML/GEXF for Gephi.