def make_sim_graph(akw, col_graph):
"""
Takes an {author: author keywords} dict and the collaboration graph for a school and creates a graph
where the nodes are authors linked if their keywords are similar.
Returns the similarity graph
"""
sim_graph = nx.Graph()
sim_threshold = 0.2
authors = akw.keys()
values = akw.values()
for i in range (0, len(authors)):
author1 = authors[i]
# make keywords into set to remove duplicates then back into list to maintain order
keywords = list(set(values[i]["keywords"]))
# Add the author to the similarity graph
add_sim_graph_node(author1, keywords, sim_graph, col_graph)
# Get a stemmed version of the author's keywords
stemmed1 = (tu.stem_word_list(keywords[:]))
# Compare author against each other author in graph
for j in range(i+1, len(authors)):
author2 = authors[j]
keywords2 = list(set(values[j]["keywords"]))
add_sim_graph_node(author2, keywords2, sim_graph, col_graph)
stemmed2 = (tu.stem_word_list(keywords2[:]))
# Check similarity of keywords
sim = tu.check_kw_sim(stemmed1, stemmed2)
# the similarity score
ratio = sim[0]
# the indices (in the longest of the two author keyword lists) of the keywords that are similar
indices = sim[1]
matched_words = []
if len(keywords) > len(keywords2):
longest = keywords
else:
longest = keywords2
# Get the keywords in the indices returned from check_sim
for i in indices:
matched_words.append(longest[i])
# If similarity score greater than threshold, add edge between authors
if ratio > sim_threshold:
sim_graph.add_edge(author1, author2, {"num_collabs":ratio, "sim_kw": matched_words})
# Indicate whether authors are actual coauthors
if col_graph.has_edge(author1, author2):
sim_graph[author1][author2]["areCoauthors"] = True
return sim_graph
def test_check_sim(self): kw1 = ["java", "python", "django"] kw2 = ["python", "programming", "django", "graphs"] result = tu.check_kw_sim(kw1, kw2) self.assertEqual(result[0], 0.5) self.assertEqual(result[1], [0, 2])
