R code for this note: http://f.briatte.org/r/collapsing-a-bipartite-co-occurrence-network

00_keywords.txt

name	keywords
Felicitas	Development, Corruption, Housing, Transportation, Religion
Lucile	Inequality, Migration, Race and Gender, Community/Neighborhood
Corentin	Inequality, Religion, Migration, Poverty, Gender
Charlotte	Energy, Environment, Smart Cities, Conflict, Gender
Alice	Environment, Gender, Migration, Urban Studies
Miranda	Politics, Gender, Human Rights
Margaux	Inequality, Urban Studies, Migration, Human Rights
Marina	Critical Theory, Social Inequality, Race and Gender
Cosima	Race and Gender and Sexuality, Spatial Inequality, Migration 
Isabelle	Slums, Poverty, Housing, Inequality
Elena	Race and Gender, Urban Inequality, Immigrant Integration, Collective Action
Gabriel	Environment, Collective Action, Urban Studies
Alexander	Transportation, Environmental Justice, Community Empowerment
Akhil	Tourism, Transportation, Politics
Gabriella	Multicultural Cohabitation, Environment, Radical Politics and Space Appropriation
Mohamed	Transport, Inequality, Poverty, Spatial Inequality
Lucien	Inequality, Poverty, Religious Segregation, Migration

01_packages.r

library(dplyr)
library(ggnetwork)
library(ggplot2)
library(igraph) # keep after ggnetwork
library(intergraph)
library(readr)
library(stringr)
library(tnet)

02_load.r

# students data
d <- read_tsv("keywords.txt") %>%
  arrange(name)

03_students.r

# student names
s <- d$name

04_keywords.r

# student keywords
k <- d$keywords %>%
  str_split(", | and ") %>%
  unlist %>%
  unique %>%
  sort

05_incidence_matrix.r

# incidence matrix
m <- sapply(d$keywords, function(x) {
  str_split(x, ", | and ") %>% unlist
}) %>%
  lapply(function(x) { as.integer(k %in% x) }) %>%
  unlist %>%
  matrix(ncol = length(k), byrow = TRUE)

06_name_dimensions.r

rownames(m) <- s # student names
colnames(m) <- k # student keywords

07_plot.r

# flatten out the network to a data frame
n <- ggnetwork(m)

# single out the primary mode (students)
n$student <- n$vertex.names %in% s

# plot two-mode network
ggplot(n, aes(x, y, xend = xend, yend = yend)) + 
  geom_edges(color = "grey50") + 
  geom_nodelabel(data = n[ !n$student, ],
                 aes(label = vertex.names),
                 color = "grey50", label.size = NA) +
  geom_nodelabel(data = n[ n$student, ],
                 aes(label = vertex.names),
                 color = "steelblue", fontface = "bold") +
  theme_blank()

08_collapse.r

# collapsed one-mode network, using student names
n <- m %*% t(m) %>%
  graph_from_adjacency_matrix(mode = "undirected", diag = FALSE, weighted = TRUE)

09_degree.r

# weighted degree
V(n)$degree <- as_adjacency_matrix(n, attr = "weight", sparse = FALSE) %>%
  degree_w %>%
  .[, 3]

10_clusters.r

V(n)$group <- cluster_louvain(n) %>%
  membership %>%
  as.character

11_plot.r

ggplot(ggnetwork(n, layout = igraph::with_kk()),
       aes(x, y, xend = xend, yend = yend)) +
  geom_edges(aes(alpha = weight)) +
  geom_nodelabel(aes(label = name, size = degree, color = group)) +
  scale_alpha_continuous(guide = FALSE) +
  scale_color_brewer(palette = "Set1", guide = FALSE) +
  scale_size_continuous(range = c(3, 6), guide = FALSE) +
  theme_blank()