单个术语上的多个共现集群答案

【问题标题】：multiple co-occurence clusters on single term单个术语上的多个共现集群
【发布时间】：2018-11-12 07:11:33
【问题描述】：

我有一个关键术语至少出现一次的语料库。由此我制作了看起来很像这样的fcm。

txts <- c("a a a b b c", "a a c e", "a c b e f g", "e d j b", "b g k l", "b a a g l", "e c b j k l", "b g w m")
total <- fcm(txts, context = "document", count = "frequency")

Feature co-occurrence matrix of: 12 by 12 features.
12 x 12 sparse Matrix of class "fcm"
    features
features a b c e f g d j k l w m
   a 5 9 6 3 1 3 0 0 0 2 0 0
   b 0 1 4 3 1 4 1 2 2 3 1 1
   c 0 0 0 3 1 1 0 1 1 1 0 0
   e 0 0 0 0 1 1 1 2 1 1 0 0
   f 0 0 0 0 0 1 0 0 0 0 0 0
   g 0 0 0 0 0 0 0 0 1 2 1 1
   d 0 0 0 0 0 0 0 1 0 0 0 0
   j 0 0 0 0 0 0 0 0 1 1 0 0
   k 0 0 0 0 0 0 0 0 0 2 0 0
   l 0 0 0 0 0 0 0 0 0 0 0 0
   w 0 0 0 0 0 0 0 0 0 0 0 1
   m 0 0 0 0 0 0 0 0 0 0 0 0

据此，我想找到“b”周围的不同簇。

着眼于缩放，我的实际 fcm 有 239104369 个元素，大小为 1.2GB。

前 10 个特征的矩阵如下所示

Feature co-occurrence matrix of: 10 by 10 features.
10 x 10 sparse Matrix of class "fcm"
           features
features        international monetary    fund development association bolivia assessment interim  poverty reduction
international       2885797  1345055 3340282    12013377      857864  199985     605036  202117  3996710   1319199
monetary                  0   227329  973979     2326677      234565   39802      93927   65773   884341    330250
fund                      0        0 1766657     6530594      621315   99900     355415  204229  2534382    927737
development               0        0       0    20054398     1683896  485906    2235294  406575 13674085   4091506
association               0        0       0           0      122947   25954      87756   47038   580721    204144
bolivia                   0        0       0           0           0   26062      35164    5336   254924     71428
assessment                0        0       0           0           0       0     203933   24196  1420850    377398
interim                   0        0       0           0           0       0          0   20595   172870     67705
poverty                   0        0       0           0           0       0          0       0  9131869   4026961
reduction                 0        0       0           0           0       0          0       0        0    642944

我的目标是可视化围绕关键术语 (https://bost.ocks.org/mike/miserables/) 的集群并从中创建术语列表。

https://www.r-bloggers.com/turning-keywords-into-a-co-occurrence-network/

https://www.r-bloggers.com/collapsing-a-bipartite-co-occurrence-network/

Co occurrence plot in R

在我的搜索中，我也偶然发现了 cooccurNet 包，但我不知道如何熟练使用它。 https://cran.r-project.org/web/packages/cooccurNet/index.html

【问题讨论】：

标签： r nlp cluster-analysis quanteda

【解决方案1】：

quanteda 有textstat_simil()，它返回一个用于层次聚类的dist 对象。此函数只需要一个 DFM，但可以使用 as.dfm() 将 FCM 转换为对象。

require(quanteda)
txt <- c("a a a b b c", "a a c e", "a c b e f g", "e d j b", "b g k l", "b a a g l", "e c b j k l", "b g w m")
dmt <- dfm(txt)
# dmt <- dfm_trim(dmt, min_termfreq = 10) # you might need this to reduce the size of fcm
fmt <- fcm(dmt, context = "document")

dist <- textstat_simil(as.dfm(fmt), margin = "features")
tree <- hclust(dist)
cutree(tree, 2)

【讨论】：