+1 @TylerRinker 的建议
- 删除停用词和
- 仅使用名词使用 Jockers 的聚类方法（我在这里有一个工作示例。
您应该尝试的另一种选择是从二元组而不是一元组创建术语文档矩阵。如果您对短语感兴趣，二元组是一个好的开始。我在这里有一个工作示例。
这是一个将停用词删除与 bigrams 相结合的工作示例。通过此示例，您可以使用不同的参数值进行迭代，以获得对您来说最合理的聚类。
获取数据...
dat <- text <- structure(list(id = c("GHQ1", "GHQ2", "GHQ3", "GHQ4", "GHQ5",
"GHQ6", "GHQ7", "GHQ8", "GHQ9", "GHQ10", "GHQ11", "GHQ12", "GHQ13",
"GHQ14", "GHQ15", "GHQ16", "GHQ17", "GHQ18", "GHQ19", "GHQ20",
"GHQ21", "GHQ22", "GHQ23", "GHQ24", "CGMH9", "GHQ25", "GHQ26",
"GHQ27", "GHQ28", "GHQ29", "GHQ30", "GHQ31", "PARQ01A-P", "PARQ02A-P",
"PARQ03A-P", "PARQ04A-P", "PARQ05A-P", "PARQ06A-P", "PARQ07A-P",
"PARQ08A-P", "PARQ09A-P", "PARQ10A-P", "PARQ11A-P", "PARQ12A-P",
"PARQ13A-P", "PARQ14A-P", "PARQ15A-P", "PARQ16A-P", "PARQ17A-P",
"PARQ18A-P", "PARQ19A-P", "PARQ20A-P", "PARQ21A-P", "PARQ22A-P",
"PARQ23A-P", "PARQ24A-P", "PARQ25A-P", "PARQ26A-P", "PARQ27A-P",
"PARQ28A-P", "PARQ29A-P", "PARQ30A-P", "PARQ31A-P", "PARQ32A-P",
"PARQ33A-P", "PARQ34A-P", "PARQ35A-P", "PARQ36A-P", "PARQ37A-P",
"PARQ38A-P", "PARQ39A-P", "PARQ40A-P", "PARQ41A-P", "PARQ42A-P",
"PARQ43A-P", "PARQ44A-P", "PARQ45A-P", "PARQ46A-P", "PARQ47A-P",
"PARQ48A-P", "PARQ49A-P", "PARQ50A-P", "PARQ51A-P", "PARQ52A-P",
"PARQ53A-P", "PARQ54A-P", "PARQ55A-P", "PARQ56A-P", "PARQ57A-P",
"PARQ58A-P", "PARQ59A-P", "PARQ60A-P", "PARQ01A-C", "PARQ02A-C",
"PARQ03A-C", "PARQ04A-C", "PARQ05A-C", "PARQ06A-C", "PARQ07A-C",
"PARQ08A-C", "PARQ09A-C", "PARQ10A-C", "PARQ11A-C", "PARQ12A-C",
"PARQ13A-C", "PARQ14A-C", "PARQ15A-C", "PARQ16A-C", "PARQ17A-C",
"PARQ18A-C", "PARQ19A-C", "PARQ20A-C", "PARQ21A-C", "PARQ22A-C",
"PARQ23A-C", "PARQ24A-C", "PARQ25A-C", "PARQ26A-C", "PARQ27A-C",
"PARQ28A-C", "PARQ29A-C", "PARQ30A-C", "PARQ31A-C", "PARQ32A-C",
"PARQ33A-C", "PARQ34A-C", "PARQ35A-C", "PARQ36A-C", "PARQ37A-C",
"PARQ38A-C", "PARQ39A-C", "PARQ40A-C", "PARQ41A-C", "PARQ42A-C",
"PARQ43A-C", "PARQ44A-C", "PARQ45A-C", "PARQ46A-C", "PARQ47A-C",
"PARQ48A-C", "PARQ49A-C", "PARQ50A-C", "PARQ51A-C", "PARQ52A-C",
"PARQ53A-C", "PARQ54A-C", "PARQ55A-C", "PARQ56A-C", "PARQ57A-C",
"PARQ58A-C", "PARQ59A-C", "PARQ60A-C"), item = c("Been feeling unhappy or depressed",
"Been feeling reasonably happy, all things considered", "Feeling edgy and bad-tempered",
"Feel constantly under strain", "Found everything getting on top of you",
"Been feeling nervous and strung-up all the time", "found at times you couldn't do anything because your nerves were too bad",
"found everything getting on top of you", "thought of the possibility that you might make away with yourself",
"found that the idea of taking your own life kept coming into your mind?",
"found yourself withing you were dead and away from it all?",
"felt that life isn't worth living", "felt that life was entirely hopeless?",
"been able to enjoy your normal day-to-day activities", "been satisfied with the way you've carried out your task",
"felt that you are playing a useful part in things", "felt on the whole you were doing things well?",
"been feeling perfectly well and in good health", "been feeling in need of a good tonic",
"been feeling run down and out of sorts?", "felt that you are ill",
"been getting any pains in your head", "been getting a feeling of tightness or pressure in your head",
"been having hot or cold spells", "Do you feel you have physical problems because of stress?",
"Lost sleep over worry", "Had difficulty in staying asleep once you are off",
"felt capable of making decisions about things", "been taking longer over the things that you do",
"been managing to keep yourself busy and occupied", "been thinking of yourself as a worthless person",
"been getting scared or panicky for no good reason ", "You say nice things about your child",
"You nag or scold your child when (s)he is bad", "You ignore your child",
"You wonder if you really love your child", "You talk to your child about daily routines and plans, and listen to what (s)he has to say",
"You complain about your child to others when (s)he does not listen to you",
"You take an interest in your child", "You want your child to bring friends home, and you try to make things pleasant for them",
"You call your child names and make fun of him/her", "You ignore your child as long as (s)he does nothing to bother you",
"You yell at your child when you are angry", "You sit close with your child so that (s)he feels free to talk about important things",
"You are harsh with your child", "You enjoy having your child around you",
"You make your child feel proud when (s)he does well", "Your hit your child even when (s)he may not deserve it, like for small mistakes",
"You forget things you are supposed to do for your child", "You see your child as an annoyance",
"You praise your child to others", "You punish your child when you are angry",
"You make sure your child has the right kind of food to eat",
"You talk to your child in a warm and loving way", "You get angry easily at your child",
"You are too busy to answer your child's questions", "You hate/despise your child",
"You say nice things to your child when (s)he deserves it, such as when (s)he does well in school",
"You are irritable with your child", "You care about who your child's friends are",
"You are really interested in what your child does", "You say many unkind things to your child",
"You pay no attention to your child when (s)he asks for help",
"You think it is your child's own fault when (s)he is having trouble",
"You make your child feel wanted and needed", "You tell your child (s)he annoys you",
"You pay a lot of attention to your child", "You tell your child how proud you are of him/her when (s)he is good",
"You hurt your child's feelings", "You forget important things your child thinks you should remember",
"When your child misbehaves, you make him/her feel unloved",
"You make your child feel what (s)he does is important", "When your child does something wrong, you frighten or threaten him/her",
"You like to spend time with your child, for example you sit and laugh together",
"You try to help your child when (s)he is scared or upset", "When your child misbehaves, you shame him/her in front of his/her friends",
"You avoid your child's company", "You complain about your child",
"You care about what your child thinks, and encourage him/her to talk about it",
"You feel other children are better than your own child", "When you make plans, you take your child's thoughts into consideration",
"You let your child do things (s)he thinks are important, even if it is hard for you",
"When your child misbehaves, you compare him/her unfavorably with other children",
"You want to leave your child in someone else's care (for example, a neighbor or relative)",
"You let your child know (s)he is not wanted", "You are interested in the things your child does",
"You try to make your child feel better when (s)he is hurt or sick",
"You tell your child you are ashamed of him/her when (s)he misbehaves",
"You let your child know you love him/her", "You treat your child gently and with kindness",
"When your child misbehaves, you make him/her feel ashamed or guilty",
"You try to make your child happy", "Says nice things about you",
"Nags or scolds you when you are bad", "Ignores you", "Does not really love you",
"Talks to you about your plans and listens to what you have to say",
"Complains about you to others when you do not listen to him",
"Takes an interest in you", "Wants you to bring your friends home, and tries to make things pleasant for them",
"Calls you names, ridicules you, and makes fun of you", "Ignores you as long as you do nothing to bother him",
"Yells at you when he is angry", "Sits close with you so that you feel free to talk about important things",
"Treats you harshly", "Enjoys having you around him", "Make you feel proud when you do well",
"Hits you even when you do not deserve it, like for small mistakes",
"Forgets things he is supposed to do for you", "Sees you as an annoyance",
"Praises you to others", "Punishes you severely when he is angry",
"Makes sure you have the right kind of food to eat", "Talks to you in a warm and loving way",
"Gets angry at you easily", "Is too busy to answer your questions",
"Seems to hate / despise you", "Says nice things to you when you deserve them, such as when you do well in school",
"Gets mad quickly and picks on you", "Wants to know who your friends are",
"Is really interested in what you do", "Says many unkind things to you",
"Pays no attention when you ask for help", "Thinks it is your own fault when you are having trouble",
"Makes you feel wanted and needed", "Tells you that you annoy him",
"Pays a lot of attention to you", "Tells you how proud he is of you when you are good",
"Goes out of his way to hurt your feelings", "Forgets important things you think he should remember",
"Makes you feel unloved if you misbehave", "Makes you feel what you do is important",
"Frightens or threatens you when you do something wrong", "Likes to spend time with you, for example you sit and laugh together",
"Tries to help you when you are scared or upset", "Shames you in front of your friends when you misbehave",
"Tries to stay away from you", "Complains about you and talks about you behind your back",
"Cares about what you think, and likes you to talk about it",
"Feels other children are better than you are no matter what you do",
"Cares about what you would like when he makes plans", "Lets you do things you think are important, even if it is hard for him",
"Thinks other children behave better than you do", "Wants other people to take care of you (for example, a neighbor or relative)",
"Lets you know you are not wanted", "Is interested in the things you do",
"Shows concern and tries to make you feel better when you are hurt or sick",
"Tells you how ashamed he is when you misbehave", "Lets you know he loves you",
"Treats you gently and with kindness", "Makes you feel ashamed or guilty when you misbehave",
"Tries to make you happy")), .Names = c("id", "item"), row.names = c(NA,
152L), class = "data.frame")
现在制作一个二元组的 tdm，然后删除包含停用词的二元组......
library("RWeka")
library("tm")
library("Matrix")
BigramTokenizer <- function(x) NGramTokenizer(x, Weka_control(min = 2, max = 2))
x <- TermDocumentMatrix(Corpus(VectorSource(text$item)), control = list(tokenize = BigramTokenizer))
# little bit of regex to remove bigrams with stopwords in them, cf. https://stackoverflow.com/a/6947724/1036500
stpwrds <- paste(stopwords("en"), collapse = "|")
x$dimnames$Terms[!grepl(stpwrds, x$dimnames$Terms)]
[1] "cold spells" "else s" "feel free" "feel proud" [5] "feel unloved" "feels free" "lost sleep"
使用软件包随附的停用词库存列表快速测试删除二元组，tm结果表明它只给我们留下 8 个二元组！显然我们需要一个更小的停用词列表，所以让我们通过在这个特定语料库中查找最常见的词并删除它们来创建一个自定义列表。
# find freq words in corpus
x <- TermDocumentMatrix(Corpus(VectorSource(text$item)))
# arbitrary choice of 10 occurances = hi freq
mystopwords <- findFreqTerms(x, 10, Inf)
您应该尝试使用 lowfreq 值，我在尝试了几次后将其设置为 10，但其他值可能会更好。
# try to filter the bigrams again with custom stopword list
x <- TermDocumentMatrix(Corpus(VectorSource(text$item)), control = list(tokenize = BigramTokenizer))
# little bit of regex to remove bigrams with mystopwords in them, cf. https://stackoverflow.com/a/6947724/1036500
mystpwrds <- paste(mystopwords, collapse = "|")
# subset tdm to keep only bigrams remaining after mystopwords removed
x <- x[x$dimnames$Terms[!grepl(mystpwrds, x$dimnames$Terms)],]
y <- sparseMatrix( i=x$i, j=x$j, x=x$v, dimnames = dimnames(x) )
plot( hclust(dist(t(y))) )
但这有点难以阅读，所以让我们像这样打印组成员
hc <- hclust(dist(t(y)))
cutree(hc, k = 100)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
1 2 3 4 5 3 6 5 3 7 8 9 10 11 12 13
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
14 15 16 17 3 18 19 20 21 22 23 24 25 26 27 28
33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48
3 29 3 30 31 32 33 34 35 36 3 37 3 3 38 39
49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64
40 41 3 3 42 43 44 45 3 46 3 3 47 48 49 50
65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80
3 51 52 53 3 3 3 54 55 56 57 58 3 3 3 59
81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96
60 61 3 62 63 47 64 65 3 3 66 3 3 67 3 68
97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112
69 70 33 71 35 72 73 74 3 75 3 76 40 41 3 73
113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128
42 43 77 45 78 79 80 81 47 48 82 83 3 3 84 85
129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144
86 87 3 88 89 90 91 92 93 3 3 59 3 94 95 96
145 146 147 148 149 150 151 152
3 97 98 99 100 3 66 3
我们看到我们在不同的组中有第 145 行和第 149 行。这是否是一个好的答案很难说，因为你没有指定你想要的输出应该是什么。这就是为什么您获得接近投票的原因，很难从您的问题中推断出一个好的答案是什么（更具体地说，问题是您要求：“推荐或找到一个工具、图书馆或最喜欢的场外资源”，这会导致自以为是的答案），SO 人群似乎更喜欢所需输出的具体示例。您可以在新的数据科学堆栈交换站点上尝试您的问题。
无论如何，希望你现在有更多的想法和更多的旋钮来处理你对数据的探索。如果您遇到与此相关的特定编程问题，请随时提出另一个问题。