Takeshi Arabiki, profile picture
Uploaded byTakeshi Arabiki
23,669 views

RではじめるTwitter解析

This document discusses analyzing Twitter data from the user @a_bicky using R. It extracts over 3,200 tweets from the user's timeline using the twitteR package. The tweets are transformed into a data frame with variables like text, date, and source. The data is then summarized using the reshape2 and ggplot2 packages to calculate metrics like average text length by day of week, month, and source. Frequency tables and heat maps are generated to explore patterns in the Twitter data over time.

Embed presentation

Downloaded 304 times
1 / 131
2 / 131
3 / 131
4 / 131
5 / 131
6 / 131
7 / 131
8 / 131
9 / 131
10 / 131
11 / 131
12 / 131
13 / 131
14 / 131
15 / 131
16 / 131
17 / 131
18 / 131
19 / 131
20 / 131
21 / 131
22 / 131
23 / 131
24 / 131
25 / 131
26 / 131
27 / 131
28 / 131
29 / 131
30 / 131
31 / 131
32 / 131
33 / 131
34 / 131
35 / 131
36 / 131
37 / 131
38 / 131
39 / 131
40 / 131
41 / 131
42 / 131
43 / 131
44 / 131
45 / 131
46 / 131
47 / 131
48 / 131
49 / 131
50 / 131
51 / 131
52 / 131
53 / 131
54 / 131
55 / 131
56 / 131
57 / 131
58 / 131
59 / 131
60 / 131
61 / 131
62 / 131
63 / 131
64 / 131
65 / 131
66 / 131
67 / 131
68 / 131
69 / 131
70 / 131
71 / 131
72 / 131
73 / 131
74 / 131
75 / 131
76 / 131
77 / 131
78 / 131
79 / 131
80 / 131
81 / 131
82 / 131
83 / 131
84 / 131
85 / 131
86 / 131
87 / 131
88 / 131
89 / 131
90 / 131
91 / 131
92 / 131
93 / 131
94 / 131
95 / 131
96 / 131
97 / 131
98 / 131
99 / 131
100 / 131
101 / 131
102 / 131
103 / 131
104 / 131
105 / 131
106 / 131
107 / 131
108 / 131
109 / 131
110 / 131
111 / 131
112 / 131
113 / 131
114 / 131
115 / 131
116 / 131
117 / 131
118 / 131
119 / 131
120 / 131
121 / 131
122 / 131
123 / 131
124 / 131
125 / 131
126 / 131
127 / 131
128 / 131
129 / 131
130 / 131
131 / 131
R Twitter R 2011 (2011/11/26) @a_bicky
• Takeshi Arabiki ‣ ‣ Twitter & : @a_bicky & id:a_bicky • R • http://d.hatena.ne.jp/a_bicky/
R Osaka.R #4 Tokyo.R #16 Tsukuba.R #9 http://www.slideshare.net/abicky/twitterr http://www.slideshare.net/abicky/r-9034336 http://www.slideshare.net/abicky/r-10128090
※
Twitter
Mentionmapp
Mentionmapp
Mentionmapp
http://twilog.org/ http://twitraq.userlocal.jp/ http://whotwi.com/ http://tweetstats.com/
http://twilog.org/ http://twitraq.userlocal.jp/ R http://whotwi.com/ http://tweetstats.com/
Twitter • • reshape2 • ggplot2 •
Twitter • • reshape2 • ggplot2 •
twitteR twitteR > library(twitteR) # twitteR > # (twitteR 0.99.15 ) > Sys.setlocale("LC_TIME", "C") [1] "C" > # @a_bicky 3,200 RT > statuses <- userTimeline("a_bicky", n = 3200)
status > # R5 > ls.str(statuses[[1]]) created : POSIXct[1:1], format: "2011-11-23 22:16:24" favorited : logi FALSE ↑ UTC id : chr "139467359571296256" initFields : Formal class 'refMethodDef' [package "methods"] with 5 slots initialize : Formal class 'refMethodDef' [package "methods"] with 5 slots replyToSID : chr(0) replyToSN : chr(0) replyToUID : chr(0) screenName : chr "a_bicky" ! Twitter statusSource : chr "<a href="http://sites.google.com/site/ yorufukurou/" rel="nofollow">YoruFukurou</a>" text : chr " " truncated : logi FALSE ↑
> statusDF <- twListToDF(statuses) > str(statusDF, vec.len = 1) 'data.frame': 3159 obs. of 10 variables: $ text : chr " " ... ↑ $ favorited : logi FALSE ... $ replyToSN : logi NA ... $ created : POSIXct, format: "2011-11-23 22:16:24" ... $ truncated : logi FALSE ... ↑ UTC $ replyToSID : logi NA ... $ id : chr "139467359571296256" ... $ replyToUID : logi NA ... $ statusSource: chr "<a href="http://sites.google.com/ site/yorufukurou/" rel="nofollow">YoruFukurou</a>" ... $ screenName : chr "a_bicky" ...
> wday.abb <- c("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun") > statusDF <- within(statusDF, { + attr(created, "tzone") <- "Asia/Tokyo" # JST + statusSource <- factor(gsub("<a .*?>(.*?)</a>", "1", statusSource)) # HTML + date <- factor(format(created, "%Y-%m-%d")) # + hour <- NULL; month <- NULL; year <- NULL; wday <- NULL + with(as.POSIXlt(created), { + hour <<- factor(hour) # + month <<- factor(mon + 1) # + year <<- factor(year + 1900) # + wday <<- factor((wday + 6) %% 7, labels = wday.abb) # + }) + textLength <- nchar(text) # + # , URL, + cleanText <- removeSpecialStr(text) + cleanTextLength <- nchar(cleanText) # URL + })
> # Twitter > topSources <- names(head(sort(table(statusDF$statusSource), decreasing = TRUE), 5)) > statusDF <- within(statusDF, { + statusSource <- as.character(statusSource) + statusSource[!statusSource %in% topSources] <- "other" + # + statusSource <- factor(statusSource, levels = names(sort(table (statusSource), dec = TRUE))) + })
Twitter • • reshape2 • ggplot2 •
reshape2
Excel 9 11 ”Twitter for iPhone”, ”YoruFukurou” Sat Mon 12 23
reshape2 > library(reshape2) > acast(melt(statusDF, id.vars = c("statusSource", "wday", "month", "hour"), + measure.vars = c("textLength")), + month + statusSource ~ wday, mean, + subset = .(statusSource %in% c("YoruFukurou", "Twitter for iPhone") + & month %in% 9:11 & hour %in% 12:23 + & wday %in% c("Mon", "Sat", "Sun"))) Mon Sat Sun 9_YoruFukurou 43 42.13333 54.76471 9_Twitter for iPhone 16 27.70000 20.50000 10_YoruFukurou 61 41.70175 56.98333 10_Twitter for iPhone NaN 27.00000 24.50000 11_YoruFukurou 35 41.08197 57.32609 11_Twitter for iPhone NaN NaN 32.00000
reshape2 > library(reshape2) > acast(melt(statusDF, id.vars = c("statusSource", "wday", "month", "hour"), + measure.vars = c("textLength")), + month + statusSource ~ wday, mean, + subset = .(statusSource %in% c("YoruFukurou", "Twitter for iPhone") + & month %in% 9:11 & hour %in% 12:23 + & wday %in% c("Mon", "Sat", "Sun"))) Mon Sat Sun 9_YoruFukurou 43 42.13333 54.76471 9_Twitter for iPhone 16 27.70000 20.50000 10_YoruFukurou 61 41.70175 56.98333 10_Twitter for iPhone NaN 27.00000 24.50000 11_YoruFukurou 35 41.08197 57.32609 11_Twitter for iPhone NaN NaN 32.00000
reshape2 > library(reshape2) > acast(melt(statusDF, id.vars = c("statusSource", "wday", "month", "hour"), + measure.vars = c("textLength")), + month + statusSource ~ wday, mean, + subset = .(statusSource %in% c("YoruFukurou", "Twitter for iPhone") + & month %in% 9:11 & hour %in% 12:23 + & wday %in% c("Mon", "Sat", "Sun"))) Mon Sat Sun 9_YoruFukurou 43 42.13333 54.76471 9_Twitter for iPhone 16 27.70000 20.50000 10_YoruFukurou 61 41.70175 56.98333 10_Twitter for iPhone NaN 27.00000 24.50000 11_YoruFukurou 35 41.08197 57.32609 11_Twitter for iPhone NaN NaN 32.00000 R
reshape2 melt melt cast melt cast > mstatus <- melt(statusDF, + id.vars = c("statusSource", "wday", "year", "month", "hour", "date"), + measure.vars = c("textLength", "cleanTextLength")) > mstatus[3157:3162, ] statusSource wday year month hour date variable value 3157 web Sun 2011 3 20 2011-03-13 textLength 72 3158 web Sun 2011 3 16 2011-03-13 textLength 24 3159 web Sun 2011 3 14 2011-03-13 textLength 82 3160 YoruFukurou Wed 2011 11 1 2011-11-23 cleanTextLength 87 3161 YoruFukurou Wed 2011 11 1 2011-11-23 cleanTextLength 14 3162 YoruFukurou Wed 2011 11 1 2011-11-23 cleanTextLength 21 id
reshape2 cast cast formula fun.aggregate > args(acast) # array acast function (data, formula, fun.aggregate = NULL, ..., margins = NULL, subset = NULL, fill = NULL, drop = TRUE, value_var = guess_value(data)) NULL > args(dcast) # data.frame dcast function (data, formula, fun.aggregate = NULL, ..., margins = NULL, subset = NULL, fill = NULL, drop = TRUE, value_var = guess_value(data)) NULL formula ... . acast hoge ~ fuga ~ piyo ※dcast 1 hoge ~ fuga + piyo
> # > acast(mstatus, . ~ wday, length, subset = .(variable == "textLength")) ↑ cleanTextLength
> # > acast(mstatus, . ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun [1,] 408 360 258 294 334 801 704 >
> # > acast(mstatus, . ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun [1,] 408 360 258 294 334 801 704 >
> # > acast(mstatus, . ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun [1,] 408 360 258 294 334 801 704 > # > acast(mstatus, hour ~ wday, length, subset = .(variable == "textLength"))
> # > acast(mstatus, . ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun [1,] 408 360 258 294 334 801 704 > # > acast(mstatus, hour ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun 0 65 69 26 46 46 49 40 1 48 19 11 15 27 44 37 2 31 24 6 16 17 23 17 3 27 19 4 11 14 17 10 4 4 15 1 7 4 5 7 5 5 11 1 4 3 4 5 6 4 14 3 6 9 8 1
> # > # > acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength"))
> # > # > acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength")) Mon_3 Mon_4 Mon_5 Mon_6 Mon_7 Mon_8 Mon_9 Mon_10 Mon_11 Tue_3 Tue_4 0 3 4 13 3 1 10 7 15 9 4 2 1 0 0 1 0 1 9 16 12 9 1 0 2 2 0 0 0 2 7 6 7 7 2 0
> # > # > acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength")) Mon_3 Mon_4 Mon_5 Mon_6 Mon_7 Mon_8 Mon_9 Mon_10 Mon_11 Tue_3 Tue_4 0 3 4 13 3 1 10 7 15 9 4 2 1 0 0 1 0 1 9 16 12 9 1 0 2 2 0 0 0 2 7 6 7 7 2 0 > # 3 > acast(mstatus, hour ~ wday ~ month, length, subset = .(variable == "textLength"))
> # > # > acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength")) Mon_3 Mon_4 Mon_5 Mon_6 Mon_7 Mon_8 Mon_9 Mon_10 Mon_11 Tue_3 Tue_4 0 3 4 13 3 1 10 7 15 9 4 2 1 0 0 1 0 1 9 16 12 9 1 0 2 2 0 0 0 2 7 6 7 7 2 0 > # 3 > acast(mstatus, hour ~ wday ~ month, length, subset = .(variable == "textLength")) , , 3 Mon Tue Wed Thu Fri Sat Sun 0 3 4 1 0 1 6 4 1 0 1 3 0 0 0 1
> # > # > acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength")) Mon_3 Mon_4 Mon_5 Mon_6 Mon_7 Mon_8 Mon_9 Mon_10 Mon_11 Tue_3 Tue_4 0 3 4 13 3 1 10 7 15 9 4 2 1 0 0 1 0 1 9 16 12 9 1 0 2 2 0 0 0 2 7 6 7 7 2 0 > # 3 > acast(mstatus, hour ~ wday ~ month, length, subset = .(variable == "textLength")) , , 3 3 Mon Tue Wed Thu Fri Sat Sun 0 3 4 1 0 1 6 4 1 0 1 3 0 0 0 1
Twitter reshape2 1 > # > dcast(mstatus, statusSource ~ ., + function(x) list(c(mean = mean(x), sd = sd(x))), + fill = list(c(mean = NaN, sd = NA)), ← + subset = .(variable == "textLength"))
Twitter reshape2 1 > # > dcast(mstatus, statusSource ~ ., + function(x) list(c(mean = mean(x), sd = sd(x))), + fill = list(c(mean = NaN, sd = NA)), ← + subset = .(variable == "textLength")) statusSource NA 1 YoruFukurou 47.51462, 32.57973 2 web 57.02720, 36.33534 3 Twitter for iPhone 33.42342, 23.06466 4 Twitter for Android 28.49048, 20.08457 5 Hatena 80.00000, 25.94212 6 other 52.58621, 33.12180 >
Twitter reshape2 1 > # > dcast(mstatus, statusSource ~ ., + function(x) list(c(mean = mean(x), sd = sd(x))), + fill = list(c(mean = NaN, sd = NA)), ← + subset = .(variable == "textLength")) statusSource NA 1 YoruFukurou 47.51462, 32.57973 2 web 57.02720, 36.33534 3 Twitter for iPhone 33.42342, 23.06466 4 Twitter for Android 28.49048, 20.08457 5 Hatena 80.00000, 25.94212 6 other 52.58621, 33.12180 >
> # t > pc <- unlist(subset(statusDF, + statusSource %in% c("YoruFukurou", "web"), + textLength)) > sp <- unlist(subset(statusDF, + grepl("(iPhone|Android)", statusSource), + textLength)) > t.test(sp, pc, var.equal = FALSE) Welch Two Sample t-test !! data: sp and pc t = -15.7921, df = 1588.246, p-value < 2.2e-16 alternative hypothesis: true difference in means is not equal to 0 95 percent confidence interval: -19.85334 -15.46645 sample estimates: mean of x mean of y 31.83945 49.49935
> # t > pc <- unlist(subset(statusDF, + statusSource %in% c("YoruFukurou", "web"), + textLength)) > sp <- unlist(subset(statusDF, + grepl("(iPhone|Android)", statusSource), + textLength)) > t.test(sp, pc, var.equal = FALSE) Welch Two Sample t-test !! data: sp and pc t = -15.7921, df = 1588.246, p-value < 2.2e-16 alternative hypothesis: true difference in means is not equal to 0 95 percent confidence interval: -19.85334 -15.46645 sample estimates: mean of x mean of y 31.83945 49.49935
> extractScreenNames <- function(text, strict = TRUE) { + if (strict) { + # Twitter screen_name + regex <- "(?:(?<!w)([@ ])((?>w+))(?![@ ])|[sS])" + } else { + # hoge@example.com + regex <- "(?:([@ ])(w+)|[sS])" + } + screenNames <- gsub(regex, "12", text, perl = TRUE) + unique(unlist(strsplit(substring(screenNames, 2), "[@ ]"))) + } > screenNames <- unlist(lapply(statusDF$text, extractScreenNames)) > head(sort(table(screenNames), decreasing = TRUE), 10) # Top 10 screenNames naopr __gfx__ hirota_inoue mandy_44 ask_a_lie 105 85 51 47 40 ken_nishi nokuno yokkuns JinJin0613 kanon19_rie 39 39 33 20 20
Twitter • • reshape2 • ggplot2 •
ggplot2
ggplot2 plot(statusDF$wday, col = "blue") ggplot2 qplot(wday, data = statusDF, fill = I("blue"), alpha = I(0.7), xlab = "", ylab = "")
ggplot2 qplot(wday, data = statusDF, fill = statusSource, xlab = "", ylab = "")
ggplot2 qplot(wday, data = statusDF, facets = ~ statusSource, fill = I("blue"), alpha = I(0.7), xlab = "", ylab = "") qplot(wday, data = statusDF, fill = statusSource, xlab = "", ylab = "")
ggplot2 qplot(wday, data = statusDF, facets = ~ statusSource, fill = I("blue"), alpha = I(0.7), xlab = "", ylab = "") qplot(wday, data = statusDF, fill = statusSource, xlab = "", ylab = "")
qplot ggplot2 > args(qplot) function (x, y = NULL, z = NULL, ..., data, facets = . ~ ., margins = FALSE, geom = "auto", stat = list(NULL), position = list(NULL), xlim = c(NA, NA), ylim = c(NA, NA), log = "", main = NULL, xlab = deparse(substitute(x)), ylab = deparse(substitute(y)), asp = NA) NULL
qplot geom geom area: bar: histogram: line: point:
qplot geom geom area: bar: histogram: line: point: qplot(as.integer(wday), data = statusDF, geom = "area", stat = "bin", fill = statusSource, xlab = "", ylab = "", binwidth = 1)
qplot geom geom area: bar: histogram: line: point: qplot(wday, data = statusDF, geom = "bar", stat = "bin", fill = statusSource, xlab = "", ylab = "")
qplot geom geom area: bar: histogram: line: point: qplot(as.integer(wday), data = statusDF, geom = "line", stat = "bin", colour = statusSource, xlab = "", ylab = "", binwidth = 1)
qplot geom geom area: bar: histogram: line: point: qplot(wday, data = statusDF, geom = "point", stat = "bin", colour = statusSource, xlab = "", ylab = "")
qplot position position geom dodge : fill : 1 jitter : stack :
qplot position position geom dodge : fill : 1 jitter : stack : qplot(wday, data = statusDF, fill = statusSource, position = "dodge", xlab = "", ylab = "")
qplot position position geom dodge : fill : 1 jitter : stack : qplot(wday, data = statusDF, fill = statusSource, position = "fill", xlab = "", ylab = "")
qplot position position geom dodge : fill : 1 jitter : stack : qplot(wday, data = statusDF, fill = statusSource, position = "jitter", xlab = "", ylab = "")
qplot position position geom dodge : fill : 1 jitter : stack : qplot(wday, data = statusDF, fill = statusSource, position = "stack", xlab = "", ylab = "")
qplot facets facets geom ~ : 1 ~ 2: 1, 2 ※reshape2 1 ~ 2 + 3
qplot facets facets geom ~ : 1 ~ 2: 1, 2 ※reshape2 1 ~ 2 + 3 qplot(wday, data = statusDF, xlab = "", ylab = "", facets = ~ statusSource)
qplot facets facets geom ~ : 1 ~ 2: 1, 2 ※reshape2 1 ~ 2 + 3 qplot(wday, data = statusDF, xlab = "", ylab = "", facets = month ~ statusSource)
qplot alpha : colour (color) : fill : linetype : size : colour, fill, linetype statusSource fill = I("blue") I (AsIs)
qplot alpha : colour (color) : fill : linetype : size : qplot(wday, data = statusDF, xlab = "", ylab = "", alpha = as.integer(wday))
qplot alpha : colour (color) : fill : linetype : size : qplot(wday, data = statusDF, xlab = "", ylab = "", colour = statusSource)
qplot alpha : colour (color) : fill : linetype : size : qplot(wday, data = statusDF, xlab = "", ylab = "", fill = statusSource)
qplot alpha : colour (color) : fill : linetype : size : qplot(wday, data = statusDF, xlab = "", ylab = "", linetype = statusSource, colour = statusSource)
whotwi http://whotwi.com/
whotwi http://whotwi.com/
whotwi > # Twitter > # melt cast xtabs > cnt <- as.data.frame(xtabs(~ hour + wday + statusSource, statusDF)) > head(cnt, 3) hour wday statusSource Freq 1 0 Mon YoruFukurou 48 2 1 Mon YoruFukurou 38 3 2 Mon YoruFukurou 25
whotwi > # Twitter > # melt cast xtabs > cnt <- as.data.frame(xtabs(~ hour + wday + statusSource, statusDF)) > head(cnt, 3) hour wday statusSource Freq 1 0 Mon YoruFukurou 48 2 1 Mon YoruFukurou 38 3 2 Mon YoruFukurou 25 > freqSources <- by(cnt, cnt[c("hour", "wday")], function(df) { + # + freqSource <- with(df, statusSource[order(Freq, decreasing = TRUE) [1]]) + cbind(df[1, c("hour", "wday")], freqSource) + }) > freqSources <- do.call(rbind, freqSources) > head(freqSources, 3) hour wday freqSource 1 0 Mon YoruFukurou 2 1 Mon YoruFukurou 3 2 Mon YoruFukurou
whotwi > # > cntSum <- as.data.frame(xtabs(Freq ~ hour + wday, cnt)) > head(cntSum, 3) hour wday Freq 1 0 Mon 65 2 1 Mon 48 3 2 Mon 31
whotwi > # > cntSum <- as.data.frame(xtabs(Freq ~ hour + wday, cnt)) > head(cntSum, 3) hour wday Freq 1 0 Mon 65 2 1 Mon 48 3 2 Mon 31 > # > data <- merge(cntSum, freqSources) > # > data$wday <- factor(data$wday, levels = rev(levels(data$wday))) > # > data$Freq <- log2(data$Freq)
whotwi > # > cntSum <- as.data.frame(xtabs(Freq ~ hour + wday, cnt)) > head(cntSum, 3) hour wday Freq 1 0 Mon 65 2 1 Mon 48 3 2 Mon 31 > # > data <- merge(cntSum, freqSources) > # > data$wday <- factor(data$wday, levels = rev(levels(data$wday))) > # > data$Freq <- log2(data$Freq) > p <- qplot(hour, wday, data = data, xlab = "", ylab = "", + geom = "point", colour = freqSource, size = Freq) > p # print(p)
whotwi > # > cntSum <- as.data.frame(xtabs(Freq ~ hour + wday, cnt)) > head(cntSum, 3) hour wday Freq 1 0 Mon 65 2 1 Mon 48 3 2 Mon 31 > # > data <- merge(cntSum, freqSources) > # > data$wday <- factor(data$wday, levels = rev(levels(data$wday))) > # > data$Freq <- log2(data$Freq) > p <- qplot(hour, wday, data = data, xlab = "", ylab = "", + geom = "point", colour = freqSource, size = Freq) > p # print(p)
whotwi
whotwi
whotwi > # whotwi theme > theme_whotwi <- function() { + opts( # + panel.background = theme_rect(fill = NA, colour = NA), + # + legend.key = theme_rect(fill = NA, colour = NA), + # + axis.ticks = theme_segment(colour = NA)) + } > p2 <- p + theme_whotwi() + scale_size(legend = FALSE) + scale_colour_hue(name = "") > p2
whotwi > # whotwi theme > theme_whotwi <- function() { + opts( # + panel.background = theme_rect(fill = NA, colour = NA), + # + legend.key = theme_rect(fill = NA, colour = NA), + # + axis.ticks = theme_segment(colour = NA)) + } > p2 <- p + theme_whotwi() + scale_size(legend = FALSE) + scale_colour_hue(name = "") > p2
whotwi
whotwi
whotwi
whotwi PC
whotwi PC PC
Twitter • • reshape2 • ggplot2 •
TweetSentiments
TweetSentiments R
1. RMeCab 2. 3.
RMeCab MeCab R > library(RMeCab) > (docDF(data.frame(" "), column = 1, type = 1)) number of extracted terms = 5 now making a data frame. wait a while! TERM POS1 POS2 Row1 1 1 2 1 3 1 4 2 5 2
http://www.lr.pi.titech.ac.jp/~takamura/pndic_ja.html : : :1 : : :0.999995 : : :0.999979 : : :0.999979 : : :0.999645 : : :0.999486 : : :0.999314 ...
> # > pndic <- read.table("http://www.lr.pi.titech.ac.jp/~takamura/pubs/ pn_ja.dic", + sep = ":", + col.names = c("term", "kana", "pos", "value"), + colClasses = c("character", "character", "factor", "numeric"), + fileEncoding = "Shift_JIS") > # > # > pndic2 <- aggregate(value ~ term + pos, pndic, mean)
> # pndic > pos <- unique(pndic2$pos) > tweetDF <- docDF(statusDF, column = "cleanText", type = 1, pos = pos) number of extracted terms = 7164 now making a data frame. wait a while! > tweetDF[2900:2904, 1:5] TERM POS1 POS2 Row1 Row2 2900 0 0 2901 0 0 2902 0 0 2903 0 0 2904 0 0 > # pndic > tweetDF <- subset(tweetDF, TERM %in% pndic2$term) > # > tweetDF <- merge(tweetDF, pndic2, by.x = c("TERM", "POS1"), by.y = c ("term", "pos"))
> # > score <- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117
> # > score <- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117
> # > score <- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765
> # > score <- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765
> # > score <- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765 > # > sum(score == 0) [1] 277
> # > score <- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765 > # > sum(score == 0) [1] 277
> # > score <- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765 > # > sum(score == 0) [1] 277
> table(ifelse(pndic$value > 0, "positive", + ifelse(pndic$value == 0, "neutral", "negative"))) negative neutral positive 49983 20 5122
> table(ifelse(pndic$value > 0, "positive", + ifelse(pndic$value == 0, "neutral", "negative"))) negative neutral positive 49983 20 5122
> m <- mean(score) > # > tweetType <- factor(ifelse(score > m, "positive", + ifelse(score == m, "neutral", "negative")), + levels = c("positive", "neutral", "negative")) > table(tweetType) tweetType positive neutral negative 1912 0 1247
> statusDF$tweetType <- droplevels(tweetType) > # > qplot(month, data = statusDF, + geom = "bar", fill = tweetType, position = "fill")
> statusDF$tweetType <- droplevels(tweetType) > # > qplot(month, data = statusDF, + geom = "bar", fill = tweetType, position = "fill")
> statusDF$tweetType <- droplevels(tweetType) > # > qplot(month, data = statusDF, + geom = "bar", fill = tweetType, position = "fill")
> statusDF$tweetType <- droplevels(tweetType) > # > qplot(month, data = statusDF, + geom = "bar", fill = tweetType, position = "fill")
twitteR • RJSONIO • • ID status ID • fav favorited TRUE • truncated TRUE • DM • status character factor
twitteR • RJSONIO • • ID status ID • fav favorited TRUE • truncated TRUE • DM • status character factor
OAuth ” ” twitteR -
• twitteR • reshape2 R • ggplot2 • RMeCab R
• twitteR • reshape2 R • ggplot2 • RMeCab R • PC •
https://github.com/abicky/rjpusers2011_abicky
status > statuses[[1]]$text [1] " " > statuses[[1]]$getText() # [1] " " > # > statuses[[1]]$text <- " " > statuses[[1]]$getText() [1] " " > statuses[[1]]$setText("ggrks") # > statuses[[1]]$getText() [1] "ggrks" > # > statuses[[1]]$getCreated() [1] "2011-11-23 22:16:24 UTC"
removeSpecialStr removeSpecialStr <- function(text) { removeURL(removeHashTag(removeScreenName(text))) }
removeScreenName removeScreenName <- function(text, strict = TRUE) { if (strict) { regex <- "(?<!w)[@ ](?>w+)(?![@ ])" } else { regex <- "[@ ]w+" } gsub(regex, "", text, perl = TRUE) }
removeURL removeURL <- function(text, strict = TRUE) { if (strict) { regex <- "(?<![-.w#@=!'"/])https?://(?:[^:]+:. +@)?(?:[0-9A-Za-z][-0-9A-Za-z]*(?<!-).)+[A-za-z]+(?:/[- w#%=+,.?!&~]*)*" } else { regex <- "https?://[-w#%=+,.?!&~/]+" } gsub(regex, "", text, perl = TRUE) }
removeHashTag removeHashTag <- function(text, strict = TRUE) { delimiters <- "s,.u3000-u3002uFF01uFF1F" # cf. http://nobu666.com/2011/07/13/914.html validJa <- "u3041-u3094u3099-u309Cu30A1-u30FA u30FCu3400-uD7A3uFF10-uFF19uFF21-uFF3AuFF41-uFF5A uFF66-uFF9E" if (strict) { regex <- sprintf("(^|[%s])(?:([# ](?>[0-9]+)(?! w))|[# ][w%s]+)", delimiters, validJa, validJa) } else { regex <- sprintf("[# ][^%s]+", delimiters) } gsub(regex, "12", text, perl = TRUE) }

More Related Content

PDF
twitteRで快適Rライフ！
byTakeshi Arabiki
 
PPTX
Coding Horrors
byMark Baker
 
PDF
ZeroMQ Is The Answer: DPC 11 Version
byIan Barber
 
PDF
How to stand on the shoulders of giants
byIan Barber
 
PDF
Palestra sobre Collections com Python
bypugpe
 
PDF
ZeroMQ: Messaging Made Simple
byIan Barber
 
PDF
Functional Pe(a)rls - the Purely Functional Datastructures edition
byosfameron
 
PPT
Functional Pe(a)rls version 2
byosfameron
 
twitteRで快適Rライフ！
byTakeshi Arabiki
 
Coding Horrors
byMark Baker
 
ZeroMQ Is The Answer: DPC 11 Version
byIan Barber
 
How to stand on the shoulders of giants
byIan Barber
 
Palestra sobre Collections com Python
bypugpe
 
ZeroMQ: Messaging Made Simple
byIan Barber
 
Functional Pe(a)rls - the Purely Functional Datastructures edition
byosfameron
 
Functional Pe(a)rls version 2
byosfameron
 

What's hot

PDF
Is Haskell an acceptable Perl?
byosfameron
 
PDF
Byterun, a Python bytecode interpreter - Allison Kaptur at NYCPython
byakaptur
 
PDF
Clustering com numpy e cython
byAnderson Dantas
 
PPTX
Data monsters probablistic data structures
byGreenM
 
PDF
Allison Kaptur: Bytes in the Machine: Inside the CPython interpreter, PyGotha...
byakaptur
 
PDF
dplyr and torrents from cpasbien
byRomain Francois
 
PPTX
The groovy puzzlers (as Presented at JavaOne 2014)
byGroovyPuzzlers
 
PDF
Malcon2017
byAndriy Brukhovetskyy
 
PDF
M12 random forest-part01
byRaman Kannan
 
PDF
Bytes in the Machine: Inside the CPython interpreter
byakaptur
 
PDF
Spark DataFrames for Data Munging
by(Susan) Xinh Huynh
 
PDF
M09-Cross validating-naive-bayes
byRaman Kannan
 
PPTX
第二讲 Python基礎
byjuzihua1102
 
PPTX
第二讲 预备-Python基礎
byanzhong70
 
PDF
"A 1,500 line (!!) switch statement powers your Python!" - Allison Kaptur, !!...
byakaptur
 
PDF
(gentle (introduction Clojure))
byGuy Taylor
 
PPT
An Elephant of a Different Colour: Hack
byVic Metcalfe
 
PDF
PLOTCON NYC: Behind Every Great Plot There's a Great Deal of Wrangling
byPlotly
 
PDF
Top 10 php classic traps
byDamien Seguy
 
PDF
MongoUK - PHP Development
byBoxed Ice
 
Is Haskell an acceptable Perl?
byosfameron
 
Byterun, a Python bytecode interpreter - Allison Kaptur at NYCPython
byakaptur
 
Clustering com numpy e cython
byAnderson Dantas
 
Data monsters probablistic data structures
byGreenM
 
Allison Kaptur: Bytes in the Machine: Inside the CPython interpreter, PyGotha...
byakaptur
 
dplyr and torrents from cpasbien
byRomain Francois
 
The groovy puzzlers (as Presented at JavaOne 2014)
byGroovyPuzzlers
 
Malcon2017
byAndriy Brukhovetskyy
 
M12 random forest-part01
byRaman Kannan
 
Bytes in the Machine: Inside the CPython interpreter
byakaptur
 
Spark DataFrames for Data Munging
by(Susan) Xinh Huynh
 
M09-Cross validating-naive-bayes
byRaman Kannan
 
第二讲 Python基礎
byjuzihua1102
 
第二讲 预备-Python基礎
byanzhong70
 
"A 1,500 line (!!) switch statement powers your Python!" - Allison Kaptur, !!...
byakaptur
 
(gentle (introduction Clojure))
byGuy Taylor
 
An Elephant of a Different Colour: Hack
byVic Metcalfe
 
PLOTCON NYC: Behind Every Great Plot There's a Great Deal of Wrangling
byPlotly
 
Top 10 php classic traps
byDamien Seguy
 
MongoUK - PHP Development
byBoxed Ice
 

Viewers also liked

PDF
Rによるテキストマイニングの一例
byLINE Corp.
 
PDF
TwitterのデータをRであれこれ
byTakeshi Arabiki
 
PDF
あんちべのすべらない話～俺のツイートがこんなにウケないはずがない～
byantibayesian 俺がS式だ
 
PDF
RでTwitterテキストマイニング
byYudai Shinbo
 
PDF
「数字を見せろ」から「コードを見せろ」へ　〜過程の透明性を確保したデータ可視化を目指す〜
byKeiichiro Ono
 
PPTX
さくっとはじめるテキストマイニング（R言語）　　スタートアップ編
byYutaka Shimada
 
PPTX
動画タグネットワーク分析を用いた ニコニコ動画における萌芽文化発見の試み ~”ゆっくり関連タグ”を例として~
byYusuke Fukasawa
 
PPTX
経済的寄与度を重視した橋梁管理手法の有効性に関するマルチエージェントシミュレーションによる分析
byYusuke Fukasawa
 
PPTX
【ニコニコ動画】"マリオメーカー問題"のもたらした影響をネットワーク分析してみた
byYusuke Fukasawa
 
PPTX
料理レシピサービスにおける検索語の意味変化に関する分析
byYusuke Fukasawa
 
PPTX
DeNAの報告書を可視化して雰囲気をつかむ
byYusuke Fukasawa
 
PPTX
ニコニコ動画における関連動画情報を用いたカテゴリ特徴の把握
byYusuke Fukasawa
 
PPTX
巨大な表を高速に扱うData.table について
byHaruka Ozaki
 
PDF
data.tableパッケージで大規模データをサクッと処理する
byShintaro Fukushima
 
PDF
LDA等のトピックモデル
byMathieu Bertin
 
PDF
トピックモデルの話
bykogecoo
 
Rによるテキストマイニングの一例
byLINE Corp.
 
TwitterのデータをRであれこれ
byTakeshi Arabiki
 
あんちべのすべらない話～俺のツイートがこんなにウケないはずがない～
byantibayesian 俺がS式だ
 
RでTwitterテキストマイニング
byYudai Shinbo
 
「数字を見せろ」から「コードを見せろ」へ　〜過程の透明性を確保したデータ可視化を目指す〜
byKeiichiro Ono
 
さくっとはじめるテキストマイニング（R言語）　　スタートアップ編
byYutaka Shimada
 
動画タグネットワーク分析を用いた ニコニコ動画における萌芽文化発見の試み ~”ゆっくり関連タグ”を例として~
byYusuke Fukasawa
 
経済的寄与度を重視した橋梁管理手法の有効性に関するマルチエージェントシミュレーションによる分析
byYusuke Fukasawa
 
【ニコニコ動画】"マリオメーカー問題"のもたらした影響をネットワーク分析してみた
byYusuke Fukasawa
 
料理レシピサービスにおける検索語の意味変化に関する分析
byYusuke Fukasawa
 
DeNAの報告書を可視化して雰囲気をつかむ
byYusuke Fukasawa
 
ニコニコ動画における関連動画情報を用いたカテゴリ特徴の把握
byYusuke Fukasawa
 
巨大な表を高速に扱うData.table について
byHaruka Ozaki
 
data.tableパッケージで大規模データをサクッと処理する
byShintaro Fukushima
 
LDA等のトピックモデル
byMathieu Bertin
 
トピックモデルの話
bykogecoo
 

Similar to RではじめるTwitter解析

PDF
08 Functions
byHadley Wickham
 
PDF
08 functions
byHadley Wickham
 
PDF
06 data
byHadley Wickham
 
PDF
Data Munging in R - Chicago R User Group
bydesignandanalytics
 
PDF
Optimization In R
bysyou6162
 
PDF
Nagoya.R #6 LT Beanplot
bySAKAUE, Tatsuya
 
PDF
20 date-times
byHadley Wickham
 
08 Functions
byHadley Wickham
 
08 functions
byHadley Wickham
 
06 data
byHadley Wickham
 
Data Munging in R - Chicago R User Group
bydesignandanalytics
 
Optimization In R
bysyou6162
 
Nagoya.R #6 LT Beanplot
bySAKAUE, Tatsuya
 
20 date-times
byHadley Wickham
 

More from Takeshi Arabiki

PDF
開発の心得
byTakeshi Arabiki
 
PDF
クックパッド特売情報 における自然言語処理 〜固有表現抽出を利用した検索システム〜
byTakeshi Arabiki
 
PDF
Introduction to Japanese Morphological Analysis
byTakeshi Arabiki
 
PDF
R による文書分類入門
byTakeshi Arabiki
 
PDF
Rのデータ構造とメモリ管理
byTakeshi Arabiki
 
PDF
HTML5 Canvas で学ぶアフィン変換
byTakeshi Arabiki
 
PDF
Introduction to Favmemo for Immature Engineers
byTakeshi Arabiki
 
PDF
Rのスコープとフレームと環境と
byTakeshi Arabiki
 
PDF
R版Getopt::Longを作ってみた
byTakeshi Arabiki
 
PDF
Rデータフレーム自由自在
byTakeshi Arabiki
 
PDF
HMM, MEMM, CRF メモ
byTakeshi Arabiki
 
PDF
文字列カーネルによる辞書なしツイート分類 〜文字列カーネル入門〜
byTakeshi Arabiki
 
PDF
Rデバッグあれこれ
byTakeshi Arabiki
 
PDF
はじめてのまっぷりでゅ〜す
byTakeshi Arabiki
 
PDF
Twitterのデータを取得する準備
byTakeshi Arabiki
 
開発の心得
byTakeshi Arabiki
 
クックパッド特売情報 における自然言語処理 〜固有表現抽出を利用した検索システム〜
byTakeshi Arabiki
 
Introduction to Japanese Morphological Analysis
byTakeshi Arabiki
 
R による文書分類入門
byTakeshi Arabiki
 
Rのデータ構造とメモリ管理
byTakeshi Arabiki
 
HTML5 Canvas で学ぶアフィン変換
byTakeshi Arabiki
 
Introduction to Favmemo for Immature Engineers
byTakeshi Arabiki
 
Rのスコープとフレームと環境と
byTakeshi Arabiki
 
R版Getopt::Longを作ってみた
byTakeshi Arabiki
 
Rデータフレーム自由自在
byTakeshi Arabiki
 
HMM, MEMM, CRF メモ
byTakeshi Arabiki
 
文字列カーネルによる辞書なしツイート分類 〜文字列カーネル入門〜
byTakeshi Arabiki
 
Rデバッグあれこれ
byTakeshi Arabiki
 
はじめてのまっぷりでゅ〜す
byTakeshi Arabiki
 
Twitterのデータを取得する準備
byTakeshi Arabiki
 

Recently uploaded

PDF
2025-CB-NCAE-slide-deck.pptx-1-83.pdf Virtual Orientation on the Administrati...
byNelizabethEsplaguera1
 
PPTX
About Computer Hardware ................
bymaratiakshaya07
 
PPTX
Design Flaws and Known Attacks in Agentic AI - ITI Business Session
byInformation Technology Institute
 
PDF
Profiting from Technological Innovation: Managing Intellectual Property to Bu...
byDavid Teece
 
PDF
6 Insights from the Patron Saint of AI Failure
byScott M. Graffius
 
PDF
Are AI Hallucinations Getting Better or Worse? We Analyzed the Data.
byScott M. Graffius
 
PDF
Special Reeling Cable Specification _ Anhui Feichun Special Cable Co., Ltd_.pdf
byAnhui Feichun Special Cable Co., Ltd.
 
PDF
Philippine Agricultural Engineering Standard_Hand Tractor .pdf
by6xfrnmdp4b
 
PPTX
Accelerate Innovation with Engineering R&D Services
byChetu
 
PDF
AI Data Analyst: Smarter Insights for Modern Real Estate Decisions
byLeni Analyst
 
PPTX
Robot Vacuums are Cleaning Up. The global robot vacuum segment has high marke...
byRobert March
 
PDF
Zniber Partners. Audience Foresight for Confident Decisions
bySamuel Zniber
 
PDF
Just Eat Data Scraping For Restaurant Reviews And Pricing.pdf
byfusiondata2026
 
PDF
2006 IEEE International Solid-State Circuits Conference
bySlide_N
 
PDF
Presentation of Cybersecurity and data protection
bytarikaityahya2
 
PPTX
Webinar: EVerest for Production: Accelerating EV Charger Development w/ Indus...
byDanBrown980551
 
PPTX
computer science class 11 ( phishing and fraud mails) .pptx
bybalajichess314
 
PDF
One String, Many Prompts: AI Code Generation
byGeoffrey Goetz
 
PDF
A paper in the leading academic journal Telecommunication Policy cited Scott ...
byScott M. Graffius
 
PDF
Unlocking Gen AI Capabilities Within UiPath Document Understanding
byDianaGray10
 
2025-CB-NCAE-slide-deck.pptx-1-83.pdf Virtual Orientation on the Administrati...
byNelizabethEsplaguera1
 
About Computer Hardware ................
bymaratiakshaya07
 
Design Flaws and Known Attacks in Agentic AI - ITI Business Session
byInformation Technology Institute
 
Profiting from Technological Innovation: Managing Intellectual Property to Bu...
byDavid Teece
 
6 Insights from the Patron Saint of AI Failure
byScott M. Graffius
 
Are AI Hallucinations Getting Better or Worse? We Analyzed the Data.
byScott M. Graffius
 
Special Reeling Cable Specification _ Anhui Feichun Special Cable Co., Ltd_.pdf
byAnhui Feichun Special Cable Co., Ltd.
 
Philippine Agricultural Engineering Standard_Hand Tractor .pdf
by6xfrnmdp4b
 
Accelerate Innovation with Engineering R&D Services
byChetu
 
AI Data Analyst: Smarter Insights for Modern Real Estate Decisions
byLeni Analyst
 
Robot Vacuums are Cleaning Up. The global robot vacuum segment has high marke...
byRobert March
 
Zniber Partners. Audience Foresight for Confident Decisions
bySamuel Zniber
 
Just Eat Data Scraping For Restaurant Reviews And Pricing.pdf
byfusiondata2026
 
2006 IEEE International Solid-State Circuits Conference
bySlide_N
 
Presentation of Cybersecurity and data protection
bytarikaityahya2
 
Webinar: EVerest for Production: Accelerating EV Charger Development w/ Indus...
byDanBrown980551
 
computer science class 11 ( phishing and fraud mails) .pptx
bybalajichess314
 
One String, Many Prompts: AI Code Generation
byGeoffrey Goetz
 
A paper in the leading academic journal Telecommunication Policy cited Scott ...
byScott M. Graffius
 
Unlocking Gen AI Capabilities Within UiPath Document Understanding
byDianaGray10
 

RではじめるTwitter解析

  • 1.
    R Twitter R 2011 (2011/11/26) @a_bicky
  • 2.
    • Takeshi Arabiki ‣ ‣ Twitter & : @a_bicky & id:a_bicky • R • http://d.hatena.ne.jp/a_bicky/
  • 3.
    R Osaka.R #4 Tokyo.R #16 Tsukuba.R #9 http://www.slideshare.net/abicky/twitterr http://www.slideshare.net/abicky/r-9034336 http://www.slideshare.net/abicky/r-10128090
  • 4.
  • 6.
  • 7.
  • 8.
  • 9.
  • 10.
    http://twilog.org/ http://twitraq.userlocal.jp/ http://whotwi.com/ http://tweetstats.com/
  • 11.
    http://twilog.org/ http://twitraq.userlocal.jp/ R http://whotwi.com/ http://tweetstats.com/
  • 12.
    Twitter • • reshape2 • ggplot2 •
  • 13.
    Twitter • • reshape2 • ggplot2 •
  • 15.
    twitteR twitteR > library(twitteR) # twitteR > # (twitteR 0.99.15 ) > Sys.setlocale("LC_TIME", "C") [1] "C" > # @a_bicky 3,200 RT > statuses <- userTimeline("a_bicky", n = 3200)
  • 16.
    status > # R5 > ls.str(statuses[[1]]) created : POSIXct[1:1], format: "2011-11-23 22:16:24" favorited : logi FALSE ↑ UTC id : chr "139467359571296256" initFields : Formal class 'refMethodDef' [package "methods"] with 5 slots initialize : Formal class 'refMethodDef' [package "methods"] with 5 slots replyToSID : chr(0) replyToSN : chr(0) replyToUID : chr(0) screenName : chr "a_bicky" ! Twitter statusSource : chr "<a href="http://sites.google.com/site/ yorufukurou/" rel="nofollow">YoruFukurou</a>" text : chr " " truncated : logi FALSE ↑
  • 17.
    > statusDF <-twListToDF(statuses) > str(statusDF, vec.len = 1) 'data.frame': 3159 obs. of 10 variables: $ text : chr " " ... ↑ $ favorited : logi FALSE ... $ replyToSN : logi NA ... $ created : POSIXct, format: "2011-11-23 22:16:24" ... $ truncated : logi FALSE ... ↑ UTC $ replyToSID : logi NA ... $ id : chr "139467359571296256" ... $ replyToUID : logi NA ... $ statusSource: chr "<a href="http://sites.google.com/ site/yorufukurou/" rel="nofollow">YoruFukurou</a>" ... $ screenName : chr "a_bicky" ...
  • 18.
    > wday.abb <-c("Mon", "Tue", "Wed", "Thu", "Fri", "Sat", "Sun") > statusDF <- within(statusDF, { + attr(created, "tzone") <- "Asia/Tokyo" # JST + statusSource <- factor(gsub("<a .*?>(.*?)</a>", "1", statusSource)) # HTML + date <- factor(format(created, "%Y-%m-%d")) # + hour <- NULL; month <- NULL; year <- NULL; wday <- NULL + with(as.POSIXlt(created), { + hour <<- factor(hour) # + month <<- factor(mon + 1) # + year <<- factor(year + 1900) # + wday <<- factor((wday + 6) %% 7, labels = wday.abb) # + }) + textLength <- nchar(text) # + # , URL, + cleanText <- removeSpecialStr(text) + cleanTextLength <- nchar(cleanText) # URL + })
  • 19.
    > # Twitter > topSources <- names(head(sort(table(statusDF$statusSource), decreasing = TRUE), 5)) > statusDF <- within(statusDF, { + statusSource <- as.character(statusSource) + statusSource[!statusSource %in% topSources] <- "other" + # + statusSource <- factor(statusSource, levels = names(sort(table (statusSource), dec = TRUE))) + })
  • 20.
    Twitter • • reshape2 • ggplot2 •
  • 21.
  • 23.
    Excel 9 11 ”Twitter for iPhone”, ”YoruFukurou” Sat Mon 12 23
  • 24.
    reshape2 > library(reshape2) > acast(melt(statusDF,id.vars = c("statusSource", "wday", "month", "hour"), + measure.vars = c("textLength")), + month + statusSource ~ wday, mean, + subset = .(statusSource %in% c("YoruFukurou", "Twitter for iPhone") + & month %in% 9:11 & hour %in% 12:23 + & wday %in% c("Mon", "Sat", "Sun"))) Mon Sat Sun 9_YoruFukurou 43 42.13333 54.76471 9_Twitter for iPhone 16 27.70000 20.50000 10_YoruFukurou 61 41.70175 56.98333 10_Twitter for iPhone NaN 27.00000 24.50000 11_YoruFukurou 35 41.08197 57.32609 11_Twitter for iPhone NaN NaN 32.00000
  • 25.
    reshape2 > library(reshape2) > acast(melt(statusDF,id.vars = c("statusSource", "wday", "month", "hour"), + measure.vars = c("textLength")), + month + statusSource ~ wday, mean, + subset = .(statusSource %in% c("YoruFukurou", "Twitter for iPhone") + & month %in% 9:11 & hour %in% 12:23 + & wday %in% c("Mon", "Sat", "Sun"))) Mon Sat Sun 9_YoruFukurou 43 42.13333 54.76471 9_Twitter for iPhone 16 27.70000 20.50000 10_YoruFukurou 61 41.70175 56.98333 10_Twitter for iPhone NaN 27.00000 24.50000 11_YoruFukurou 35 41.08197 57.32609 11_Twitter for iPhone NaN NaN 32.00000
  • 26.
    reshape2 > library(reshape2) > acast(melt(statusDF,id.vars = c("statusSource", "wday", "month", "hour"), + measure.vars = c("textLength")), + month + statusSource ~ wday, mean, + subset = .(statusSource %in% c("YoruFukurou", "Twitter for iPhone") + & month %in% 9:11 & hour %in% 12:23 + & wday %in% c("Mon", "Sat", "Sun"))) Mon Sat Sun 9_YoruFukurou 43 42.13333 54.76471 9_Twitter for iPhone 16 27.70000 20.50000 10_YoruFukurou 61 41.70175 56.98333 10_Twitter for iPhone NaN 27.00000 24.50000 11_YoruFukurou 35 41.08197 57.32609 11_Twitter for iPhone NaN NaN 32.00000 R
  • 27.
    reshape2 melt melt cast melt cast > mstatus <- melt(statusDF, + id.vars = c("statusSource", "wday", "year", "month", "hour", "date"), + measure.vars = c("textLength", "cleanTextLength")) > mstatus[3157:3162, ] statusSource wday year month hour date variable value 3157 web Sun 2011 3 20 2011-03-13 textLength 72 3158 web Sun 2011 3 16 2011-03-13 textLength 24 3159 web Sun 2011 3 14 2011-03-13 textLength 82 3160 YoruFukurou Wed 2011 11 1 2011-11-23 cleanTextLength 87 3161 YoruFukurou Wed 2011 11 1 2011-11-23 cleanTextLength 14 3162 YoruFukurou Wed 2011 11 1 2011-11-23 cleanTextLength 21 id
  • 28.
    reshape2 cast cast formula fun.aggregate > args(acast) # array acast function (data, formula, fun.aggregate = NULL, ..., margins = NULL, subset = NULL, fill = NULL, drop = TRUE, value_var = guess_value(data)) NULL > args(dcast) # data.frame dcast function (data, formula, fun.aggregate = NULL, ..., margins = NULL, subset = NULL, fill = NULL, drop = TRUE, value_var = guess_value(data)) NULL formula ... . acast hoge ~ fuga ~ piyo ※dcast 1 hoge ~ fuga + piyo
  • 29.
    > # > acast(mstatus,. ~ wday, length, subset = .(variable == "textLength")) ↑ cleanTextLength
  • 30.
    > # > acast(mstatus,. ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun [1,] 408 360 258 294 334 801 704 >
  • 31.
    > # > acast(mstatus,. ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun [1,] 408 360 258 294 334 801 704 >
  • 32.
    > # > acast(mstatus, . ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun [1,] 408 360 258 294 334 801 704 > # > acast(mstatus, hour ~ wday, length, subset = .(variable == "textLength"))
  • 33.
    > # > acast(mstatus,. ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun [1,] 408 360 258 294 334 801 704 > # > acast(mstatus, hour ~ wday, length, subset = .(variable == "textLength")) Mon Tue Wed Thu Fri Sat Sun 0 65 69 26 46 46 49 40 1 48 19 11 15 27 44 37 2 31 24 6 16 17 23 17 3 27 19 4 11 14 17 10 4 4 15 1 7 4 5 7 5 5 11 1 4 3 4 5 6 4 14 3 6 9 8 1
  • 34.
    > # > # >acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength"))
  • 35.
    > # > # >acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength")) Mon_3 Mon_4 Mon_5 Mon_6 Mon_7 Mon_8 Mon_9 Mon_10 Mon_11 Tue_3 Tue_4 0 3 4 13 3 1 10 7 15 9 4 2 1 0 0 1 0 1 9 16 12 9 1 0 2 2 0 0 0 2 7 6 7 7 2 0
  • 36.
    > # > # >acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength")) Mon_3 Mon_4 Mon_5 Mon_6 Mon_7 Mon_8 Mon_9 Mon_10 Mon_11 Tue_3 Tue_4 0 3 4 13 3 1 10 7 15 9 4 2 1 0 0 1 0 1 9 16 12 9 1 0 2 2 0 0 0 2 7 6 7 7 2 0 > # 3 > acast(mstatus, hour ~ wday ~ month, length, subset = .(variable == "textLength"))
  • 37.
    > # > # >acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength")) Mon_3 Mon_4 Mon_5 Mon_6 Mon_7 Mon_8 Mon_9 Mon_10 Mon_11 Tue_3 Tue_4 0 3 4 13 3 1 10 7 15 9 4 2 1 0 0 1 0 1 9 16 12 9 1 0 2 2 0 0 0 2 7 6 7 7 2 0 > # 3 > acast(mstatus, hour ~ wday ~ month, length, subset = .(variable == "textLength")) , , 3 Mon Tue Wed Thu Fri Sat Sun 0 3 4 1 0 1 6 4 1 0 1 3 0 0 0 1
  • 38.
    > # > # >acast(mstatus, hour ~ wday + month, length, subset = .(variable == "textLength")) Mon_3 Mon_4 Mon_5 Mon_6 Mon_7 Mon_8 Mon_9 Mon_10 Mon_11 Tue_3 Tue_4 0 3 4 13 3 1 10 7 15 9 4 2 1 0 0 1 0 1 9 16 12 9 1 0 2 2 0 0 0 2 7 6 7 7 2 0 > # 3 > acast(mstatus, hour ~ wday ~ month, length, subset = .(variable == "textLength")) , , 3 3 Mon Tue Wed Thu Fri Sat Sun 0 3 4 1 0 1 6 4 1 0 1 3 0 0 0 1
  • 42.
    Twitter reshape2 1 > # > dcast(mstatus, statusSource ~ ., + function(x) list(c(mean = mean(x), sd = sd(x))), + fill = list(c(mean = NaN, sd = NA)), ← + subset = .(variable == "textLength"))
  • 43.
    Twitter reshape2 1 > # > dcast(mstatus, statusSource ~ ., + function(x) list(c(mean = mean(x), sd = sd(x))), + fill = list(c(mean = NaN, sd = NA)), ← + subset = .(variable == "textLength")) statusSource NA 1 YoruFukurou 47.51462, 32.57973 2 web 57.02720, 36.33534 3 Twitter for iPhone 33.42342, 23.06466 4 Twitter for Android 28.49048, 20.08457 5 Hatena 80.00000, 25.94212 6 other 52.58621, 33.12180 >
  • 44.
    Twitter reshape2 1 > # > dcast(mstatus, statusSource ~ ., + function(x) list(c(mean = mean(x), sd = sd(x))), + fill = list(c(mean = NaN, sd = NA)), ← + subset = .(variable == "textLength")) statusSource NA 1 YoruFukurou 47.51462, 32.57973 2 web 57.02720, 36.33534 3 Twitter for iPhone 33.42342, 23.06466 4 Twitter for Android 28.49048, 20.08457 5 Hatena 80.00000, 25.94212 6 other 52.58621, 33.12180 >
  • 45.
    > # t > pc <- unlist(subset(statusDF, + statusSource %in% c("YoruFukurou", "web"), + textLength)) > sp <- unlist(subset(statusDF, + grepl("(iPhone|Android)", statusSource), + textLength)) > t.test(sp, pc, var.equal = FALSE) Welch Two Sample t-test !! data: sp and pc t = -15.7921, df = 1588.246, p-value < 2.2e-16 alternative hypothesis: true difference in means is not equal to 0 95 percent confidence interval: -19.85334 -15.46645 sample estimates: mean of x mean of y 31.83945 49.49935
  • 46.
    > # t > pc <- unlist(subset(statusDF, + statusSource %in% c("YoruFukurou", "web"), + textLength)) > sp <- unlist(subset(statusDF, + grepl("(iPhone|Android)", statusSource), + textLength)) > t.test(sp, pc, var.equal = FALSE) Welch Two Sample t-test !! data: sp and pc t = -15.7921, df = 1588.246, p-value < 2.2e-16 alternative hypothesis: true difference in means is not equal to 0 95 percent confidence interval: -19.85334 -15.46645 sample estimates: mean of x mean of y 31.83945 49.49935
  • 47.
    > extractScreenNames <-function(text, strict = TRUE) { + if (strict) { + # Twitter screen_name + regex <- "(?:(?<!w)([@ ])((?>w+))(?![@ ])|[sS])" + } else { + # [email protected] + regex <- "(?:([@ ])(w+)|[sS])" + } + screenNames <- gsub(regex, "12", text, perl = TRUE) + unique(unlist(strsplit(substring(screenNames, 2), "[@ ]"))) + } > screenNames <- unlist(lapply(statusDF$text, extractScreenNames)) > head(sort(table(screenNames), decreasing = TRUE), 10) # Top 10 screenNames naopr __gfx__ hirota_inoue mandy_44 ask_a_lie 105 85 51 47 40 ken_nishi nokuno yokkuns JinJin0613 kanon19_rie 39 39 33 20 20
  • 48.
    Twitter • • reshape2 • ggplot2 •
  • 49.
  • 50.
    ggplot2 plot(statusDF$wday, col ="blue") ggplot2 qplot(wday, data = statusDF, fill = I("blue"), alpha = I(0.7), xlab = "", ylab = "")
  • 51.
    ggplot2 qplot(wday, data =statusDF, fill = statusSource, xlab = "", ylab = "")
  • 52.
    ggplot2 qplot(wday, data = statusDF, facets = ~ statusSource, fill = I("blue"), alpha = I(0.7), xlab = "", ylab = "") qplot(wday, data = statusDF, fill = statusSource, xlab = "", ylab = "")
  • 53.
    ggplot2 qplot(wday, data = statusDF, facets = ~ statusSource, fill = I("blue"), alpha = I(0.7), xlab = "", ylab = "") qplot(wday, data = statusDF, fill = statusSource, xlab = "", ylab = "")
  • 54.
    qplot ggplot2 > args(qplot) function (x, y = NULL, z = NULL, ..., data, facets = . ~ ., margins = FALSE, geom = "auto", stat = list(NULL), position = list(NULL), xlim = c(NA, NA), ylim = c(NA, NA), log = "", main = NULL, xlab = deparse(substitute(x)), ylab = deparse(substitute(y)), asp = NA) NULL
  • 55.
    qplot geom geom area: bar: histogram: line: point:
  • 56.
    qplot geom geom area: bar: histogram: line: point: qplot(as.integer(wday), data = statusDF, geom = "area", stat = "bin", fill = statusSource, xlab = "", ylab = "", binwidth = 1)
  • 57.
    qplot geom geom area: bar: histogram: line: point: qplot(wday, data = statusDF, geom = "bar", stat = "bin", fill = statusSource, xlab = "", ylab = "")
  • 58.
    qplot geom geom area: bar: histogram: line: point: qplot(as.integer(wday), data = statusDF, geom = "line", stat = "bin", colour = statusSource, xlab = "", ylab = "", binwidth = 1)
  • 59.
    qplot geom geom area: bar: histogram: line: point: qplot(wday, data = statusDF, geom = "point", stat = "bin", colour = statusSource, xlab = "", ylab = "")
  • 60.
    qplot position position geom dodge : fill : 1 jitter : stack :
  • 61.
    qplot position position geom dodge : fill : 1 jitter : stack : qplot(wday, data = statusDF, fill = statusSource, position = "dodge", xlab = "", ylab = "")
  • 62.
    qplot position position geom dodge : fill : 1 jitter : stack : qplot(wday, data = statusDF, fill = statusSource, position = "fill", xlab = "", ylab = "")
  • 63.
    qplot position position geom dodge : fill : 1 jitter : stack : qplot(wday, data = statusDF, fill = statusSource, position = "jitter", xlab = "", ylab = "")
  • 64.
    qplot position position geom dodge : fill : 1 jitter : stack : qplot(wday, data = statusDF, fill = statusSource, position = "stack", xlab = "", ylab = "")
  • 65.
    qplot facets facets geom ~ : 1 ~ 2: 1, 2 ※reshape2 1 ~ 2 + 3
  • 66.
    qplot facets facets geom ~ : 1 ~ 2: 1, 2 ※reshape2 1 ~ 2 + 3 qplot(wday, data = statusDF, xlab = "", ylab = "", facets = ~ statusSource)
  • 67.
    qplot facets facets geom ~ : 1 ~ 2: 1, 2 ※reshape2 1 ~ 2 + 3 qplot(wday, data = statusDF, xlab = "", ylab = "", facets = month ~ statusSource)
  • 68.
    qplot alpha : colour (color) : fill : linetype : size : colour, fill, linetype statusSource fill = I("blue") I (AsIs)
  • 69.
    qplot alpha : colour (color) : fill : linetype : size : qplot(wday, data = statusDF, xlab = "", ylab = "", alpha = as.integer(wday))
  • 70.
    qplot alpha : colour (color) : fill : linetype : size : qplot(wday, data = statusDF, xlab = "", ylab = "", colour = statusSource)
  • 71.
    qplot alpha : colour (color) : fill : linetype : size : qplot(wday, data = statusDF, xlab = "", ylab = "", fill = statusSource)
  • 72.
    qplot alpha : colour (color) : fill : linetype : size : qplot(wday, data = statusDF, xlab = "", ylab = "", linetype = statusSource, colour = statusSource)
  • 75.
    whotwi http://whotwi.com/
  • 76.
    whotwi http://whotwi.com/
  • 77.
    whotwi > # Twitter > # melt cast xtabs > cnt <- as.data.frame(xtabs(~ hour + wday + statusSource, statusDF)) > head(cnt, 3) hour wday statusSource Freq 1 0 Mon YoruFukurou 48 2 1 Mon YoruFukurou 38 3 2 Mon YoruFukurou 25
  • 78.
    whotwi > # Twitter > # melt cast xtabs > cnt <- as.data.frame(xtabs(~ hour + wday + statusSource, statusDF)) > head(cnt, 3) hour wday statusSource Freq 1 0 Mon YoruFukurou 48 2 1 Mon YoruFukurou 38 3 2 Mon YoruFukurou 25 > freqSources <- by(cnt, cnt[c("hour", "wday")], function(df) { + # + freqSource <- with(df, statusSource[order(Freq, decreasing = TRUE) [1]]) + cbind(df[1, c("hour", "wday")], freqSource) + }) > freqSources <- do.call(rbind, freqSources) > head(freqSources, 3) hour wday freqSource 1 0 Mon YoruFukurou 2 1 Mon YoruFukurou 3 2 Mon YoruFukurou
  • 79.
    whotwi > # > cntSum<- as.data.frame(xtabs(Freq ~ hour + wday, cnt)) > head(cntSum, 3) hour wday Freq 1 0 Mon 65 2 1 Mon 48 3 2 Mon 31
  • 80.
    whotwi > # > cntSum<- as.data.frame(xtabs(Freq ~ hour + wday, cnt)) > head(cntSum, 3) hour wday Freq 1 0 Mon 65 2 1 Mon 48 3 2 Mon 31 > # > data <- merge(cntSum, freqSources) > # > data$wday <- factor(data$wday, levels = rev(levels(data$wday))) > # > data$Freq <- log2(data$Freq)
  • 81.
    whotwi > # > cntSum<- as.data.frame(xtabs(Freq ~ hour + wday, cnt)) > head(cntSum, 3) hour wday Freq 1 0 Mon 65 2 1 Mon 48 3 2 Mon 31 > # > data <- merge(cntSum, freqSources) > # > data$wday <- factor(data$wday, levels = rev(levels(data$wday))) > # > data$Freq <- log2(data$Freq) > p <- qplot(hour, wday, data = data, xlab = "", ylab = "", + geom = "point", colour = freqSource, size = Freq) > p # print(p)
  • 82.
    whotwi > # > cntSum<- as.data.frame(xtabs(Freq ~ hour + wday, cnt)) > head(cntSum, 3) hour wday Freq 1 0 Mon 65 2 1 Mon 48 3 2 Mon 31 > # > data <- merge(cntSum, freqSources) > # > data$wday <- factor(data$wday, levels = rev(levels(data$wday))) > # > data$Freq <- log2(data$Freq) > p <- qplot(hour, wday, data = data, xlab = "", ylab = "", + geom = "point", colour = freqSource, size = Freq) > p # print(p)
  • 83.
  • 84.
  • 85.
    whotwi > # whotwitheme > theme_whotwi <- function() { + opts( # + panel.background = theme_rect(fill = NA, colour = NA), + # + legend.key = theme_rect(fill = NA, colour = NA), + # + axis.ticks = theme_segment(colour = NA)) + } > p2 <- p + theme_whotwi() + scale_size(legend = FALSE) + scale_colour_hue(name = "") > p2
  • 86.
    whotwi > # whotwitheme > theme_whotwi <- function() { + opts( # + panel.background = theme_rect(fill = NA, colour = NA), + # + legend.key = theme_rect(fill = NA, colour = NA), + # + axis.ticks = theme_segment(colour = NA)) + } > p2 <- p + theme_whotwi() + scale_size(legend = FALSE) + scale_colour_hue(name = "") > p2
  • 87.
  • 88.
  • 89.
  • 90.
  • 91.
  • 92.
    Twitter • • reshape2 • ggplot2 •
  • 95.
  • 96.
  • 97.
  • 98.
    RMeCab MeCab R > library(RMeCab) > (docDF(data.frame(" "), column = 1, type = 1)) number of extracted terms = 5 now making a data frame. wait a while! TERM POS1 POS2 Row1 1 1 2 1 3 1 4 2 5 2
  • 99.
    http://www.lr.pi.titech.ac.jp/~takamura/pndic_ja.html : : :1 : : :0.999995 : : :0.999979 : : :0.999979 : : :0.999645 : : :0.999486 : : :0.999314 ...
  • 100.
    > # > pndic<- read.table("http://www.lr.pi.titech.ac.jp/~takamura/pubs/ pn_ja.dic", + sep = ":", + col.names = c("term", "kana", "pos", "value"), + colClasses = c("character", "character", "factor", "numeric"), + fileEncoding = "Shift_JIS") > # > # > pndic2 <- aggregate(value ~ term + pos, pndic, mean)
  • 101.
    > # pndic >pos <- unique(pndic2$pos) > tweetDF <- docDF(statusDF, column = "cleanText", type = 1, pos = pos) number of extracted terms = 7164 now making a data frame. wait a while! > tweetDF[2900:2904, 1:5] TERM POS1 POS2 Row1 Row2 2900 0 0 2901 0 0 2902 0 0 2903 0 0 2904 0 0 > # pndic > tweetDF <- subset(tweetDF, TERM %in% pndic2$term) > # > tweetDF <- merge(tweetDF, pndic2, by.x = c("TERM", "POS1"), by.y = c ("term", "pos"))
  • 102.
    > # > score<- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117
  • 103.
    > # > score<- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117
  • 104.
    > # > score<- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765
  • 105.
    > # > score<- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765
  • 106.
    > # > score<- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765 > # > sum(score == 0) [1] 277
  • 107.
    > # > score<- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765 > # > sum(score == 0) [1] 277
  • 108.
    > # > score<- colSums(tweetDF[4:(ncol(tweetDF) - 1)] * tweetDF$value) > # > sum(score > 0) [1] 117 > # > sum(score < 0) [1] 2765 > # > sum(score == 0) [1] 277
  • 109.
    > table(ifelse(pndic$value >0, "positive", + ifelse(pndic$value == 0, "neutral", "negative"))) negative neutral positive 49983 20 5122
  • 110.
    > table(ifelse(pndic$value >0, "positive", + ifelse(pndic$value == 0, "neutral", "negative"))) negative neutral positive 49983 20 5122
  • 111.
    > m <-mean(score) > # > tweetType <- factor(ifelse(score > m, "positive", + ifelse(score == m, "neutral", "negative")), + levels = c("positive", "neutral", "negative")) > table(tweetType) tweetType positive neutral negative 1912 0 1247
  • 112.
    > statusDF$tweetType <-droplevels(tweetType) > # > qplot(month, data = statusDF, + geom = "bar", fill = tweetType, position = "fill")
  • 113.
    > statusDF$tweetType <-droplevels(tweetType) > # > qplot(month, data = statusDF, + geom = "bar", fill = tweetType, position = "fill")
  • 114.
    > statusDF$tweetType <-droplevels(tweetType) > # > qplot(month, data = statusDF, + geom = "bar", fill = tweetType, position = "fill")
  • 115.
    > statusDF$tweetType <-droplevels(tweetType) > # > qplot(month, data = statusDF, + geom = "bar", fill = tweetType, position = "fill")
  • 118.
    twitteR • RJSONIO • • ID status ID • fav favorited TRUE • truncated TRUE • DM • status character factor
  • 119.
    twitteR • RJSONIO • • ID status ID • fav favorited TRUE • truncated TRUE • DM • status character factor
  • 120.
    OAuth ” ” twitteR -
  • 122.
    • twitteR • reshape2 R • ggplot2 • RMeCab R
  • 123.
    • twitteR • reshape2 R • ggplot2 • RMeCab R • PC •
  • 126.
  • 127.
    status > statuses[[1]]$text [1] " " > statuses[[1]]$getText() # [1] " " > # > statuses[[1]]$text <- " " > statuses[[1]]$getText() [1] " " > statuses[[1]]$setText("ggrks") # > statuses[[1]]$getText() [1] "ggrks" > # > statuses[[1]]$getCreated() [1] "2011-11-23 22:16:24 UTC"
  • 128.
    removeSpecialStr removeSpecialStr <- function(text){ removeURL(removeHashTag(removeScreenName(text))) }
  • 129.
    removeScreenName removeScreenName <- function(text,strict = TRUE) { if (strict) { regex <- "(?<!w)[@ ](?>w+)(?![@ ])" } else { regex <- "[@ ]w+" } gsub(regex, "", text, perl = TRUE) }
  • 130.
    removeURL removeURL <- function(text,strict = TRUE) { if (strict) { regex <- "(?<![-.w#@=!'"/])https?://(?:[^:]+:. +@)?(?:[0-9A-Za-z][-0-9A-Za-z]*(?<!-).)+[A-za-z]+(?:/[- w#%=+,.?!&~]*)*" } else { regex <- "https?://[-w#%=+,.?!&~/]+" } gsub(regex, "", text, perl = TRUE) }
  • 131.
    removeHashTag removeHashTag <- function(text,strict = TRUE) { delimiters <- "s,.u3000-u3002uFF01uFF1F" # cf. http://nobu666.com/2011/07/13/914.html validJa <- "u3041-u3094u3099-u309Cu30A1-u30FA u30FCu3400-uD7A3uFF10-uFF19uFF21-uFF3AuFF41-uFF5A uFF66-uFF9E" if (strict) { regex <- sprintf("(^|[%s])(?:([# ](?>[0-9]+)(?! w))|[# ][w%s]+)", delimiters, validJa, validJa) } else { regex <- sprintf("[# ][^%s]+", delimiters) } gsub(regex, "12", text, perl = TRUE) }