# tidytext 패키지 로드 및 감성 데이터셋 확인
library(tidytext)
sentiments
## # A tibble: 6,786 × 2
## word sentiment
## <chr> <chr>
## 1 2-faces negative
## 2 abnormal negative
## 3 abolish negative
## 4 abominable negative
## 5 abominably negative
## 6 abominate negative
## 7 abomination negative
## 8 abort negative
## 9 aborted negative
## 10 aborts negative
## # ℹ 6,776 more rows
install.packages("textdata")
## Error in contrib.url(repos, "source"): trying to use CRAN without setting a mirror
library(textdata)
# 각 감성 사전 로드 (afinn, nrc, loughran)
get_sentiments(lexicon="afinn")
## # A tibble: 2,477 × 2
## word value
## <chr> <dbl>
## 1 abandon -2
## 2 abandoned -2
## 3 abandons -2
## 4 abducted -2
## 5 abduction -2
## 6 abductions -2
## 7 abhor -3
## 8 abhorred -3
## 9 abhorrent -3
## 10 abhors -3
## # ℹ 2,467 more rows
get_sentiments(lexicon="nrc")
## # A tibble: 13,872 × 2
## word sentiment
## <chr> <chr>
## 1 abacus trust
## 2 abandon fear
## 3 abandon negative
## 4 abandon sadness
## 5 abandoned anger
## 6 abandoned fear
## 7 abandoned negative
## 8 abandoned sadness
## 9 abandonment anger
## 10 abandonment fear
## # ℹ 13,862 more rows
get_sentiments(lexicon="loughran")
## # A tibble: 4,150 × 2
## word sentiment
## <chr> <chr>
## 1 abandon negative
## 2 abandoned negative
## 3 abandoning negative
## 4 abandonment negative
## 5 abandonments negative
## 6 abandons negative
## 7 abdicated negative
## 8 abdicates negative
## 9 abdicating negative
## 10 abdication negative
## # ℹ 4,140 more rows
# 필요 패키지 로드
library(dplyr)
##
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
library(tibble)
library(purrr)
library(readr)
install.packages("lubridate")
## Error in contrib.url(repos, "source"): trying to use CRAN without setting a mirror
##
## Attaching package: 'lubridate'
## The following objects are masked from 'package:base':
##
## date, intersect, setdiff, union
# 트윗 데이터를 다운로드하여 로컬에 저장
url <- "https://archive.ics.uci.edu/ml/machine-learning-databases/00438/Health-News-Tweets.zip"
local.copy <- tempfile()
download.file(url, destfile=local.copy, mode="wb")
Sys.setlocale("LC_TIME", "en_US.UTF-8") #로케일 설정
# 트윗 데이터를 읽고 결합하여 tibble로 변환
health.twitter <-
map(unzip(zipfile=local.copy,
files=c("Health-Tweets/bbchealth.txt",
"Health-Tweets/cnnhealth.txt",
"Health-Tweets/foxnewshealth.txt",
"Health-Tweets/NBChealth.txt")),
read_delim, delim="|", quote="",
col_types=list(col_character(), col_character(), col_character()),
col_names=c("id", "datetime", "tweet")) %>%
map2(c("bbc", "cnn", "foxnews", "nbc"), ~cbind(.x, source=.y)) %>%
reduce(bind_rows) %>%
as_tibble() %>%
mutate(datetime=ymd_hms(strptime(datetime, "%a %b %d %H:%M:%S +0000 %Y")))
## Warning: One or more parsing issues, call `problems()` on your data frame for details,
## e.g.:
## dat <- vroom(...)
## problems(dat)
unlink(local.copy)
Sys.setlocale() # 로케일 원래대로 복원
## [1] "en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8"
# 출처별 트윗 수 확인
health.twitter %>%
count(source)
## # A tibble: 4 × 2
## source n
## <chr> <int>
## 1 bbc 3929
## 2 cnn 4061
## 3 foxnews 2000
## 4 nbc 4215
# 텍스트 전처리 수행
library(tidytext)
library(stringr)
health.words <- health.twitter %>%
select(-id) %>%
mutate(tweet=str_replace_all(tweet, pattern="(f|ht)tp\\S+\\s*", replacement="")) %>%
mutate(tweet=str_replace_all(tweet, pattern="\\d+", replacement="")) %>%
mutate(tweet=str_replace_all(tweet, pattern="\\bRT", replacement="")) %>%
mutate(tweet=str_replace_all(tweet, pattern="@\\S+", replacement="")) %>%
mutate(tweet=str_replace_all(tweet, pattern="&", replacement="")) %>%
unnest_tokens(word, tweet) # 토큰화하여 단어 추출
health.words
## # A tibble: 124,559 × 3
## datetime source word
## <dttm> <chr> <chr>
## 1 2015-04-09 01:31:50 bbc breast
## 2 2015-04-09 01:31:50 bbc cancer
## 3 2015-04-09 01:31:50 bbc risk
## 4 2015-04-09 01:31:50 bbc test
## 5 2015-04-09 01:31:50 bbc devised
## 6 2015-04-08 23:30:18 bbc gp
## 7 2015-04-08 23:30:18 bbc workload
## 8 2015-04-08 23:30:18 bbc harming
## 9 2015-04-08 23:30:18 bbc care
## 10 2015-04-08 23:30:18 bbc bma
## # ℹ 124,549 more rows
# 감성 분석 및 상위 10개 단어 추출
health.sentiment <- health.words %>%
inner_join(get_sentiments("bing"), by="word") %>%
count(word, sentiment, sort=TRUE) %>%
group_by(sentiment) %>%
slice_max(order_by=n, n=10) %>%
ungroup() %>%
mutate(nsign=ifelse(sentiment=="negative", -n, n))
health.sentiment
## # A tibble: 20 × 4
## word sentiment n nsign
## <chr> <chr> <int> <int>
## 1 cancer negative 701 -701
## 2 risk negative 317 -317
## 3 death negative 193 -193
## 4 outbreak negative 188 -188
## 5 virus negative 147 -147
## 6 fat negative 103 -103
## 7 loss negative 103 -103
## 8 deadly negative 102 -102
## 9 warning negative 96 -96
## 10 pain negative 92 -92
## 11 work positive 210 210
## 12 healthy positive 158 158
## 13 patient positive 155 155
## 14 good positive 139 139
## 15 like positive 139 139
## 16 better positive 135 135
## 17 support positive 102 102
## 18 free positive 95 95
## 19 love positive 90 90
## 20 best positive 87 87
# [그림 7-9] 그래프 생성
library(ggplot2)
library(scales)
##
## Attaching package: 'scales'
## The following object is masked from 'package:readr':
##
## col_factor
## The following object is masked from 'package:purrr':
##
## discard
ggplot(health.sentiment,
aes(x=reorder(word, nsign), y=nsign,
fill=factor(sentiment, levels=c("positive", "negative")))) +
geom_col(color="lightslategray", width=0.8) +
geom_text(aes(label=n), size=3, color="black",
hjust=ifelse(health.sentiment$nsign < 0, 1.1, -0.1)) +
scale_fill_manual(values=c("cornflowerblue", "tomato")) +
scale_y_continuous(breaks=pretty(health.sentiment$nsign),
labels=abs(pretty(health.sentiment$nsign))) +
coord_flip() +
labs(x=NULL, y="Count",
title="Health News Tweets") +
theme_minimal() +
theme(legend.position="bottom",
legend.title=element_blank(),
plot.title=element_text(face="bold"),
axis.text=element_text(face="bold", size=10))
# 의료 관련 용어 제외 후 분석
health.sentiment <- health.words %>%
inner_join(get_sentiments("bing"), by="word") %>%
filter(!(word %in% c("patient", "cancer", "virus"))) %>%
count(word, sentiment, sort=TRUE) %>%
group_by(sentiment) %>%
slice_max(order_by=n, n=10) %>%
ungroup() %>%
mutate(nsign=ifelse(sentiment=="negative", -n, n))
# [그림 7-10] 그래프 생성
library(ggplot2)
ggplot(health.sentiment,
aes(x=reorder(word, n), y=n,
fill=factor(sentiment, levels=c("positive", "negative")))) +
geom_col(color="lightslategray", width=0.6, show.legend = FALSE) +
geom_text(aes(label=n), size=3, color="black", hjust=1.2) +
scale_fill_manual(values=c("lightsteelblue1", "lightsalmon1")) +
facet_wrap(~ factor(sentiment, levels=c("positive", "negative")),
ncol=2, scales="free") +
coord_flip() +
labs(x=NULL, y="Count",
title="Health News Tweets") +
theme_light() +
theme(plot.title=element_text(face="bold"),
axis.line=element_line(color="gray"),
axis.text=element_text(face="bold", size=10))
# [그림 7-12] 건강뉴스 트윗 (회사별 상위 10개 긍/부정 단어)
health.sentiment <- health.words %>%
inner_join(get_sentiments("bing"), by="word") %>%
filter(!(word %in% c("patient", "cancer", "virus"))) %>%
count(word, sentiment, source, sort=TRUE) %>%
ungroup() %>%
group_by(source, sentiment) %>%
slice_max(order_by = n, n=10) %>%
ungroup()
health.sentiment
## # A tibble: 86 × 4
## word sentiment source n
## <chr> <chr> <chr> <int>
## 1 risk negative bbc 89
## 2 death negative bbc 72
## 3 warning negative bbc 54
## 4 crisis negative bbc 35
## 5 abuse negative bbc 29
## 6 outbreak negative bbc 26
## 7 threat negative bbc 26
## 8 fat negative bbc 22
## 9 poor negative bbc 22
## 10 strike negative bbc 22
## # ℹ 76 more rows
# 그래프 그리기
library(ggplot2)
ggplot(health.sentiment,
aes(reorder_within(x=word, by=n, within=source), n, fill=source)) +
geom_col(show.legend = FALSE) +
facet_wrap(~ factor(source,
labels=c("BBC", "CNN", "Fox News", "NBC")) + sentiment,
ncol=2, scales="free") +
scale_x_reordered() +
coord_flip() +
labs(x=NULL, y="Count",
title="Health News Tweets") +
theme_light() +
theme(strip.background=element_blank(),
strip.text=element_text(color="goldenrod4", face="bold"),
plot.title=element_text(face="bold"),
axis.line=element_line(color="gray"),
axis.text=element_text(face="bold", size=10),
panel.grid.minor=element_blank())
# 시간 흐름에 따른 추이 분석
library(lubridate)
health.sentiment <- health.words %>%
inner_join(get_sentiments("bing"), by="word") %>%
filter(!(word %in% c("patient", "cancer", "virus"))) %>%
mutate(time=floor_date(x=datetime, unit="month")) %>%
count(sentiment, time) %>%
group_by(sentiment) %>%
slice(2:(n()-1)) %>%
ungroup()
health.sentiment
## # A tibble: 62 × 3
## sentiment time n
## <chr> <dttm> <int>
## 1 negative 2012-09-01 00:00:00 86
## 2 negative 2012-10-01 00:00:00 94
## 3 negative 2012-11-01 00:00:00 54
## 4 negative 2012-12-01 00:00:00 65
## 5 negative 2013-01-01 00:00:00 135
## 6 negative 2013-02-01 00:00:00 188
## 7 negative 2013-03-01 00:00:00 183
## 8 negative 2013-04-01 00:00:00 197
## 9 negative 2013-05-01 00:00:00 212
## 10 negative 2013-06-01 00:00:00 176
## # ℹ 52 more rows
# [그림 7-13] 그래프 생성하기
Sys.setlocale("LC_TIME", "en_US.UTF-8")
library(ggplot2)
ggplot(health.sentiment, aes(x=time, y=n, fill=sentiment, color=sentiment)) +
geom_area(position="identity", alpha=0.3) +
geom_line(size=1.5) +
scale_fill_manual(labels=c("Negative", "Positive"),
values=c("orangered", "deepskyblue2")) +
scale_color_manual(labels=c("Negative", "Positive"),
values=c("orangered", "deepskyblue2")) +
scale_x_datetime(date_labels="%b %Y", date_breaks="6 months") +
labs(x=NULL, y="Count",
title="Health News Twetts") +
theme_minimal() +
theme(plot.title=element_text(face="bold"),
axis.text=element_text(face="bold"),
legend.position="bottom",
legend.title=element_blank())
## Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
## ℹ Please use `linewidth` instead.
## This warning is displayed once every 8 hours.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.
## [1] "en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8"
# [그림 7-14] 회사별 긍/부정 단어 출현 빈도 추이 분석 그래프 생성
Sys.setlocale("LC_TIME", "en_US.UTF-8")
library(ggplot2)
health.words %>%
inner_join(get_sentiments("bing"), by="word") %>%
filter(!(word %in% c("patient", "cancer", "virus"))) %>%
mutate(time=floor_date(datetime, unit="month")) %>%
count(source, sentiment, time) %>%
group_by(source, sentiment) %>%
slice(2:(n()-1)) %>%
ungroup() %>%
ggplot(aes(x=time, y=n, fill=sentiment, color=sentiment)) +
geom_area(position="identity", alpha=0.3) +
geom_line(size=1.5) +
facet_wrap(~ factor(source,
labels=c("BBC", "CNN", "Fox News", "NBC")),
nrow=4, scales="free") +
scale_fill_manual(labels=c("Negative", "Positive"),
values=c("coral", "cornflowerblue")) +
scale_color_manual(labels=c("Negative", "Positive"),
values=c("coral", "cornflowerblue")) +
scale_x_datetime(date_labels="%b %Y", date_breaks="2 months") +
labs(x=NULL, y="Count",
title="Health News Tweets") +
theme(plot.title=element_text(face="bold"),
axis.text.x=element_text(size=8),
legend.position="bottom",
legend.title=element_blank())
