term-topic probability

install.packages("topicmodels")
install.packages("tidytext")
install.packages("broom")
install.packages("tidyverse")
library(topicmodels)
library(tidytext)
library(broom)
library(tidyverse)

 

 

data("AssociatedPress")

 

 

# 2개의 topic LDA 분석
ap_lda <- topicmodels::LDA(AssociatedPress, k = 2, control = list(seed = 1234))

 

 

# broom 패키지에서 제공하는 tidy 형태로 변경
ap_topics <- broom::tidy(ap_lda, matrix = "beta")

 

 

ap_topics %>%
     dplyr::group_by(topic) %>%
     dplyr::top_n(n = 10, wt = beta) %>%
     dplyr::ungroup() %>%
     dplyr::arrange(topic, desc(beta)) %>%
     dplyr::mutate(term = reorder(term, beta)) %>%
     ggplot2::ggplot(mapping = aes(x = term, y = beta, fill = factor(topic))) +
     ggplot2::geom_col(show.legend = FALSE) +
     ggplot2::facet_wrap(~ topic, scales = "free") +
     ggplot2::coord_flip()

 

ap_topics %>%
     dplyr::mutate(topic = paste0("topic", topic)) %>%
     tidyr::spread(topic, beta) %>%
     dplyr::filter(topic1 > 0.001 | topic2 > 0.001) %>%
     dplyr::mutate(log_ratio = log2(topic2/topic1)) %>%
     dplyr::arrange(desc(log_ratio)) %>%
     ggplot2::ggplot(mapping = aes(x = reorder(term, -log_ratio), y = log_ratio)) +
     ggplot2::geom_col(show.legend = FALSE) +
     ggplot2::coord_flip()

 

[출처] R로 배우는 텍스트마이닝, 줄리아 실기/데이비드 로빈스 지음, 박진수 옮김, 제이펍, p107 ~ 111