classifier.Rmd

---
title: "songs"
output: html_document
---

```{r setup, include=FALSE}
knitr::opts_chunk$set(echo = TRUE)
```


```{r message=FALSE, warning=FALSE, error=FALSE}
library(jiebaR)
library(stringr)

# Get txt file paths
fps_m <- list.files('male', full.names = T)
fps_f <- list.files('female', full.names = T)

# Initialize jiebaR
seg <- worker()


# determining lyrics
lyric_m <- vector('character', length(fps_m))

for (i in seq_along(fps_m)) {
  l_m <- readLines(fps_m[i], encoding = 'UTF-8') %>% str_squish()
  segged_m <- segment(l_m, seg)
  lyric_m[i] <- paste0(segged_m, collapse = ' ')
}

lyric_f <- vector('character', length(fps_f))

for (i in seq_along(fps_f)) {
  l_f <- readLines(fps_f[i], encoding = 'UTF-8') %>% str_squish()
  segged_f <- segment(l_f, seg)
  lyric_f[i] <- paste0(segged_f, collapse = ' ')
}

# Combine results into dfs
male_df <- tibble::tibble(性別 = '男', 歌詞 = lyric_m)
female_df <- tibble::tibble(性別 = '女', 歌詞 = lyric_f)

```



```{r message=FALSE, warning=FALSE}
library(quanteda)
library(caret)
library(dplyr)

both_df <- rbind(male_df, female_df) %>% mutate(編號 = 1:(nrow(male_df)+nrow(female_df)))
songs_corpus <- corpus(both_df, text_field = '歌詞')

set.seed(500)
train_id <- sample(1:nrow(both_df), 20000, replace = FALSE)
head(train_id, 10)

# get training set

songs_train_df <- filter(both_df, 編號 %in% train_id)
songs_train_dtm <- corpus(songs_train_df, text_field = '歌詞') %>%
  dfm(remove = stopwords(), stem = T)

# get test set (documents not in id_train)
songs_test_df <- filter(both_df, !編號 %in% train_id)
songs_test_dtm <- corpus(songs_test_df, text_field = '歌詞') %>%
  dfm(remove = stopwords(), stem = T)

# train cassifier model
var <- songs_train_df$性別
var <- as.factor(var)
tmod_song <- textmodel_nb(songs_train_dtm, var)

# inspect performance
songs_matched <- dfm_match(songs_test, features = featnames(songs_training))

actual_class <- songs_test_df$性別
predicted_class <- predict(tmod_song, newdata = songs_matched)
tab_class <- table(actual_class, predicted_class)
tab_class

confusionMatrix(tab_class, mode = 'everything')
```