如何在R语言中为数据集添加关联原数据行的情感分析列

Got it, let's solve this problem step by step—getting per-row sentiment scores tied back to your original dataset is totally doable with a couple of solid R packages. Here's how to pull it off, tailored to your data structure (case IDs + text):

First, let's assume your imported data looks like this (replace with your actual BI-loaded data):

# 加载必备工具包
library(tidyverse)

# 模拟你的数据结构（实际用read_csv/read.table导入BI数据）
your_data <- tibble(
  case_id = c("12345", "67890", "13579"),
  text = c("I am so happy with the service today!", 
           "This product is terrible, I'm really disappointed.",
           "It's okay, not great but not bad either.")
)

方法1：使用tidytext（灵活支持多种情感词典） #

tidytext lets you use well-known sentiment dictionaries (like AFINN for numeric scores, or BING for positive/negative categories) and easily tie results back to your original rows.

选项A：获取数值型情感分数（AFINN词典） #

This gives you a continuous score where positive values mean positive sentiment, negative mean negative:

library(tidytext)

# 拆分文本为单词，匹配AFINN词典，按case_id求和得到每行情感分
sentiment_scores <- your_data %>%
  unnest_tokens(word, text) %>%  # 把每行文本拆成单个单词
  inner_join(get_sentiments("afinn")) %>%  # 匹配情感词典
  group_by(case_id) %>%
  summarise(sentiment_score = sum(value))  # 按案例ID求和得到该行总情感分

# 把情感分合并回原数据集
your_data_with_sentiment <- your_data %>%
  left_join(sentiment_scores, by = "case_id") %>%
  replace_na(list(sentiment_score = 0))  # 没有匹配到情感词的行填充0

选项B：获取情感分类（BING词典） #

If you prefer categorical results (positive/negative/neutral) instead of numeric scores:

sentiment_categories <- your_data %>%
  unnest_tokens(word, text) %>%
  inner_join(get_sentiments("bing")) %>%  # 匹配BING的正负词
  group_by(case_id, sentiment) %>%
  count() %>%
  pivot_wider(names_from = sentiment, values_from = n, values_fill = 0) %>%
  # 根据正负词数量判断最终分类
  mutate(sentiment_category = case_when(
    positive > negative ~ "positive",
    negative > positive ~ "negative",
    TRUE ~ "neutral"
  )) %>%
  select(case_id, sentiment_category)

# 合并回原数据
your_data_with_sentiment <- your_data %>%
  left_join(sentiment_categories, by = "case_id") %>%
  replace_na(list(sentiment_category = "neutral"))

方法2：使用sentimentr（智能处理上下文） #

If you want better handling of context (like negations: "not happy" gets recognized as negative, which basic word-based methods miss), sentimentr is perfect—it calculates sentiment directly on full lines of text without splitting words:

library(sentimentr)

# 直接给每行文本计算情感分数，自动关联case_id
your_data_with_sentiment <- your_data %>%
  mutate(
    sentiment_score = sentiment_by(text, by = case_id)$ave_sentiment,
    # 可选：把分数转成分类
    sentiment_category = case_when(
      sentiment_score > 0 ~ "positive",
      sentiment_score < 0 ~ "negative",
      TRUE ~ "neutral"
    )
  )

关键注意事项 #

• 数据导入: When loading your BI data, make sure case_id is treated as a character (not numeric) to avoid merging issues. Use read_csv("your_data.csv", col_types = cols(case_id = col_character())) if needed.
• 缺失值: Rows with no sentiment-related words will get NA scores—use replace_na() to fill these with 0 (for scores) or "neutral" (for categories) as shown above.
• 词典选择: AFINN gives nuanced numeric scores, BING is simple for binary classification, and sentimentr handles context best. Pick the one that fits your use case!

内容的提问来源于stack exchange，提问作者Josh