如何在R中基于openNLP词性标注统计各词性数量？

Great job getting the POS tagging pipeline set up with openNLP! Let's turn those annotated words into the category counts you need for each observation—here's a straightforward breakdown:

Step 1: Extract POS Tags from Your Annotations #

First, we'll pull the POS tags out of the features column in your a3w object. Since each entry in features is a list, we can use sapply to grab the POS value for every word:

# Extract POS tags from word annotations
pos_tags <- sapply(a3w$features, function(x) x$POS)

Step 2: Define POS Tag Groups #

We need to map standard Penn Treebank POS tags to the categories you care about. Here are the common tags for nouns, verbs, and adjectives:

# Define tag sets for each category
noun_tags <- c("NN", "NNS", "NNP", "NNPS")  # Singular/plural common/proper nouns
verb_tags <- c("VB", "VBD", "VBG", "VBN", "VBP", "VBZ")  # All verb tenses/forms
adj_tags <- c("JJ", "JJR", "JJS")  # Positive/comparative/superlative adjectives

Step 3: Calculate Counts #

Now count how many tags fall into each category:

# Compute counts for each POS category
noun_count <- sum(pos_tags %in% noun_tags)
verb_count <- sum(pos_tags %in% verb_tags)
adj_count <- sum(pos_tags %in% adj_tags)

# View results
cat("Nouns:", noun_count, "\nVerbs:", verb_count, "\nAdjectives:", adj_count)

Step 4: Scale to a Dataset (For Multiple Observations) #

Since you need counts per observation, wrap this logic into a function you can apply to every text entry in your dataset:

# Function to compute POS counts for a single text string
count_pos_features <- function(text) {
  # Convert text to openNLP-compatible String
  s <- as.String(text)
  
  # Run full annotation pipeline
  sent_annotator <- Maxent_Sent_Token_Annotator()
  word_annotator <- Maxent_Word_Token_Annotator()
  pos_annotator <- Maxent_POS_Tag_Annotator()
  
  a2 <- annotate(s, list(sent_annotator, word_annotator))
  a3 <- annotate(s, pos_annotator, a2)
  
  # Subset to only word annotations
  a3w <- subset(a3, type == "word")
  
  # Extract POS tags
  pos_tags <- sapply(a3w$features, function(x) x$POS)
  
  # Define tag groups
  noun_tags <- c("NN", "NNS", "NNP", "NNPS")
  verb_tags <- c("VB", "VBD", "VBG", "VBN", "VBP", "VBZ")
  adj_tags <- c("JJ", "JJR", "JJS")
  
  # Return counts as a data frame row
  data.frame(
    noun_count = sum(pos_tags %in% noun_tags),
    verb_count = sum(pos_tags %in% verb_tags),
    adj_count = sum(pos_tags %in% adj_tags),
    stringsAsFactors = FALSE
  )
}

# Example: Apply to a dataset with multiple texts
your_dataset <- data.frame(
  text = c(
    "Pierre Vinken, 61 years old, will join the board as a nonexecutive director Nov. 29.",
    "Mr. Vinken is chairman of Elsevier N.V., the Dutch publishing group."
  ),
  stringsAsFactors = FALSE
)

# Add POS counts to the original dataset
pos_counts <- do.call(rbind, lapply(your_dataset$text, count_pos_features))
your_dataset_with_counts <- cbind(your_dataset, pos_counts)

# View the final result
print(your_dataset_with_counts)

This will give you a data frame where each row includes the original text plus the three count variables you need for your analysis.

内容的提问来源于stack exchange，提问作者Ceri