R语言bibliometrix包主题图关键词重叠问题解决问询

Hey there! I get it—dealing with overlapping keywords in your bibliometrix thematic map can be super frustrating, especially when you’ve already put in the work to extract and analyze your keywords. Let’s walk through some practical fixes to get every label displaying clearly on your network or topic map.

Quick Fixes: Adjust Built-in Parameters #

The easiest place to start is tweaking the arguments directly in the thematicMap() function. These parameters control how labels are rendered:

• Shrink label size: Use the cex argument to reduce font size. Smaller text means less chance of overlap. Example:

  # Reduce label size to 0.7 (default is usually around 1)
  thematicMap(your_data_object, cex = 0.7)
  

• Add space between labels and nodes: The offset parameter moves labels away from their corresponding nodes, creating more breathing room. Pair this with smaller text for better results:

  thematicMap(your_data_object, cex = 0.7, offset = 0.1)
  

• Switch layout algorithms: The default layout might be packing nodes too tightly. Try alternative layouts like "kk" (Kamada-Kawai) or "fr" (Fruchterman-Reingold) to spread nodes out more naturally:

  thematicMap(your_data_object, layout = "kk", cex = 0.7)
  

Advanced Fix: Use Automatic Label Repulsion with ggrepel #

If adjusting built-in parameters isn’t enough, you can leverage the ggrepel package to automatically reposition labels to avoid overlap. Here’s how:

1. First, generate the thematic map as a ggplot object (instead of plotting it immediately):

   # Get the ggplot object without plotting
   tm_plot <- thematicMap(your_data_object, plot = FALSE)
   

2. Load ggrepel and replace the default text layer with geom_text_repel():

   library(ggrepel)
   
   # Update the plot with repelling labels
   tm_plot +
     geom_text_repel(
       aes(label = label),
       size = 3,          # Adjust label size as needed
       box.padding = 0.5, # Add padding around labels
       max.overlaps = 10  # Allow some overlaps only if absolutely necessary
     )
   

This method is highly effective because ggrepel calculates optimal positions for each label in real-time.

Preventative Step: Filter High-Frequency Keywords Only #

If your map has too many keywords (including low-occurrence ones), it’s bound to get crowded. Filter for only the most relevant, high-frequency keywords before generating the map:

# Build a keyword co-occurrence network with only keywords appearing ≥5 times
keyword_net <- biblioNetwork(
  your_data_object,
  analysis = "co-occurrences",
  network = "keywords",
  minDegree = 5  # Adjust this threshold based on your dataset
)

# Generate the thematic map with the filtered network
thematicMap(keyword_net, cex = 0.8)

Start with the quick parameter tweaks first—they’re fast and often solve the problem. If you still see overlap, the ggrepel trick is a game-changer. Let me know if you need help fine-tuning any of these steps!

内容的提问来源于stack exchange，提问作者Bayuo Blaise