colMassPlot()

Visualize Temporal Dimensions of Corpus Structure

The colMassPlot() function creates a multi-line plot showing how four key corpus metrics evolve over time. This provides crucial insights into corpus growth patterns, data collection trends, and temporal coverage.

πŸ”Ή Function Definition

colMassPlot(
  data,
  sc = c(1, 10, 10, 1),
  r = 1,
  textty = "text",
  themety = "light",
  size_b = 2.5,
  x_lab = "years"
)

🎯 Purpose

Understanding the temporal structure of your corpus is essential for interpreting frequency trends. The colMassPlot() function helps you:

  1. Visualize corpus growth β€” See how the corpus expands over time
  2. Identify collection patterns β€” Detect periods of intensive/sparse data collection
  3. Assess temporal balance β€” Evaluate whether time periods are comparably represented
  4. Detect anomalies β€” Spot unusual spikes or drops in corpus size
  5. Contextualize term frequencies β€” Understand how corpus size affects frequency patterns
  6. Validate data quality β€” Ensure corpus metadata is consistent and complete

This function is typically used after importData() and alongside rowMassPlot() for comprehensive data exploration before normalization.

βš™οΈ Arguments

Argument Type Default Description
data List required A list object returned by importData(), containing the TDM and corpus metadata.
sc Numeric vector c(1, 10, 10, 1) Scaling factors for the four metrics (in order):
1. nDoc (number of documents)
2. dimCorpus (total tokens)
3. Csum (sum of keyword frequencies)
4. Mcf (maximum keyword frequency)
Adjust to make lines visually comparable.
r Integer 1 Interval for x-axis label thinning. r = 2 shows every 2nd year, r = 5 shows every 5th year, etc. Useful for long time series.
textty Character "text" Label for the unit of analysis in the legend (e.g., "text", "document", "paper", "article").
themety Character "light" Visual theme for the plot:
β€’ "light": Light background (default)
β€’ "dark": Dark background
size_b Numeric 2.5 Line thickness for the plot. Increase for bolder lines, decrease for finer lines.
x_lab Character "years" Label for the x-axis (e.g., "Years", "Time Period", "Publication Year").

πŸ“Š Corpus Metrics Explained

1. nDoc β€” Number of Documents

The count of documents/texts in the corpus for each time period.

Insights: - Shows data collection intensity - Reveals archival coverage patterns - Indicates periods of high/low publication activity

2. dimCorpus β€” Total Tokens

The total number of words/tokens in all documents for each time period.

Insights: - Represents overall corpus size - Important for normalization decisions - Shows writing/publication volume trends

3. Csum β€” Sum of Keyword Frequencies

The total frequency of all keywords across all documents in each time period.

Insights: - Indicates keyword density in the corpus - Shows overall vocabulary coverage - Helps assess keyword selection adequacy

4. Mcf β€” Maximum Keyword Frequency

The highest frequency of any single keyword in each time period.

Insights: - Identifies periods dominated by specific terms - Shows potential outliers or dominant topics - Indicates vocabulary concentration patterns

🎨 Understanding Scaling Factors

Different metrics have vastly different scales: - nDoc might range from 10-100 - dimCorpus might range from 10,000-1,000,000 - Csum might range from 5,000-500,000 - Mcf might range from 50-5,000

Scaling makes visual comparison possible by bringing all metrics to similar ranges.

Default scaling: c(1, 10, 10, 1) - nDoc: Γ—1 (no scaling) - dimCorpus: Γ—10 (reduces by factor of 10) - Csum: Γ—10 (reduces by factor of 10) - Mcf: Γ—1 (no scaling)

Adjust scaling if lines are too far apart or overlapping too much.

πŸ“¦ Output

Returns a ggplot2 object with the following characteristics:

Element Description
X-axis Time periods (years) from the corpus metadata
Y-axis Rescaled values of the four corpus metrics
Lines Four colored lines representing each metric’s temporal evolution
Legend Shows metric names with scaling factors (e.g., β€œnDoc (Γ—1)”, β€œdimCorpus (Γ—10)”)
Theme Light or dark background based on themety parameter

πŸ’‘ Usage Examples

Basic Usage

library(cccc)

# Import data
corpus <- importData("tdm.csv", "corpus_info.csv")

# Create temporal plot with default settings
colMassPlot(corpus)

πŸ” Interpreting the Plot

Common Patterns

1. Parallel Growth

All four lines increase proportionally over time.

πŸ“ˆ All metrics ↗️

Interpretation: Consistent corpus expansion with stable composition

3. Spikes or Drops

Sudden changes in one or more metrics.

πŸ“ˆ nDoc: sudden spike in specific year

Interpretation: Intensive data collection period or archival event

4. Plateau Patterns

Metrics level off after initial growth.

πŸ“ˆ Early growth β†’ β†’ β†’ flat

Interpretation: Complete historical coverage or saturation

🎯 What to Look For

1. Temporal Coverage

  • Are all time periods equally represented?
  • Are there gaps or sparse periods?
  • Does coverage increase toward recent years?

2. Growth Patterns

  • Linear growth (steady increase)
  • Exponential growth (accelerating increase)
  • Irregular patterns (varying collection intensity)

3. Proportionality

  • Do nDoc and dimCorpus grow together? (expected)
  • Does Csum follow dimCorpus? (indicates keyword coverage)
  • Are there periods where Mcf spikes? (term dominance)

4. Anomalies

  • Sudden drops (potential data quality issues)
  • Isolated spikes (special events or archival additions)
  • Plateaus (collection boundaries)

πŸ“ˆ Use Cases

1. Data Quality Assessment

Verify that corpus metadata is complete and consistent across time periods.

2. Normalization Decision

Determine whether normalization is necessary based on corpus size variation.

3. Historical Context

Understand how corpus collection reflects historical publication/archival patterns.

4. Methodology Documentation

Create figures for research papers showing corpus characteristics.

5. Comparative Studies

Compare temporal structures of different corpora or subcorpora.

πŸ’‘ Tips & Best Practices

  1. Always check this plot before normalization to understand corpus growth patterns
  2. Experiment with scaling to find the most informative visualization
  3. Compare with rowMassPlot() for comprehensive data exploration
  4. Document temporal patterns in your methodology section
  5. Use r parameter for long time series (>30 years) to avoid label clutter
  6. Save high-resolution versions for publications
  7. Consider normalization if you see large variations in corpus size across time

πŸ“š See Also

  • importData() β€” Import corpus data and metadata
  • rowMassPlot() β€” Visualize keyword frequency distribution
  • normalization() β€” Normalize frequencies (often needed when corpus size varies)
  • curvePlot() β€” Visualize individual keyword trajectories