Detecting copied or similar text blocks

Question

I have a bunch of texts about programming in Markdown format. There is a build process that is capable of converting those texts into Word/HTML and also perform simple validation rules like spell checking or checking if document has required header structure. I would like to extend that build code to also check for copy-pasted or similar chunks within all texts.

Is there any existing Java/Groovy library that can help me with that analysis?

My first idea was to use PMD's CopyPasteDetector, but it is too much oriented to analyse real code. I don't see how I can use it to analyse normal text.

Answer 1

I ended up using CPD and Groovy after all. Here is the code if some one is interested:

import net.sourceforge.pmd.cpd.Tokens
import net.sourceforge.pmd.cpd.TokenEntry
import net.sourceforge.pmd.cpd.Tokenizer
import net.sourceforge.pmd.cpd.CPDNullListener
import net.sourceforge.pmd.cpd.MatchAlgorithm
import net.sourceforge.pmd.cpd.SourceCode
import net.sourceforge.pmd.cpd.SourceCode.StringCodeLoader
import net.sourceforge.pmd.cpd.SimpleRenderer

// Prepare empty token data.
TokenEntry.clearImages()
def tokens = new Tokens()

// List all source files with text.
def source = new TreeMap<String, SourceCode>()
new File('.').eachFile { file ->
  if (file.isFile() && file.name.endsWith('.txt')) {
    def analyzedText = file.text
    def sourceCode = new SourceCode(new StringCodeLoader(analyzedText, file.name))
    source.put(sourceCode.fileName, sourceCode)
    analyzedText.eachLine { line, lineNumber ->
      line.split('[\\W\\s\\t\\f]+').each { token ->
        token = token.trim()
        if (token) {
          tokens.add(new TokenEntry(token, sourceCode.fileName, lineNumber + 1))
        }
      }
    }
    tokens.add(TokenEntry.getEOF())
  }
}

// Run matching algorithm.
def maxTokenChain = 15
def matchAlgorithm = new MatchAlgorithm(source, tokens, maxTokenChain, new CPDNullListener())
matchAlgorithm.findMatches()

// Produce report.
matchAlgorithm.matches().each { match ->
  println "  ========================================"
  match.iterator().each { mark ->
    println "  DUPLICATION ERROR: <${mark.tokenSrcID}:${mark.beginLine}> [DUPLICATION] Found a ${match.lineCount} line (${match.tokenCount} tokens) duplication!"
  }
  def indentedTextSlice = ""
  match.sourceCodeSlice.eachLine { line ->
    indentedTextSlice += "  $line\n"
  }
  println "  ----------------------------------------"
  println indentedTextSlice
  println "  ========================================"
}