利用余弦相似度在大量文章中找出抄袭的文章

我前面的2篇文章分别讲了如果利用余弦相似度来判断2篇文章的相似度，来确定文章是否存在抄袭，和余弦相似度的原理，即余弦相似度到底是怎么来判断文章的相似性高低的等等。这一篇再说下，对于文章字数多和大量文章时，如果找到两篇相似度高的文章。这里就需要考虑内存溢出的风险了。所以对第一篇的代码进行了改造。在一定程度上降低了内存溢出的风险。

pom依赖

<dependency><groupId>org.apache.commons</groupId><artifactId>commons-math3</artifactId><version>3.6.1</version>
</dependency>

这里和第一篇略有不同，即第一篇采用的hankcs包实现的余弦相似度算法。本篇文章时通过math3包实现的。但是原理相同。

代码如下：

package com.lsl.config;import org.apache.commons.math3.linear.ArrayRealVector;
import org.apache.commons.math3.linear.RealVector;import java.io.BufferedReader;
import java.io.FileReader;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;
import java.util.*;public class PlagiarismDetector {// 计算余弦相似度public static double cosineSimilarity(RealVector vectorA, RealVector vectorB) {double dotProduct = vectorA.dotProduct(vectorB);double normA = vectorA.getNorm();double normB = vectorB.getNorm();return dotProduct / (normA * normB);}// 将文本转换为词频向量public static Map<String, Integer> textToWordFrequency(String text) {Map<String, Integer> wordFrequency = new HashMap<>();String[] words = text.split("\\s+");for (String word : words) {wordFrequency.put(word, wordFrequency.getOrDefault(word, 0) + 1);}return wordFrequency;}// 将词频映射转换为向量public static RealVector wordFrequencyToVector(Map<String, Integer> wordFrequency, List<String> vocabulary) {double[] vector = new double[vocabulary.size()];for (int i = 0; i < vocabulary.size(); i++) {vector[i] = wordFrequency.getOrDefault(vocabulary.get(i), 0);}return new ArrayRealVector(vector);}// 读取文件内容（流式读取）public static String readFile(String filePath) throws IOException {StringBuilder content = new StringBuilder();try (BufferedReader br = new BufferedReader(new FileReader(filePath))) {String line;while ((line = br.readLine()) != null) {content.append(line).append("\n");}}return content.toString();}// 构建词汇表（增量构建）public static List<String> buildVocabulary(Path papersDir) throws IOException {Set<String> vocabulary = new HashSet<>();Files.list(papersDir).forEach(path -> {try {String content = readFile(path.toString());String[] words = content.split("\\s+");vocabulary.addAll(Arrays.asList(words));} catch (IOException e) {e.printStackTrace();}});return new ArrayList<>(vocabulary);}// 主函数public static void main(String[] args) throws IOException {// 论文文件目录Path papersDir = Paths.get("D:\\codeabc");// 构建词汇表List<String> vocabulary = buildVocabulary(papersDir);// 存储每篇论文的词频向量List<RealVector> vectors = new ArrayList<>();// 逐篇处理论文Files.list(papersDir).forEach(path -> {try {String content = readFile(path.toString());Map<String, Integer> wordFrequency = textToWordFrequency(content);RealVector vector = wordFrequencyToVector(wordFrequency, vocabulary);vectors.add(vector);} catch (IOException e) {e.printStackTrace();}});System.err.println("共有=" + vectors.size() + "文章");// 比较每对论文的相似度for (int i = 0; i < vectors.size(); i++) {for (int j = i + 1; j < vectors.size(); j++) {double similarity = cosineSimilarity(vectors.get(i), vectors.get(j));if (similarity > 0.9) { // 假设相似度大于0.9认为是抄袭System.out.printf("Paper %d and Paper %d are similar with cosine similarity: %.2f%n", i, j, similarity);}}}}
}

运行截图如下：