java - 获取 mallet 中所有文档的实例和主题序列

Question

我正在使用 mallet 库进行主题建模。我的数据集位于 filePath 路径中，并且 csvIterator 似乎可以读取数据，因为 model.getData() 有大约 27000 行，等于我的数据集。我编写了一个循环，打印 10 个第一个文档的实例和主题序列，但标记的大小为 0。我哪里出错了？

在下面，我想显示主题中前 5 个单词以及第 10 个文档的比例，但所有输出都是相同的。

控制台中的输出示例:

----文档0

0 0.200 com (1723) twitter (1225) http (871) cbr (688) 堪培拉 (626)

1 0.200 com (981) twitter (901) 天 (205) 五月 (159) 周三 (156)

2 0.200 twitter (1068) com (947) act (433) actvcc (317) 堪培拉 (302)

3 0.200 http (1039) 堪培拉 (841) 工作 (378) dlvr (313) com (228)

4 0.200 com (1185) www (1074) http (831) 新闻 (708) 堪培拉时报 (560)

----文档1

0 0.200 com (1723) twitter (1225) http (871) cbr (688) 堪培拉 (626)

1 0.200 com (981) twitter (901) 天 (205) 五月 (159) 周三 (156)

2 0.200 twitter (1068) com (947) act (433) actvcc (317) 堪培拉 (302)

3 0.200 http (1039) 堪培拉 (841) 工作 (378) dlvr (313) com (228)

4 0.200 com (1185) www (1074) http (831) 新闻 (708) 堪培拉时报 (560)

据我所知，LDA 模型生成每个文档并将它们分配给主题的单词。那么为什么每个文件的结果都是一样的呢？

ArrayList<Pipe> pipeList = new ArrayList<Pipe>();
   pipeList.add(new CharSequenceLowercase());
    pipeList.add(new CharSequence2TokenSequence(Pattern.compile("\\p{L}[\\p{L}\\p{P}]+\\p{L}")));
    //stoplists/en.txt
    pipeList.add(new TokenSequenceRemoveStopwords(new File(pathStopWords), "UTF-8", false, false, false));
    pipeList.add(new TokenSequence2FeatureSequence());

    InstanceList instances = new InstanceList(new SerialPipes(pipeList));

    Reader fileReader = new InputStreamReader(new FileInputStream(new File(filePath)), "UTF-8");
//header of my data set
// row,location,username,hashtaghs,text,retweets,date,favorites,numberOfComment
    CsvIterator csvIterator = new CsvIterator(fileReader,
            Pattern.compile("^(\\d+)[,]*[^,]*[,]*[^,]*[,]*[^,]*[,]*([^,]*)[,]*[^,]*[,]*[^,]*[,]*[^,]*[,]*[^,]*$"),
            2, 0, 1);
    instances.addThruPipe(csvIterator); // data, label, name fields

    int numTopics = 5;
    ParallelTopicModel model = new ParallelTopicModel(numTopics, 1.0, 0.01);

    model.addInstances(instances);

    model.setNumThreads(2);


    model.setNumIterations(50);
    model.estimate();

    Alphabet dataAlphabet = instances.getDataAlphabet();
    ArrayList<TopicAssignment> arrayTopics = model.getData();

    for (int i = 0; i < 10; i++) {
        System.out.println("---- document " + i);
        FeatureSequence tokens = (FeatureSequence) model.getData().get(i).instance.getData();
        LabelSequence topics = model.getData().get(i).topicSequence;

        Formatter out = new Formatter(new StringBuilder(), Locale.US);
        for (int position = 0; position < tokens.getLength(); position++) {
            out.format("%s-%d ", dataAlphabet.lookupObject(tokens.getIndexAtPosition(position)),
                    topics.getIndexAtPosition(position));
        }
        System.out.println(out);

        double[] topicDistribution = model.getTopicProbabilities(i);

        ArrayList<TreeSet<IDSorter>> topicSortedWords = model.getSortedWords();


        for (int topic = 0; topic < numTopics; topic++) {
            Iterator<IDSorter> iterator = topicSortedWords.get(topic).iterator();
            out = new Formatter(new StringBuilder(), Locale.US);
            out.format("%d\t%.3f\t", topic, topicDistribution[topic]);
            int rank = 0;
            while (iterator.hasNext() && rank < 5) {
                IDSorter idCountPair = iterator.next();
                out.format("%s (%.0f) ", dataAlphabet.lookupObject(idCountPair.getID()), idCountPair.getWeight());
                rank++;
            }
            System.out.println(out);
        }

        StringBuilder topicZeroText = new StringBuilder();
        Iterator<IDSorter> iterator = topicSortedWords.get(0).iterator();

        int rank = 0;
        while (iterator.hasNext() && rank < 5) {
            IDSorter idCountPair = iterator.next();
            topicZeroText.append(dataAlphabet.lookupObject(idCountPair.getID()) + " ");
            rank++;
        }

    }

Answer 1

主题是在模型级别定义的，而不是在文档级别定义的。它们对于所有人来说都应该是相同的。

看起来您的所有文本都是网址。将 PrintInputPipe 添加到导入序列可能有助于调试。