Examples with structures._ParentDoc structures._ParentDoc used on opensource projects

Example 21 with structures._ParentDoc

use of structures._ParentDoc in project IR_Base by Linda-sunshine.

the class languageModelBaseLine method rankChild4StnByLanguageModel.

protected HashMap<String, Double> rankChild4StnByLanguageModel(_Stn stnObj, _ParentDoc pDoc) {
    HashMap<String, Double> childLikelihoodMap = new HashMap<String, Double>();
    double smoothingMu = 1000;
    for (_ChildDoc cDoc : pDoc.m_childDocs) {
        int cDocLen = cDoc.getTotalDocLength();
        _SparseFeature[] fv = cDoc.getSparse();
        double stnLogLikelihood = 0;
        double alphaDoc = smoothingMu / (smoothingMu + cDocLen);
        _SparseFeature[] sv = stnObj.getFv();
        for (_SparseFeature svWord : sv) {
            double featureLikelihood = 0;
            int wid = svWord.getIndex();
            double stnVal = svWord.getValue();
            int featureIndex = Utils.indexOf(fv, wid);
            double docVal = 0;
            if (featureIndex != -1) {
                docVal = fv[featureIndex].getValue();
            }
            double smoothingProb = (1 - alphaDoc) * docVal / (cDocLen);
            smoothingProb += alphaDoc * getReferenceProb(wid);
            featureLikelihood = Math.log(smoothingProb);
            stnLogLikelihood += stnVal * featureLikelihood;
        }
        childLikelihoodMap.put(cDoc.getName(), stnLogLikelihood);
    }
    return childLikelihoodMap;
}

Example 22 with structures._ParentDoc

use of structures._ParentDoc in project IR_Base by Linda-sunshine.

the class outputFile method outputFiles.

public static void outputFiles(String filePrefix, _Corpus c) {
    try {
        String selectedSentencesinParentFile = filePrefix + "/selected_Stn.txt";
        String selectedCommentsFile = filePrefix + "/selected_Comments.txt";
        String sctmFormatParentFile = filePrefix + "/abagf.AT.txt";
        String sctmFormatChildFile = filePrefix + "/cbagf.AT.txt";
        String sctmWordFile = filePrefix + "/words.AT.txt";
        String stnLengthFile = filePrefix + "/selected_StnLength.txt";
        String shortStnFile = filePrefix + "/selected_ShortStn.txt";
        String longStnFile = filePrefix + "/selected_LongStn.txt";
        if (c.getFeatureSize() != 0) {
            PrintWriter wordPW = new PrintWriter(new File(sctmWordFile));
            for (int i = 0; i < c.getFeatureSize(); i++) {
                String wordName = c.getFeature(i);
                wordPW.println(wordName);
            }
            wordPW.flush();
            wordPW.close();
        }
        PrintWriter stnLengthPW = new PrintWriter(new File(stnLengthFile));
        PrintWriter shortParentPW = new PrintWriter(new File(shortStnFile));
        PrintWriter longParentPW = new PrintWriter(new File(longStnFile));
        PrintWriter parentPW = new PrintWriter(new File(selectedSentencesinParentFile));
        PrintWriter childPW = new PrintWriter(new File(selectedCommentsFile));
        PrintWriter sctmParentPW = new PrintWriter(new File(sctmFormatParentFile));
        PrintWriter sctmChildPW = new PrintWriter(new File(sctmFormatChildFile));
        int totoalParentNum = 0;
        TreeMap<Integer, _ParentDoc> parentMap = new TreeMap<Integer, _ParentDoc>();
        int totalStnNum = 0;
        ArrayList<_Doc> m_trainSet = c.getCollection();
        ArrayList<Integer> parentNameList = new ArrayList<Integer>();
        for (_Doc d : m_trainSet) {
            if (d instanceof _ParentDoc) {
                // HashMap<Integer, _Stn> stnMap = ((_ParentDoc)
                // d).m_sentenceMap;
                totoalParentNum += 1;
                String parentName = d.getName();
                parentMap.put(Integer.parseInt(parentName), (_ParentDoc) d);
                parentNameList.add(Integer.parseInt(parentName));
            }
        }
        ArrayList<Double> parentDocLenList = new ArrayList<Double>();
        ArrayList<Double> childDocLenList = new ArrayList<Double>();
        double parentDocLenSum = 0;
        double childDocLenSum = 0;
        for (int parentID : parentMap.keySet()) {
            _ParentDoc parentObj = parentMap.get(parentID);
            double parentDocLen = parentObj.getTotalDocLength();
            parentDocLenSum += parentDocLen;
            parentDocLenList.add(parentDocLen);
            for (_ChildDoc cDoc : parentObj.m_childDocs) {
                double childDocLen = cDoc.getTotalDocLength();
                childDocLenList.add(childDocLen);
                childDocLenSum += childDocLen;
            }
            _Stn[] sentenceArray = parentObj.getSentences();
            int selectedStn = 0;
            for (int i = 0; i < sentenceArray.length; i++) {
                _Stn stnObj = sentenceArray[i];
                if (stnObj == null)
                    continue;
                selectedStn += 1;
                stnLengthPW.println(stnObj.getLength());
            // if(stnObj==null)
            // continue;
            // selectedStn += 1;
            }
            totalStnNum += selectedStn;
            parentPW.print(parentID + "\t" + selectedStn + "\t");
            shortParentPW.print(parentID + "\t");
            longParentPW.print(parentID + "\t");
            for (int i = 0; i < sentenceArray.length; i++) {
                _Stn stnObj = sentenceArray[i];
                if (stnObj == null)
                    continue;
                if (stnObj.getLength() < 15)
                    shortParentPW.print((stnObj.getIndex() + 1) + "\t");
                else
                    longParentPW.print((stnObj.getIndex() + 1) + "\t");
                parentPW.print((stnObj.getIndex() + 1) + "\t");
            }
            parentPW.println();
            longParentPW.println();
            shortParentPW.println();
        }
        System.out.println("longest child\t" + Collections.max(childDocLenList));
        System.out.println("shortest child\t" + Collections.min(childDocLenList));
        System.out.println("parent doc len\t" + parentDocLenSum / parentDocLenList.size());
        System.out.println("child doc len\t" + childDocLenSum / childDocLenList.size());
        parentPW.flush();
        parentPW.close();
        stnLengthPW.flush();
        stnLengthPW.close();
        shortParentPW.flush();
        shortParentPW.close();
        longParentPW.flush();
        longParentPW.close();
        sctmParentPW.println(totoalParentNum);
        sctmChildPW.println(totoalParentNum);
        System.out.println("stnNum" + totalStnNum);
        for (int parentID : parentMap.keySet()) {
            _ParentDoc d = parentMap.get(parentID);
            // HashMap<Integer, _Stn> stnMap = ((_ParentDoc)
            // d).m_sentenceMap;
            _Stn[] sentenceArray = (d).getSentences();
            int selectedStn = 0;
            for (int i = 0; i < sentenceArray.length; i++) {
                _Stn stnObj = sentenceArray[i];
                if (stnObj == null)
                    continue;
                selectedStn += 1;
            }
            sctmParentPW.println(selectedStn);
            for (int i = 0; i < sentenceArray.length; i++) {
                _Stn stnObj = sentenceArray[i];
                if (stnObj == null)
                    continue;
                _SparseFeature[] sv = stnObj.getFv();
                sctmParentPW.print((int) stnObj.getLength() + "\t");
                for (int j = 0; j < sv.length; j++) {
                    int index = sv[j].getIndex();
                    double value = sv[j].getValue();
                    for (int v = 0; v < value; v++) sctmParentPW.print(index + "\t");
                }
                sctmParentPW.println();
            }
            ArrayList<_ChildDoc> childDocs = ((_ParentDoc) d).m_childDocs;
            sctmChildPW.println(childDocs.size());
            String parentName = d.getName();
            TreeMap<Integer, _ChildDoc> childMap = new TreeMap<Integer, _ChildDoc>();
            for (_ChildDoc cDoc : childDocs) {
                String childName = cDoc.getName();
                int childID = Integer.parseInt(childName.replace(parentName + "_", ""));
                childMap.put(childID, cDoc);
            }
            childPW.print(parentName + "\t");
            for (int t : childMap.keySet()) {
                _ChildDoc cDoc = childMap.get(t);
                sctmChildPW.print((int) cDoc.getTotalDocLength() + "\t");
                childPW.print(cDoc.getName() + "\t");
                // System.out.println(cDoc.getName() + "\t");
                _SparseFeature[] fv = cDoc.getSparse();
                for (int j = 0; j < fv.length; j++) {
                    int index = fv[j].getIndex();
                    double value = fv[j].getValue();
                    for (int v = 0; v < value; v++) {
                        sctmChildPW.print(index + "\t");
                    }
                }
                sctmChildPW.println();
            }
            childPW.println();
        }
        sctmParentPW.flush();
        sctmParentPW.close();
        sctmChildPW.flush();
        sctmChildPW.close();
        childPW.flush();
        childPW.close();
    } catch (Exception e) {
        e.printStackTrace();
    }
}

Example 23 with structures._ParentDoc

use of structures._ParentDoc in project IR_Base by Linda-sunshine.

the class ACCTM method sampleInParentDoc.

protected void sampleInParentDoc(_Doc d) {
    _ParentDoc pDoc = (_ParentDoc) d;
    int wid, tid;
    double normalizedProb;
    for (_Word w : pDoc.getWords()) {
        wid = w.getIndex();
        tid = w.getTopic();
        pDoc.m_sstat[tid]--;
        if (m_collectCorpusStats) {
            word_topic_sstat[tid][wid]--;
            m_sstat[tid]--;
        }
        normalizedProb = 0;
        for (tid = 0; tid < number_of_topics; tid++) {
            double pWordTopic = parentWordByTopicProb(tid, wid);
            double pTopicPDoc = parentTopicInDocProb(tid, pDoc);
            double pTopicCDoc = parentChildInfluenceProb(tid, pDoc);
            m_topicProbCache[tid] = pWordTopic * pTopicPDoc * pTopicCDoc;
            normalizedProb += m_topicProbCache[tid];
        }
        normalizedProb *= m_rand.nextDouble();
        for (tid = 0; tid < number_of_topics; tid++) {
            normalizedProb -= m_topicProbCache[tid];
            if (normalizedProb <= 0)
                break;
        }
        if (tid == number_of_topics)
            tid--;
        w.setTopic(tid);
        pDoc.m_sstat[tid]++;
        if (m_collectCorpusStats) {
            word_topic_sstat[tid][wid]++;
            m_sstat[tid]++;
        }
    }
}

Example 24 with structures._ParentDoc

use of structures._ParentDoc in project IR_Base by Linda-sunshine.

the class ACCTM method computeMu4Doc.

protected void computeMu4Doc(_ChildDoc d) {
    _ParentDoc tempParent = d.m_parentDoc;
    double mu = Utils.cosine(tempParent.getSparse(), d.getSparse());
    d.setMu(mu);
}

Example 25 with structures._ParentDoc

use of structures._ParentDoc in project IR_Base by Linda-sunshine.

the class ACCTM method parentChildInfluenceProb.

protected double parentChildInfluenceProb(int tid, _ParentDoc pDoc) {
    double term = 1.0;
    if (tid == 0)
        return term;
    double topicSum = Utils.sumOfArray(pDoc.m_sstat);
    for (_ChildDoc cDoc : pDoc.m_childDocs) {
        double muDp = cDoc.getMu() / topicSum;
        term *= gammaFuncRatio((int) cDoc.m_sstat[tid], muDp, d_alpha + pDoc.m_sstat[tid] * muDp) / gammaFuncRatio((int) cDoc.m_sstat[0], muDp, d_alpha + pDoc.m_sstat[0] * muDp);
    }
    return term;
}