Package edu.gslis.ttg.main

Source Code of edu.gslis.ttg.main.RunTTGBaseline

package edu.gslis.ttg.main;

import java.io.FileReader;
import java.io.PrintStream;
import java.io.UnsupportedEncodingException;
import java.text.DecimalFormat;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.HashSet;
import java.util.Iterator;
import java.util.List;
import java.util.Map;
import java.util.Set;

import org.apache.thrift.TException;
import org.json.simple.JSONArray;
import org.json.simple.JSONObject;
import org.json.simple.parser.JSONParser;

import cc.twittertools.search.api.TrecSearchThriftClient;
import cc.twittertools.thrift.gen.TResult;
import edu.gslis.eval.Qrels;
import edu.gslis.queries.GQueries;
import edu.gslis.queries.GQueriesJsonImpl;
import edu.gslis.queries.GQuery;
import edu.gslis.ttg.clusters.Cluster;
import edu.gslis.ttg.clusters.Clusters;
import edu.gslis.ttg.clusters.clusterers.SimpleJaccardClusterer;
import edu.gslis.ttg.searchers.SimpleSearcher;
import edu.gslis.utils.ParameterBroker;

public class RunTTGBaseline {
  private static final String DEFAULT_RUNTAG = "ttgbaseline";
 
  private static final String HOST_OPTION = "host";
  private static final String TRAINING_PORT = "training_port";
  private static final String TESTING_PORT = "testing_port";
  private static final String QUERIES_OPTION = "queries";
  private static final String NUM_RESULTS_OPTION = "num_results";
  private static final String GROUP_OPTION = "group";
  private static final String TOKEN_OPTION = "token";
  private static final String RUNTAG_OPTION = "runtag";
  private static final String JACCARD_STEP_SIZE = "jaccard_step";
  private static final String TRAINING_QUERIES = "training_queries";
  private static final String TRAINING_CLUSTERS = "training_clusters";
  private static final String QRELS_OPTION = "qrels";
  private static final String EVALUATION_OPTION = "evaluation_type";

  public static void main(String[] args) throws NumberFormatException, TException, UnsupportedEncodingException {
    ParameterBroker params = new ParameterBroker("./config/run_params.json");

    PrintStream out = new PrintStream(System.out, true, "UTF-8");
    PrintStream err = new PrintStream(System.err, true, "UTF-8");
   
    GQueries trainingQueries = new GQueriesJsonImpl();
    trainingQueries.setMetadataField("querytweettime");
    trainingQueries.read(params.getParamValue(TRAINING_QUERIES));
   
    GQueries queries = new GQueriesJsonImpl();
    queries.setMetadataField("querytweettime");
    queries.read(params.getParamValue(QUERIES_OPTION));
   
    Qrels qrels = new Qrels(params.getParamValue(QRELS_OPTION), false, 1);
   
    // max number of docs to send to output
    int numResults = 1000;
    try {
      if (params.getParamValue(NUM_RESULTS_OPTION) != null) {
        numResults = Integer.parseInt(params.getParamValue(NUM_RESULTS_OPTION));
      }
    } catch (NumberFormatException e) {
      err.println("Invalid " + NUM_RESULTS_OPTION + ": " + params.getParamValue(NUM_RESULTS_OPTION));
      System.exit(-1);
    }
   
    // authentication credentials
    String group = params.getParamValue(GROUP_OPTION);
    if(group==null) {
      err.println("Invalid " + GROUP_OPTION + ": must set a valid group ID");
      System.exit(-1);
    }
    String token = params.getParamValue(TOKEN_OPTION);
    if(group==null) {
      err.println("Invalid " + TOKEN_OPTION + ": must set a valid authentication token");
      System.exit(-1);
    }
   
    // ports
    int trainingPort = 9090;
    try {
      if (params.getParamValue(TRAINING_PORT) != null) {
        trainingPort = Integer.parseInt(params.getParamValue(TRAINING_PORT));
      }
    } catch (NumberFormatException e) {
      err.println("Invalid " + TRAINING_PORT + ": " + params.getParamValue(TRAINING_PORT));
      System.exit(-1);
    }
    int testingPort = 9091;
    try {
      if (params.getParamValue(TESTING_PORT) != null) {
        testingPort = Integer.parseInt(params.getParamValue(TESTING_PORT));
      }
    } catch (NumberFormatException e) {
      err.println("Invalid " + TESTING_PORT + ": " + params.getParamValue(TESTING_PORT));
      System.exit(-1);
    }
   
    // run tag
    String runTag = params.getParamValue(RUNTAG_OPTION);
    if(runTag==null) {
      runTag = DEFAULT_RUNTAG;
    }
   
    // jaccard step size
    double stepSize = 0.1;
    try {
      if (params.getParamValue(JACCARD_STEP_SIZE) != null) {
        stepSize = Double.parseDouble(params.getParamValue(JACCARD_STEP_SIZE));
      }
    } catch (NumberFormatException e) {
      err.println("Invalid " + JACCARD_STEP_SIZE + ": " + params.getParamValue(JACCARD_STEP_SIZE));
      System.exit(-1);
    }
   
    // weighted or unweighted evaluation
    String evalType = "weighted";
    try {
      if (params.getParamValue(EVALUATION_OPTION) != null) {
        evalType = params.getParamValue(EVALUATION_OPTION);
      }
    } catch (Exception e) {
      err.println("Invalid " + EVALUATION_OPTION + ": " + params.getParamValue(EVALUATION_OPTION));
      System.exit(-1);
    }
   
    // 2 decimal places
    DecimalFormat df = new DecimalFormat("#.##");
   
    // read in training data
    String trainingFile = params.getParamValue(TRAINING_CLUSTERS);
    if (trainingFile==null) {
      err.println("Invalid " + TRAINING_CLUSTERS + ": please provide valid file.");
      System.exit(-1);
    }
   
    // parse training data into clusters
    Map<String, Clusters> clusterMembership = new HashMap<String, Clusters>();
    JSONParser parser = new JSONParser();
    try {
      JSONObject parseObj = (JSONObject) parser.parse(new FileReader(trainingFile));
      JSONObject topicObj = (JSONObject) parseObj.get("topics");
      Set<String> topics = topicObj.keySet();
      Iterator<String> topicIt = topics.iterator();
      while (topicIt.hasNext()) { // for each topic
        String topic = topicIt.next();
        clusterMembership.put(topic, new Clusters());
        JSONArray clusters = (JSONArray) ((JSONObject) topicObj.get(topic)).get("clusters");
        Iterator<JSONArray> clusterIt = clusters.iterator();
        while (clusterIt.hasNext()) { // for each cluster in the topic
          JSONArray cluster = (JSONArray) clusterIt.next();
          Cluster c = new Cluster();
          Iterator<String> clusterMemberIt = cluster.iterator();
          while (clusterMemberIt.hasNext()) { // for each docId in the cluster
            String member = clusterMemberIt.next();
            long memberId = Long.parseLong(member);
            c.add(memberId);
          }
          clusterMembership.get(topic).add(c);
        }
      }
    } catch (Exception e) {
      err.println("Error reading training data.");
      e.printStackTrace();
      System.exit(-1);
    }
   
    // instantiate search client
    TrecSearchThriftClient client = new TrecSearchThriftClient(params.getParamValue(HOST_OPTION),
        trainingPort, group, token);

    SimpleSearcher searcher = new SimpleSearcher(client, numResults);
   
    err.println("=== Train Queries ===");
   
    List<Double> thresholds = new ArrayList<Double>();
    double averageThreshold = 0;
    Iterator<GQuery> queryIterator = trainingQueries.iterator();
    while(queryIterator.hasNext()) {
      GQuery query = queryIterator.next();
     
      Map<Long, TResult> seenResults = searcher.search(query);
     
      SimpleJaccardClusterer clusterer = new SimpleJaccardClusterer(new ArrayList<TResult>(seenResults.values()));
     
      // sweep through jaccard steps, calculating F1
      double maxF1 = 0;
      double maxF1Threshold = 1;
      for (double j = 1.0; j >= 0.0; j -= stepSize) { // for each jaccard threshold step
        Clusters clusters = clusterer.cluster(j);
       
        // all clusters are created now, get a finalized set of results
        Set<Long> allResults = new HashSet<Long>(seenResults.keySet());
        allResults.removeAll(clusters.getAllClusteredResults()); // allResults includes unclustered plus one representative from each cluster
        for (Cluster c : clusters) {
          allResults.add(c.getFirstMember());
        }
       
        // calculate f1 on the finalized set
        Clusters seenClusters = new Clusters();
        Clusters trueClusters = clusterMembership.get(query.getTitle());
        Iterator<Long> resultIt = allResults.iterator();
        while (resultIt.hasNext()) {
          long result = resultIt.next();
          Cluster trueCluster = trueClusters.findCluster(result);
          if (trueCluster != null) { // if it is relevant, it will have a true cluster; if this is null, it's non-relevant
            seenClusters.add(trueCluster);
          }
        }
       
        int numRetrievedClusters = seenClusters.size();
        int numResultsReturned = allResults.size();
        int numTrueClusters = trueClusters.size();

        double precision = 0;
        double recall = 0;
        double f1 = 0;
        if (evalType.equals("unweighted")) {
          precision = numRetrievedClusters / (double) numResultsReturned;
          recall = numRetrievedClusters / (double) numTrueClusters;
          f1 = 2 * precision * recall / (precision + recall);
        } else {       
          // for weighted measurements, we need the weight of each cluster
          int retrievedWeight = 0;
          for (Cluster cluster : seenClusters) {
            int w = cluster.getWeight(query, qrels);
            retrievedWeight += w;
          }
          int resultsWeight = 0;
          for (long result : allResults) {
            int w = 0;
            if (seenClusters.findCluster(result) == null)
            resultsWeight += w;
          }
          int trueWeight = 0;
          for (Cluster cluster : trueClusters) {
            int w = cluster.getWeight(query, qrels);
            trueWeight += w;
          }
         
          precision = retrievedWeight / (double) resultsWeight; // <--- ??????
          recall = retrievedWeight / (double) trueWeight;
          f1 = 2 * precision * recall / (precision + recall);
        }
        if (f1 > maxF1) {
          maxF1 = f1;
          maxF1Threshold = j;
        }
      }
      thresholds.add(maxF1Threshold);
      err.println("F1: "+df.format(maxF1)+"; Jaccard: "+df.format(maxF1Threshold));
     
    }
   
    // get the average threshold
    for (double threshold : thresholds) {
      averageThreshold += threshold;
    }
    averageThreshold /= thresholds.size();
    err.println("Average Jaccard: "+averageThreshold);
   
    err.println("=== Test Queries ===");
   
    // now cluster the test queries and output
    queryIterator = queries.iterator();
    while(queryIterator.hasNext()) {
      GQuery query = queryIterator.next();
      err.println(query.getTitle());
     
      client = new TrecSearchThriftClient(params.getParamValue(HOST_OPTION), testingPort, group, token);
      searcher = new SimpleSearcher(client, numResults);
      Map<Long, TResult> seenResults = searcher.search(query);
     
      SimpleJaccardClusterer clusterer = new SimpleJaccardClusterer(new ArrayList<TResult>(seenResults.values()));
      Clusters clusters = clusterer.cluster(averageThreshold);
     
      // all clusters are created now, get a finalized set of results
      Set<Long> allResults = new HashSet<Long>(seenResults.keySet());
      allResults.removeAll(clusters.getAllClusteredResults()); // allResults includes unclustered plus one representative from each cluster
      for (Cluster c : clusters) {
        allResults.add(c.getFirstMember());
      }
     
      int i = 0;
      for (long result : allResults) {
        TResult hit = seenResults.get(result);
        out.println(String.format("%s Q0 %s %d %f %s", query.getTitle(), hit.getId(), i, hit.getRsv(), runTag));
        i++;
      }
    }
    out.close();
    err.close();
  }

}
TOP

Related Classes of edu.gslis.ttg.main.RunTTGBaseline

TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.