From 98076556c7f1f33096e35ca33ee218c27baf26f1 Mon Sep 17 00:00:00 2001
From: Eike Cochu <eike@cochu.com>
Date: Sat, 2 Jan 2016 15:01:43 +0100
Subject: [PATCH] updated
---
TODO | 3 +
.../main/java/de/vipra/cmd/file/Filebase.java | 2 -
.../java/de/vipra/cmd/file/FilebaseIndex.java | 14 +++-
.../de/vipra/cmd/file/FilebaseVocabulary.java | 10 ++-
.../java/de/vipra/cmd/file/JGibbFilebase.java | 41 ++++++++----
.../java/de/vipra/cmd/file/LdacFilebase.java | 4 +-
.../de/vipra/cmd/lda/JGibbLDAAnalyzer.java | 58 ++++++++++++++++-
.../java/de/vipra/cmd/lda/LDAAnalyzer.java | 9 ++-
.../de/vipra/cmd/lda/LdacLDAAnalyzer.java | 23 -------
.../de/vipra/cmd/option/ImportCommand.java | 36 +++++-----
.../de/vipra/cmd/option/StatsCommand.java | 2 +-
.../main/java/de/vipra/util/Constants.java | 4 +-
.../de/vipra/util/model/ArticleStats.java | 3 +
.../main/java/de/vipra/util/model/Topic.java | 65 +++++++++++++++++++
.../java/de/vipra/util/model/TopicWord.java | 31 +++++++++
15 files changed, 236 insertions(+), 69 deletions(-)
delete mode 100644 vipra-cmd/src/main/java/de/vipra/cmd/lda/LdacLDAAnalyzer.java
create mode 100644 vipra-util/src/main/java/de/vipra/util/model/Topic.java
create mode 100644 vipra-util/src/main/java/de/vipra/util/model/TopicWord.java
diff --git a/TODO b/TODO
index 1da64172..286b0293 100644
--- a/TODO
+++ b/TODO
@@ -2,6 +2,9 @@ cmd
☐ implement delete operation
☐ implement filebase remove
☐ implement elasticsearch indexing
+ ☐ allow other document input formats
+ ☐ do not read whole file into memory
+ ☐ on save topics: retain topic names?
rest
☐ implement etag caching
\ No newline at end of file
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/file/Filebase.java b/vipra-cmd/src/main/java/de/vipra/cmd/file/Filebase.java
index 02dcda7b..2ea001f7 100644
--- a/vipra-cmd/src/main/java/de/vipra/cmd/file/Filebase.java
+++ b/vipra-cmd/src/main/java/de/vipra/cmd/file/Filebase.java
@@ -66,8 +66,6 @@ public abstract class Filebase implements Closeable {
public static Filebase getFilebase(Config config) throws FilebaseException, ConfigException {
File dataDir = config.getDataDirectory();
switch (Constants.Analyzer.fromString(config.getString(Key.ANALYZER))) {
- case LDAC:
- return new LdacFilebase(dataDir);
case JGIBB:
case DEFAULT:
default:
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/file/FilebaseIndex.java b/vipra-cmd/src/main/java/de/vipra/cmd/file/FilebaseIndex.java
index 516a5caa..dae34993 100644
--- a/vipra-cmd/src/main/java/de/vipra/cmd/file/FilebaseIndex.java
+++ b/vipra-cmd/src/main/java/de/vipra/cmd/file/FilebaseIndex.java
@@ -4,12 +4,13 @@ import java.io.Closeable;
import java.io.File;
import java.io.IOException;
import java.util.ArrayList;
+import java.util.Iterator;
import java.util.List;
import de.vipra.util.Constants;
import de.vipra.util.FileUtils;
-public class FilebaseIndex implements Closeable {
+public class FilebaseIndex implements Closeable, Iterable<String> {
private final File file;
private final List<String> index;
@@ -23,7 +24,7 @@ public class FilebaseIndex implements Closeable {
}
}
- private void write() throws IOException {
+ public void write() throws IOException {
FileUtils.writeLines(file, Constants.FB_ENCODING.name(), index, null, false);
}
@@ -40,6 +41,10 @@ public class FilebaseIndex implements Closeable {
return index.indexOf(id);
}
+ public String get(int i) {
+ return index.get(i);
+ }
+
public boolean remove(String id) {
return index.remove(id);
}
@@ -49,4 +54,9 @@ public class FilebaseIndex implements Closeable {
write();
}
+ @Override
+ public Iterator<String> iterator() {
+ return index.iterator();
+ }
+
}
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/file/FilebaseVocabulary.java b/vipra-cmd/src/main/java/de/vipra/cmd/file/FilebaseVocabulary.java
index 227fa33c..a910d941 100644
--- a/vipra-cmd/src/main/java/de/vipra/cmd/file/FilebaseVocabulary.java
+++ b/vipra-cmd/src/main/java/de/vipra/cmd/file/FilebaseVocabulary.java
@@ -4,12 +4,13 @@ import java.io.Closeable;
import java.io.File;
import java.io.IOException;
import java.util.ArrayList;
+import java.util.Iterator;
import java.util.List;
import de.vipra.util.Constants;
import de.vipra.util.FileUtils;
-public class FilebaseVocabulary implements Closeable {
+public class FilebaseVocabulary implements Closeable, Iterable<String> {
private File file;
private List<String> vocables;
@@ -23,7 +24,7 @@ public class FilebaseVocabulary implements Closeable {
}
}
- private void write() throws IOException {
+ public void write() throws IOException {
FileUtils.writeLines(file, Constants.FB_ENCODING.name(), vocables, null, false);
}
@@ -49,4 +50,9 @@ public class FilebaseVocabulary implements Closeable {
return vocables.indexOf(word);
}
+ @Override
+ public Iterator<String> iterator() {
+ return vocables.iterator();
+ }
+
}
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/file/JGibbFilebase.java b/vipra-cmd/src/main/java/de/vipra/cmd/file/JGibbFilebase.java
index b92fcf9a..1195e099 100644
--- a/vipra-cmd/src/main/java/de/vipra/cmd/file/JGibbFilebase.java
+++ b/vipra-cmd/src/main/java/de/vipra/cmd/file/JGibbFilebase.java
@@ -17,6 +17,8 @@ public class JGibbFilebase extends Filebase {
private final FilebaseVocabulary vocab;
private final List<Article> articles;
+ private final int bufferMaxSize = 100;
+
public JGibbFilebase(File dataDir) throws FilebaseException {
super(dataDir, "jgibb");
this.modelFile = getModelFile();
@@ -26,11 +28,19 @@ public class JGibbFilebase extends Filebase {
}
@Override
- public void add(Article article) {
+ public void add(Article article) throws FilebaseException {
String[] words = article.getProcessedText().getText().split("\\s+");
vocab.addVocabulary(words);
index.add(article.getId());
articles.add(article);
+
+ if (articles.size() >= bufferMaxSize) {
+ try {
+ write();
+ } catch (IOException e) {
+ throw new FilebaseException(e);
+ }
+ }
}
@Override
@@ -40,20 +50,23 @@ public class JGibbFilebase extends Filebase {
@Override
public void write() throws IOException {
- boolean linesep = modelFile.exists();
- RandomAccessFile raf = new RandomAccessFile(modelFile, "rw");
-
- // write articles
- raf.seek(raf.length());
- for (Article a : articles) {
- if (linesep)
- raf.writeBytes(System.lineSeparator());
- else
- linesep = true;
- raf.writeBytes(a.getProcessedText().getText());
- }
+ if (!articles.isEmpty()) {
+ boolean linesep = modelFile.exists();
+ RandomAccessFile raf = new RandomAccessFile(modelFile, "rw");
+
+ // write articles
+ raf.seek(raf.length());
+ for (Article a : articles) {
+ if (linesep)
+ raf.writeBytes(System.lineSeparator());
+ else
+ linesep = true;
+ raf.writeBytes(a.getProcessedText().getText());
+ }
- raf.close();
+ raf.close();
+ articles.clear();
+ }
}
}
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/file/LdacFilebase.java b/vipra-cmd/src/main/java/de/vipra/cmd/file/LdacFilebase.java
index 67b210a9..2360d4af 100644
--- a/vipra-cmd/src/main/java/de/vipra/cmd/file/LdacFilebase.java
+++ b/vipra-cmd/src/main/java/de/vipra/cmd/file/LdacFilebase.java
@@ -15,13 +15,13 @@ public class LdacFilebase extends Filebase {
}
@Override
- public void add(Article article) {
+ public void add(Article article) throws FilebaseException {
// TODO Auto-generated method stub
}
@Override
- public void remove(String id) {
+ public void remove(String id) throws FilebaseException {
throw new NotImplementedException();
}
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/lda/JGibbLDAAnalyzer.java b/vipra-cmd/src/main/java/de/vipra/cmd/lda/JGibbLDAAnalyzer.java
index 41e21180..ffebc4d5 100644
--- a/vipra-cmd/src/main/java/de/vipra/cmd/lda/JGibbLDAAnalyzer.java
+++ b/vipra-cmd/src/main/java/de/vipra/cmd/lda/JGibbLDAAnalyzer.java
@@ -1,13 +1,25 @@
package de.vipra.cmd.lda;
+import java.io.BufferedReader;
import java.io.File;
+import java.io.FileInputStream;
+import java.io.IOException;
+import java.io.InputStreamReader;
+import java.util.ArrayList;
+import java.util.List;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import de.vipra.cmd.ex.LDAAnalyzerException;
+import de.vipra.cmd.model.Article;
import de.vipra.util.Config;
+import de.vipra.util.FileUtils;
import de.vipra.util.ex.ConfigException;
+import de.vipra.util.ex.DatabaseException;
+import de.vipra.util.model.Topic;
+import de.vipra.util.model.TopicWord;
+import de.vipra.util.service.DatabaseService;
import jgibblda.Estimator;
import jgibblda.Inferencer;
import jgibblda.LDACmdOption;
@@ -38,7 +50,7 @@ public class JGibbLDAAnalyzer extends LDAAnalyzer {
modelDir = new File(dataDir, "jgibb");
options.dir = modelDir.getAbsolutePath();
- options.estc = new File(modelDir, "model-final.tassign").exists();
+ options.estc = new File(modelDir, "jgibb.tassign").exists();
options.est = !options.estc;
modelFile = new File(modelDir, "jgibb");
@@ -53,7 +65,6 @@ public class JGibbLDAAnalyzer extends LDAAnalyzer {
throw new LDAAnalyzerException("model file does not exist: " + modelFile.getAbsolutePath());
}
estimate();
- // inference();
}
private void estimate() {
@@ -62,10 +73,53 @@ public class JGibbLDAAnalyzer extends LDAAnalyzer {
estimator.estimate();
}
+ @SuppressWarnings("unused")
private void inference() {
Inferencer inferencer = new Inferencer();
inferencer.init(options);
Model newModel = inferencer.inference();
}
+ private List<Topic> readTopics() throws IOException {
+ File twords = new File(modelDir, "jgibb.twords");
+ List<String> lines = FileUtils.readFile(twords);
+ List<Topic> topics = new ArrayList<>();
+ List<TopicWord> topicWords = null;
+ for (String line : lines) {
+ if (line.startsWith("\t")) {
+ String[] parts = line.trim().split("\\s+");
+ topicWords.add(new TopicWord(parts[0], Double.parseDouble(parts[1])));
+ } else {
+ if (topicWords != null)
+ topics.add(new Topic(topicWords));
+ topicWords = new ArrayList<>();
+ }
+ }
+ return topics;
+ }
+
+ @Override
+ public void save(DatabaseService<Article> dbArticles, DatabaseService<Topic> dbTopics) throws LDAAnalyzerException {
+ try {
+ List<Topic> topics = readTopics();
+
+ // recreate topics in database
+ dbTopics.drop();
+ for (Topic topic : topics) {
+ dbTopics.createSingle(topic);
+ }
+
+ // read document topics
+ BufferedReader reader = new BufferedReader(
+ new InputStreamReader(new FileInputStream(new File(modelDir, "jgibb.tassign"))));
+ String line;
+ while ((line = reader.readLine()) != null) {
+ String[] parts = line.trim().split("\\s+");
+
+ }
+ } catch (IOException | DatabaseException e) {
+ throw new LDAAnalyzerException(e);
+ }
+ }
+
}
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/lda/LDAAnalyzer.java b/vipra-cmd/src/main/java/de/vipra/cmd/lda/LDAAnalyzer.java
index b7285a72..c352eb89 100644
--- a/vipra-cmd/src/main/java/de/vipra/cmd/lda/LDAAnalyzer.java
+++ b/vipra-cmd/src/main/java/de/vipra/cmd/lda/LDAAnalyzer.java
@@ -1,8 +1,11 @@
package de.vipra.cmd.lda;
import de.vipra.cmd.ex.LDAAnalyzerException;
+import de.vipra.cmd.model.Article;
import de.vipra.util.Config;
import de.vipra.util.Constants;
+import de.vipra.util.model.Topic;
+import de.vipra.util.service.DatabaseService;
import de.vipra.util.Config.Key;
public abstract class LDAAnalyzer {
@@ -21,12 +24,12 @@ public abstract class LDAAnalyzer {
public abstract void analyze() throws LDAAnalyzerException;
+ public abstract void save(DatabaseService<Article> dbArticles, DatabaseService<Topic> dbTopics)
+ throws LDAAnalyzerException;
+
public static LDAAnalyzer getAnalyzer(Config config) throws LDAAnalyzerException {
LDAAnalyzer analyzer = null;
switch (Constants.Analyzer.fromString(config.getString(Key.ANALYZER))) {
- case LDAC:
- analyzer = new LdacLDAAnalyzer();
- break;
case JGIBB:
case DEFAULT:
default:
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/lda/LdacLDAAnalyzer.java b/vipra-cmd/src/main/java/de/vipra/cmd/lda/LdacLDAAnalyzer.java
deleted file mode 100644
index 0b65ea36..00000000
--- a/vipra-cmd/src/main/java/de/vipra/cmd/lda/LdacLDAAnalyzer.java
+++ /dev/null
@@ -1,23 +0,0 @@
-package de.vipra.cmd.lda;
-
-import de.vipra.cmd.ex.LDAAnalyzerException;
-import de.vipra.util.Config;
-
-public class LdacLDAAnalyzer extends LDAAnalyzer {
-
- protected LdacLDAAnalyzer() {
- super("lda-c Analyzer");
- }
-
- @Override
- public void init(Config config) throws LDAAnalyzerException {
- // TODO Auto-generated method stub
-
- }
-
- @Override
- public void analyze() throws LDAAnalyzerException {
- // TODO Auto-generated method stub
- }
-
-}
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/option/ImportCommand.java b/vipra-cmd/src/main/java/de/vipra/cmd/option/ImportCommand.java
index a61a8da3..5b6c8226 100644
--- a/vipra-cmd/src/main/java/de/vipra/cmd/option/ImportCommand.java
+++ b/vipra-cmd/src/main/java/de/vipra/cmd/option/ImportCommand.java
@@ -6,7 +6,6 @@ import java.io.FileReader;
import java.io.FilenameFilter;
import java.io.IOException;
import java.util.ArrayList;
-import java.util.List;
import org.json.simple.JSONArray;
import org.json.simple.JSONObject;
@@ -27,6 +26,7 @@ import de.vipra.util.Constants;
import de.vipra.util.StringUtils;
import de.vipra.util.Timer;
import de.vipra.util.model.ArticleStats;
+import de.vipra.util.model.Topic;
import de.vipra.util.service.DatabaseService;
public class ImportCommand implements Command {
@@ -38,6 +38,7 @@ public class ImportCommand implements Command {
private JSONParser parser = new JSONParser();
private Config config;
private DatabaseService<Article> dbArticles;
+ private DatabaseService<Topic> dbTopics;
private Filebase filebase;
private Processor preprocessor;
private LDAAnalyzer analyzer;
@@ -82,7 +83,7 @@ public class ImportCommand implements Command {
* @return
* @throws ImportException
*/
- Article importArticle(JSONObject obj) throws ImportException {
+ void importArticle(JSONObject obj) throws ImportException {
out.debug("importing \"" + StringUtils.ellipsize(obj.get("title").toString(), 80) + "\"");
Article article = new Article();
article.fromJSON(obj);
@@ -99,8 +100,6 @@ public class ImportCommand implements Command {
// add article to filebase
filebase.add(article);
-
- return article;
} catch (Exception e) {
throw new ImportException(e, article.getId());
}
@@ -116,21 +115,22 @@ public class ImportCommand implements Command {
* @throws ImportException
* @throws Exception
*/
- private List<Article> importFile(File file)
- throws FileNotFoundException, IOException, ParseException, ImportException {
+ private long importFile(File file) throws FileNotFoundException, IOException, ParseException, ImportException {
Object data = parser.parse(new FileReader(file));
- List<Article> articles = new ArrayList<Article>();
+ long imported = 0;
if (data instanceof JSONArray) {
for (Object object : (JSONArray) data) {
- articles.add(importArticle((JSONObject) object));
+ importArticle((JSONObject) object);
+ imported++;
}
} else if (data instanceof JSONObject) {
- articles.add(importArticle((JSONObject) data));
+ importArticle((JSONObject) data);
+ imported++;
}
- return articles;
+ return imported;
}
@Override
@@ -138,6 +138,7 @@ public class ImportCommand implements Command {
try {
config = Config.getConfig();
dbArticles = DatabaseService.getDatabaseService(config, Constants.Collection.ARTICLES, Article.class);
+ dbTopics = DatabaseService.getDatabaseService(config, Constants.Collection.TOPICS, Topic.class);
filebase = Filebase.getFilebase(config);
preprocessor = Processor.getPreprocessor(config);
analyzer = LDAAnalyzer.getAnalyzer(config);
@@ -151,25 +152,28 @@ public class ImportCommand implements Command {
// import files into database and filebase
out.info("file import");
- List<Article> articles = new ArrayList<>();
+ long imported = 0;
for (File file : files) {
- articles.addAll(importFile(file));
+ imported += importFile(file);
}
long durImport = timer.lap();
// write filebase
out.info("writing file index");
filebase.close();
- long durIndex = timer.lap();
+ timer.lap();
// do topic modeling
out.info("topic modeling");
analyzer.analyze();
long durAnalyze = timer.lap();
- out.info("imported " + articles.size() + " " + (articles.size() == 1 ? "article" : "articles"));
- out.info("import: " + StringUtils.timeString(durImport) + ", analyze: " + StringUtils.timeString(durAnalyze)
- + ", reindex: " + StringUtils.timeString(durIndex));
+ out.info("saving topic models");
+ analyzer.save(dbArticles, dbTopics);
+
+ out.info("imported " + imported + " " + (imported == 1 ? "article" : "articles"));
+ out.info("import: " + StringUtils.timeString(durImport) + ", analyze: "
+ + StringUtils.timeString(durAnalyze));
} catch (Exception e) {
throw new ExecutionException(e);
}
diff --git a/vipra-cmd/src/main/java/de/vipra/cmd/option/StatsCommand.java b/vipra-cmd/src/main/java/de/vipra/cmd/option/StatsCommand.java
index 6cf6800b..53d7a738 100644
--- a/vipra-cmd/src/main/java/de/vipra/cmd/option/StatsCommand.java
+++ b/vipra-cmd/src/main/java/de/vipra/cmd/option/StatsCommand.java
@@ -7,7 +7,7 @@ public class StatsCommand implements Command {
@Override
public void run() throws ExecutionException {
// TODO Auto-generated method stub
-
+
}
}
diff --git a/vipra-util/src/main/java/de/vipra/util/Constants.java b/vipra-util/src/main/java/de/vipra/util/Constants.java
index 34e8043a..0594c27d 100644
--- a/vipra-util/src/main/java/de/vipra/util/Constants.java
+++ b/vipra-util/src/main/java/de/vipra/util/Constants.java
@@ -30,7 +30,8 @@ public class Constants {
"then", "there", "these", "they", "this", "to", "was", "will", "with");
public static enum Collection {
- ARTICLES("articles");
+ ARTICLES("articles"),
+ TOPICS("topics");
public final String name;
@@ -67,7 +68,6 @@ public class Constants {
}
public static enum Analyzer {
- LDAC("ldac"),
JGIBB("jgibb"),
DEFAULT(JGIBB);
diff --git a/vipra-util/src/main/java/de/vipra/util/model/ArticleStats.java b/vipra-util/src/main/java/de/vipra/util/model/ArticleStats.java
index 2b4cc36c..fca89724 100644
--- a/vipra-util/src/main/java/de/vipra/util/model/ArticleStats.java
+++ b/vipra-util/src/main/java/de/vipra/util/model/ArticleStats.java
@@ -47,6 +47,7 @@ public class ArticleStats implements BsonDocument {
stats.setWordCount(words.length);
Map<String, TermFrequency> uniqueWords = new HashMap<>();
long maxFrequency = 0;
+
// loop and count unique words
// also remember maximum frequency
for (String word : words) {
@@ -60,10 +61,12 @@ public class ArticleStats implements BsonDocument {
}
uniqueWords.put(word, tf);
}
+
// normalize frequencies
for (Map.Entry<String, TermFrequency> entry : uniqueWords.entrySet()) {
entry.getValue().normalizeTermFrequency(maxFrequency);
}
+
stats.setUniqueWordCount(uniqueWords.size());
stats.setUniqueWords(uniqueWords);
return stats;
diff --git a/vipra-util/src/main/java/de/vipra/util/model/Topic.java b/vipra-util/src/main/java/de/vipra/util/model/Topic.java
new file mode 100644
index 00000000..168a4c94
--- /dev/null
+++ b/vipra-util/src/main/java/de/vipra/util/model/Topic.java
@@ -0,0 +1,65 @@
+package de.vipra.util.model;
+
+import java.io.File;
+import java.io.IOException;
+import java.util.List;
+
+import org.bson.Document;
+
+public class Topic extends Model {
+
+ private List<String> names;
+ private List<TopicWord> words;
+
+ public Topic() {}
+
+ public Topic(List<TopicWord> words) {
+ this.words = words;
+ }
+
+ public List<String> getNames() {
+ return names;
+ }
+
+ public void setNames(List<String> names) {
+ this.names = names;
+ }
+
+ public List<TopicWord> getWords() {
+ return words;
+ }
+
+ public void setWords(List<TopicWord> words) {
+ this.words = words;
+ }
+
+ @Override
+ public String getType() {
+ return Topic.class.getSimpleName().toLowerCase();
+ }
+
+ @Override
+ public void fromDocument(Document document) {
+ // TODO Auto-generated method stub
+
+ }
+
+ @Override
+ public Document toDocument() {
+ // TODO Auto-generated method stub
+ return null;
+ }
+
+ @Override
+ public void fromFile(File file) throws IOException {
+ // TODO Auto-generated method stub
+
+ }
+
+ @Override
+ public String toFileString() {
+ // TODO Auto-generated method stub
+ return null;
+ }
+
+}
diff --git a/vipra-util/src/main/java/de/vipra/util/model/TopicWord.java b/vipra-util/src/main/java/de/vipra/util/model/TopicWord.java
new file mode 100644
index 00000000..22fe8ab0
--- /dev/null
+++ b/vipra-util/src/main/java/de/vipra/util/model/TopicWord.java
@@ -0,0 +1,31 @@
+package de.vipra.util.model;
+
+public class TopicWord {
+
+ private String word;
+ private double likeliness;
+
+ public TopicWord() {}
+
+ public TopicWord(String word, double likeliness) {
+ this.word = word;
+ this.likeliness = likeliness;
+ }
+
+ public String getWord() {
+ return word;
+ }
+
+ public void setWord(String word) {
+ this.word = word;
+ }
+
+ public double getLikeliness() {
+ return likeliness;
+ }
+
+ public void setLikeliness(double likeliness) {
+ this.likeliness = likeliness;
+ }
+
+}
--
GitLab