added word-position to ranking (this is only a first step)

git-svn-id: https://svn.berlios.de/svnroot/repos/yacy/trunk@1395 6c8d7289-2bf4-0310-a012-ef5d649a1542

source/de/anomic/plasma/plasmaCondenser.java

@@ -114,40 +114,6 @@ public final class plasmaCondenser {
         return words.entrySet().iterator();
     }
     
-    /*
-    public int wordCount(String word) {
-        // number of occurrences of one word
-        // if the word did not occur, this simply returns 0
-        wordStatProp sp = (wordStatProp) words.get(word);
-        if (sp == null) return 0;
-        return sp.count;
-    }
-
-    public int wordPositionInText(String word) {
-        // position of word in text
-        // if unknown and word does not exist, the method returns 0
-        wordStatProp sp = (wordStatProp) words.get(word);
-        if (sp == null) return 0;
-        return sp.posInText;
-    }
-
-    public int wordPositionInPhrase(String word) {
-        // position of word in text
-        // if unknown and word does not exist, the method returns 0
-        wordStatProp sp = (wordStatProp) words.get(word);
-        if (sp == null) return 0;
-        return sp.posInPhrase;
-    }
-
-    public int wordNumberOfPhrase(String word) {
-        // position of word in text
-        // if unknown and word does not exist, the method returns 0
-        wordStatProp sp = (wordStatProp) words.get(word);
-        if (sp == null) return 0;
-        return sp.numOfPhrase;
-    }
-    */
-    
     public static class wordStatProp {
         // object carries statistics for words and sentences
         

source/de/anomic/plasma/plasmaSearchPreOrder.java

@@ -137,7 +137,9 @@ public final class plasmaSearchPreOrder {
             else if (query.order[i].equals(plasmaSearchQuery.ORDER_YBR))  ranking  = factor * ybr_p(indexEntry.getUrlHash());
             factor = factor / 4096L;
         }
-
+        int wordpos = indexEntry.posintext();
+        if (wordpos == 0) wordpos = 1000;
+        ranking = ranking + 1000 - wordpos + indexEntry.hitcount();
         pageAcc.put(serverCodings.encodeHex(ranking, 16) + indexEntry.getUrlHash(), indexEntry);
     }
 

source/de/anomic/plasma/plasmaSearchResult.java

@@ -157,6 +157,9 @@ public final class plasmaSearchResult {
                 else if (query.order[j].equals(plasmaSearchQuery.ORDER_YBR))  ranking += factor * plasmaSearchPreOrder.ybr_p(indexEntry.getUrlHash());
                 factor = factor / 4096L;
             }
+            int wordpos = indexEntry.posintext();
+            if (wordpos == 0) wordpos = 1000;
+            ranking = ranking + 1000 - wordpos + indexEntry.hitcount();
             
             // apply 'common-sense' heuristic using references
             for (int j = 0; j < urlcomps.length; j++) if (commonSense.contains(urlcomps[j])) ranking += 10L*4096L*4096L / urlcomps.length;

source/de/anomic/plasma/plasmaWordIndexEntry.java

@@ -73,13 +73,13 @@ public final class plasmaWordIndexEntry {
     private final String urlHash;
 
     // discrete values
-    private int    hitcount;       // words in file
-    private int    wordcount;
-    private int    phrasecount;
+    private int    hitcount;    // number of this words in file
+    private int    wordcount;   // number of all words in the file
+    private int    phrasecount; // number of all phrases in the file
     private int    posintext;   // first position of the word in text as number of word; 0=unknown or irrelevant position
     private int    posinphrase; // position within a phrase of the word
     private int    posofphrase; // position of the phrase in the text as count of sentences; 0=unknown; 1=path; 2=keywords; 3=headline; >4: in text
-    private int    worddistance;
+    private int    worddistance;// distance between the words, only used if the index is artificial (from a conjunction)
     private long   lastModified;// calculated by using last-modified
     private int    quality;     // result of a heuristic on the source file
     private byte[] language;    // essentially the country code (the TLD as heuristic), two letters lowercase only
@@ -101,12 +101,12 @@ public final class plasmaWordIndexEntry {
 
     // appearance locations: (used for flags)
     public static final int AP_TITLE  =  0; // title tag from html header
-    public static final int AP_H1     = 1; // h0-tag
-    public static final int AP_H2     = 2;
-    public static final int AP_H3     = 3;
-    public static final int AP_H4     = 4;
-    public static final int AP_H5     = 5;
-    public static final int AP_H6     = 6;
+    public static final int AP_H1     =  1; // h1-tag
+    public static final int AP_H2     =  2; // h2-tag
+    public static final int AP_H3     =  3; // h3-tag
+    public static final int AP_H4     =  4; // h4-tag
+    public static final int AP_H5     =  5; // h5-tag
+    public static final int AP_H6     =  6; // h6-tag
     public static final int AP_ANCHOR =  7; // anchor description
     public static final int AP_URL    =  8; // word inside an url
     public static final int AP_IMG    =  9; // tag inside image references
@@ -254,6 +254,9 @@ public final class plasmaWordIndexEntry {
         this.worddistance = (code.length() >= 19) ? (int) kelondroBase64Order.enhancedCoder.decodeLong(code.substring(18, 20)) : 0;
         this.wordcount = (code.length() >= 21) ? (int) kelondroBase64Order.enhancedCoder.decodeLong(code.substring(20, 22)) : 0;
         this.phrasecount = (code.length() >= 23) ? (int) kelondroBase64Order.enhancedCoder.decodeLong(code.substring(22, 24)) : 0;
+        if (hitcount == 0) hitcount = 1;
+        if (wordcount == 0) wordcount = 1000;
+        if (phrasecount == 0) phrasecount = 100;
     }
     
     public plasmaWordIndexEntry(String external) {
@@ -335,7 +338,7 @@ public final class plasmaWordIndexEntry {
     public int getQuality() { return quality; }
     public int getVirtualAge() { return plasmaWordIndex.microDateDays(lastModified); }
     public long getLastModified() { return lastModified; }
-    public int getCount() { return hitcount; }
+    public int hitcount() { return hitcount; }
     public int posintext() { return posintext; }
     public int posinphrase() { return posinphrase; }
     public int posofphrase() { return posofphrase; }