Use ArticleItem object during clustering as well as display; add sort…

… by date function and use that to order each cluster.
Daniel-Mietchen · Nov 29, 2018 · 8d0e563cbd2cde49edf2a6dff9de27a8a919348f · 8d0e563
1 parent 71fb84e
commit 8d0e563cbd2cde49edf2a6dff9de27a8a919348f
Unified Split

Showing with 49 additions and 40 deletions.

+14 −17 index.php

+9 −0 lib/article_model.php

+25 −22 lib/clustering.php

+1 −1 resources/html/index_bs4.html
diff --git a/index.php b/index.php
@@ -68,7 +68,6 @@ function getORCIDurl ( $s ) {
 foreach ( $items_authors AS $q ) $to_load[] = $q ;
 $wil->loadItems ( $to_load ) ;
 
-
 $delete_statements = array() ;
 if ( $action == 'add' ) {
 	print "<form method='post' class='form' action='https://tools.wmflabs.org/quickstatements/api.php'>" ;
@@ -111,26 +110,28 @@ function getORCIDurl ( $s ) {
 <input type='hidden' name='fuzzy' value='$fuzzy' />
 <input type='hidden' name='name' value='" . escape_attribute($name) . "' />" ;
 
-$clusters = cluster_articles ( $wil, $items_papers ) ;
-#print "<pre>" ; print_r ( $clusters ) ; print "</pre>" ;
-
-// P50 authors
 $to_load = array() ;
+$article_items = array();
 foreach ( $items_papers AS $q ) {
 	$i = $wil->getItem ( $q ) ;
 	if ( !isset($i) ) continue ;
-	$claims = $i->getClaims ( 'P50' ) ;
-	foreach ( $claims AS $c ) $to_load[] = $i->getTarget ( $c ) ;
-	$claims = $i->getClaims ( 'P1433' ) ;
-	foreach ( $claims AS $c ) $to_load[] = $i->getTarget ( $c ) ;
+
+	$article = new WikidataArticleEntry( $i ) ;
+	$article_items[] = $article ;
+
+	foreach ( $article->authors AS $auth ) $to_load[] = $auth ;
+	foreach ( $article->published_in AS $pub ) $to_load[] = $pub ;
+	foreach ( $article->topics AS $topic ) $to_load[] = $topic ;
 }
 $wil->loadItems ( $to_load ) ;
 
+$clusters = cluster_articles ( $article_items ) ;
+
+#print "<pre>" ; print_r ( $clusters ) ; print "</pre>" ;
 // Publications
 $name_counter = array() ;
 print "<h2>Potential publications</h2>" ;
-print "<p>" . count($items_papers) . " publications found</p>" ;
-
+print "<p>" . count($article_items) . " publications found</p>" ;
 
 $is_first_group = true ;
 foreach ( $clusters AS $cluster_name => $cluster ) {
@@ -148,11 +149,8 @@ function getORCIDurl ( $s ) {
 	print "<th>Author Name Strings</th><th>Identified Authors</th>" ;
 	print "<th>Published In</th><th>Identifier(s)</th>" ;
 	print "<th>Topic</th><th>Published Date</th></tr>" ;
-	foreach ( $cluster AS $q ) {
-		$q = "Q$q" ;
-		$i = $wil->getItem ( $q ) ;
-		if ( !isset($i) ) continue ;
-		$article = new WikidataArticleEntry( $i ) ;
+	foreach ( $cluster AS $article ) {
+		$q = $article->q ;
 
 		$out = array() ;
 		foreach ( $article->author_names AS $a ) {
@@ -198,7 +196,6 @@ function getORCIDurl ( $s ) {
 		if ( count($article->topics) > 0 ) {
 			$topics = [] ;
 			foreach ( $article->topics AS $qt ) {
-				$wil->loadItem ( $qt ) ;
 				$i2 = $wil->getItem($qt) ;
 				if ( !isset($i2) ) continue ;
 				$topics[] = "<a href='https://www.wikidata.org/wiki/" . $i2->getQ() . "' target='_blank' style='color:green'>" . $i2->getLabel() . "</a>" ;

diff --git a/lib/article_model.php b/lib/article_model.php
@@ -75,5 +75,14 @@ public function formattedPublicationDate () {
 		if ( $this->publication_date != '' ) $formatted_date = DateTime::createFromFormat( '\+Y-m-d\TH:i:s\Z', $this->publication_date )->format( "Y-m-d" );
 		return $formatted_date ;
 	}
+
+	public static function dateCompare ($a, $b) {
+		$adate = $a->publication_date ;
+		$bdate = $b->publication_date ;
+		if ($adate == $bdate) {
+			return 0;
+		}
+		return ($adate > $bdate) ? -1 : 1 ;
+	}
 }
 ?>
diff --git a/lib/clustering.php b/lib/clustering.php
@@ -3,21 +3,17 @@
 // Heuristic clustering algorithm
 $min_authors_for_cluster = 4 ;
 $score_cache = array() ;
-function compareArticles( $wil, $q1 , $q2 ) {
+function compareArticles( $article1 , $article2 ) {
 	global $score_cache , $min_authors_for_cluster ;
 
-	if ( $q1 > $q2 ) return compareArticles ( $wil, $q2 , $q1 ) ; // Enforce $q1 <= $q2
+	$q1 = $article1->q ;
+	$q2 = $article2->q ;
+	if ( $q1 > $q2 ) return compareArticles ( $article2 , $article1 ) ; // Enforce $q1 <= $q2
 	$key = "$q1|$q2" ;
 	if ( isset($score_cache[$key]) ) return $score_cache[$key] ;
 
-	$i1 = $wil->getItem ( $q1 ) ;
-	$i2 = $wil->getItem ( $q2 ) ;
-	if ( !isset($i1) or !isset($i2) ) return 0 ;
-	$authors1 = $i1->getStrings ( 'P2093' ) ;
-	$authors2 = $i2->getStrings ( 'P2093' ) ;
-	
-	foreach ( $i1->getClaims('P50') AS $claim ) $authors1[] = $i1->getTarget ( $claim ) ;
-	foreach ( $i2->getClaims('P50') AS $claim ) $authors2[] = $i2->getTarget ( $claim ) ;
+	$authors1 = array_merge( $article1->author_names, $article1->authors ) ;
+	$authors2 = array_merge( $article2->author_names, $article2->authors ) ;
 	
 	$score = 0 ;
 	if ( count($authors1) < $min_authors_for_cluster or count($authors2) < $min_authors_for_cluster ) {
@@ -39,36 +35,43 @@ function compareArticles( $wil, $q1 , $q2 ) {
 	return $score ;
 }
 
-function cluster_articles ( $wil, $items_papers ) {
+function cluster_articles ( $article_items ) {
 	$clusters = array() ;
 	$min_score = 30 ;
 	$is_in_cluster = array() ;
-	foreach ( $items_papers AS $q1 ) {
+	foreach ( $article_items AS $article ) {
+		$q1 = $article->q ;
 		if ( isset($is_in_cluster[$q1]) ) continue ;
-		$base_score = compareArticles ( $wil, $q1 , $q1 ) ;
+		$base_score = compareArticles ( $article , $article ) ;
 		if ( $base_score == 0 ) continue ;
 		$cluster = array() ;
-		foreach ( $items_papers AS $q2 ) {
+		foreach ( $article_items AS $article2 ) {
+			$q2 = $article2->q ;
 			if ( $q1 == $q2 ) continue ;
 			if ( isset($is_in_cluster[$q2]) ) continue ;
-			$score = compareArticles ( $wil, $q1 , $q2 ) ;
+			$score = compareArticles ( $article , $article2 ) ;
 			$score = 100 * $score / $base_score ;
 			if ( $score >= $min_score ) {
-				if ( count($cluster) == 0 ) $cluster[] = $q1 ;
-				$cluster[] = $q2 ;
+				if ( count($cluster) == 0 ) $cluster[] = $article ;
+				$cluster[] = $article2 ;
 			}
 		}
 	
 		if ( count($cluster) == 0 ) continue ;
-		foreach ( $cluster AS $q ) $is_in_cluster[$q] = $q ;
+		usort( $cluster, 'WikidataArticleEntry::dateCompare' );
+		foreach ( $cluster AS $c ) $is_in_cluster[$c->q] = 1 ;
 		$clusters['Group #'.(count($clusters)+1)] = $cluster ;
 	}
 	$cluster = array() ;
-	foreach ( $items_papers AS $q1 ) {
-		if ( isset($is_in_cluster[$q1]) ) continue ;
-		$cluster[] = $q1 ;
+	foreach ( $article_items AS $article ) {
+		if ( isset($is_in_cluster[$article->q]) ) continue ;
+		$cluster[] = $article ;
 	}
-	if ( count($cluster) > 0 ) $clusters['Misc'] = $cluster ;
+	if ( count($cluster) > 0 ) {
+		usort( $cluster, 'WikidataArticleEntry::dateCompare' );
+		$clusters['Misc'] = $cluster ;
+	}
+
 	return $clusters;
 }
 

diff --git a/resources/html/index_bs4.html b/resources/html/index_bs4.html
@@ -52,4 +52,4 @@
 </div>
 </div>
 </body>
-</html>
+</html>