<?php
namespace headstart;
require_once "autoload.inc.php";
use headstart\persistence;
use headstart\library;
$INI_DIR = dirname(__FILE__) . "/../../preprocessing/conf/";
$ini_array = library\Toolkit::loadIni($INI_DIR);
$persistence = new persistence\SQLitePersistence($ini_array["connection"]["sqlite_db"]);
//$persistence->createTables();
//$persistence->createVisualization("test_id", "My Test Vis", "{[some json string;]}");
echo $persistence->getLastVersion("vis_id2");
//$persistence->writeRevision("test_id", "{[some other json string;]}");
//echo $persistence->getLastVersion("test_id");
public function performNaming($working_dir)
{
$ini = $this->ini_array["naming"];
$ini_general = $this->ini_array["general"];
$ini_output = $this->ini_array["output"];
$ini_connection = $this->ini_array["connection"];
$WORKING_DIR = $working_dir;
//Output of scaling and clustering script
$CLUSTERS = $WORKING_DIR . $ini_output["output_scaling_clustering"];
//Output file
$OUTPUT_FILE = $WORKING_DIR . $ini_output["output_naming"];
//Output file for the full API responses
$FULL_ZEMANTA = $WORKING_DIR . "full_responses/zemanta/";
$FULL_CALAIS = $WORKING_DIR . "full_responses/calais/";
$cluster = array();
$cluster_details = array("title" => array(), "abstracts" => array());
$counts = array();
$stop_words = array();
$output = array();
$cluster_text_file = library\Toolkit::openFileForReading($CLUSTERS);
$stop_words_file = library\Toolkit::openFileForReading($ini_general["preprocessing_dir"] . $ini["stop_words"]);
while (($line = fgetcsv($stop_words_file, null, "\t")) !== false) {
$this->stop_words[] = $line[0];
}
$row = 0;
while (($line = fgetcsv($cluster_text_file, null)) !== false) {
if ($row == 0) {
$output[] = $line;
$row++;
continue;
}
$output[] = $line;
$line_cluster_id = intval($ini["line_cluster_id"]);
$line_title = intval($ini["line_title"]);
$line_abstract = intval($ini["line_abstract"]);
if (!isset($line[$line_cluster_id])) {
throw new \Exception("Error in line: " . $line[0]);
}
if (!isset($cluster[$line[$line_cluster_id]])) {
$cluster[$line[$line_cluster_id]] = $line[$line_title] . ". " . $line[$line_abstract];
$cluster_details[$line[$line_cluster_id]]["title"] = $line[$line_title] . ".";
$cluster_details[$line[$line_cluster_id]]["abstracts"] = $line[$line_abstract];
$counts[$line[$line_cluster_id]] = 1;
} else {
$cluster[$line[$line_cluster_id]] .= "\n" . $line[$line_title] . ". " . $line[$line_abstract];
$cluster_details[$line[$line_cluster_id]]["title"] .= "\n" . $line[$line_title] . ".";
$cluster_details[$line[$line_cluster_id]]["abstracts"] .= "\n" . $line[$line_abstract];
$counts[$line[$line_cluster_id]]++;
}
}
$topics = $this->executeCurl($cluster);
$cluster_names = array();
foreach ($cluster as $id => $text) {
library\Toolkit::info($text);
$categories = array("topics" => array(), "topics_title" => array());
$categories_one = array("topics" => array());
//get 1-grams
$response_object_one = $this->getNgrams($cluster[$id], 1);
$this->processNgrams($response_object_one, "topics_title", $categories_one, $ini["threshold_single_words"]);
//get 2-, 3-, and 4-grams
for ($n = 4; $n >= 2; $n--) {
$response_object = $this->getNgrams($cluster[$id], $n);
$response_object_title = $this->getNgrams($cluster_details[$id]["title"], $n);
arsort($response_object);
$this->processNgrams($response_object_title, "topics_title", $categories, $ini["threshold_title_ngrams"]);
$this->processNgrams($response_object, "topics", $categories, $ini["threshold_title_abstract_ngrams"]);
}
library\Toolkit::info($id . ": " . print_r($categories, true));
$cluster_names_calais = $topics["calais"][$id];
$cluster_names_zemanta = $topics["zemanta"][$id];
$cluster_name = "";
//Search for 4-, 3-, and 2-title-grams in Calais concepts
$cluster_name = $this->compareConcepts($cluster_names_calais, $categories, "topics_title");
//If that fails, search for 4-, 3-, and 2-title-grams in Zemanta concepts
if ($cluster_name == "") {
$cluster_name = $this->compareConcepts($cluster_names_zemanta, $categories, "topics_title");
}
//If that fails, search for 4-, 3-, and 2-grams in Zemanta concepts
if ($cluster_name == "") {
$cluster_name = $this->compareConcepts($cluster_names_zemanta, $categories, "topics");
}
//If that fails, search for 1-grams in Zemanta concepts
if ($cluster_name == "") {
$count_new = 0;
$filtered_array = array_filter($categories_one["topics_title"], function ($item) {
return !in_array($item, $this->ini_array["naming"]["forbidden_names"]);
});
library\Toolkit::info("Filtered Array: " . print_r($filtered_array, true));
foreach ($cluster_names_zemanta["topics_format"] as $name) {
$key = array_search($name, $filtered_array);
if ($key !== false) {
$cluster_name = $cluster_names_zemanta["topics"][$count_new];
break;
}
$count_new++;
}
}
//If everything above fails, name the cluster after the most important concept
//returned by (1) Zemanta or (2) Calais. Finally, name the cluster
//"Miscellaneous"
if ($cluster_name == "") {
if (isset($cluster_names_zemanta["topics"][0])) {
$cluster_name = $cluster_names_zemanta["topics"][0];
} elseif (isset($cluster_names_calais["topics"][0])) {
$cluster_name = $cluster_names_calais["topics"][0];
} else {
$cluster_name = "Miscellaneous";
}
}
$cluster_id = library\Toolkit::generateUriFromString($cluster_name);
$cluster_temp = $cluster_id;
$count = 1;
foreach ($cluster_names as $attributes) {
if ($attributes["uri"] == $cluster_id) {
$cluster_id = $cluster_temp . "-" . $count;
$count++;
}
}
$cluster_names[$id] = array("name" => $cluster_name, "uri" => $cluster_id);
library\Toolkit::info("*** CLUSTER NAME: " . $cluster_name . "\n");
//Write full response for later consultation
$this->getFullResponseZemanta($text, $cluster_id, $FULL_ZEMANTA);
$this->getFullResponseCalais($text, $cluster_id, $FULL_CALAIS);
}
//add areas to output array
array_push($output[0], "area_uri", "area");
library\Toolkit::info(sizeof($output) . "\n");
$size = sizeof($output);
for ($counter = 1; $counter < $size; $counter++) {
$cluster_id = $output[$counter][$line_cluster_id];
array_push($output[$counter], $cluster_names[$cluster_id]["uri"], $cluster_names[$cluster_id]["name"]);
library\Toolkit::info("{$counter}\n");
}
$output_handle = library\Toolkit::openOrCreateFile($OUTPUT_FILE);
foreach ($output as $line) {
fputcsv($output_handle, $line);
}
fclose($output_handle);
$UNIQUE_ID = $ini_output["unique_id"];
$TITLE = $ini_output["title"];
$persistence = new persistence\SQLitePersistence($ini_connection["sqlite_db"]);
$header = array_shift($output);
$json_array = array();
foreach ($output as $row) {
$json_array[] = array_combine($header, $row);
}
$json = json_encode($json_array);
$persistence->createVisualization($UNIQUE_ID, $TITLE, $json);
}
