private $dbh;
private $score_modifier = 50;
private $sql_prefix = "ttrss_plugin_af_sort_bayes";
+ private $auto_categorize_threshold = 10000;
function about() {
return array(1.0,
function init($host) {
require_once __DIR__ . "/lib/class.naivebayesian.php";
+ //require_once __DIR__ . "/lib/class.naivebayesian_ngram.php";
require_once __DIR__ . "/lib/class.naivebayesianstorage.php";
$this->host = $host;
$article_id = (int) $_REQUEST["article_id"];
$train_up = sql_bool_to_bool($_REQUEST["train_up"]);
- $category = $train_up ? "GOOD" : "NEUTRAL";
+ //$category = $train_up ? "GOOD" : "UGLY";
+ $dst_category = "UGLY";
$nbs = new NaiveBayesianStorage($_SESSION["uid"]);
$nb = new NaiveBayesian($nbs);
$this->dbh->query("BEGIN");
- if ($nb->untrain($guid, $content)) {
- if ($score >= $this->score_modifier) $score -= $this->score_modifier;
+ $ref = $nbs->getReference($guid, false);
+
+ if (isset($ref['category_id'])) {
+ $current_category = $nbs->getCategoryById($ref['category_id']);
+ } else {
+ $current_category = "UGLY";
}
- $nb->train($guid, $nbs->getCategoryByName($category), $content);
+ // set score to fixed value for now
+
+ if ($train_up) {
+ switch ($current_category) {
+ case "UGLY":
+ $dst_category = "GOOD";
+ $score = $this->score_modifier;
+ break;
+ case "BAD":
+ $dst_category = "UGLY";
+ $score = 0;
+ break;
+ case "GOOD":
+ $dst_category = "GOOD";
+ break;
+ }
+ } else {
+ switch ($current_category) {
+ case "UGLY":
+ $dst_category = "BAD";
+ $score = -$this->score_modifier;
+ break;
+ case "BAD":
+ $dst_category = "BAD";
+ break;
+ case "GOOD":
+ $dst_category = "UGLY";
+ $score = 0;
+ break;
+ }
+ }
- if ($category == "GOOD") $score += $this->score_modifier;
+ $nb->untrain($guid, $content);
+ $nb->train($guid, $nbs->getCategoryByName($dst_category), $content);
$this->dbh->query("UPDATE ttrss_user_entries SET score = '$score' WHERE ref_id = $article_id AND owner_uid = " . $_SESSION["uid"]);
}
- print "$article_id :: $category";
+ print "$article_id :: $dst_category :: $score";
}
function get_js() {
function hook_article_button($line) {
return "<img src=\"plugins/af_sort_bayes/thumb_up.png\"
style=\"cursor : pointer\" style=\"cursor : pointer\"
- onclick=\"bayesTrain(".$line["id"].", true)\"
+ onclick=\"bayesTrain(".$line["id"].", true, event)\"
class='tagsPic' title='".__('+1')."'>" .
"<img src=\"plugins/af_sort_bayes/thumb_down.png\"
style=\"cursor : pointer\" style=\"cursor : pointer\"
- onclick=\"bayesTrain(".$line["id"].", false)\"
- class='tagsPic' title='".__('-1')."'>";
+ onclick=\"bayesTrain(".$line["id"].", false, event)\"
+ class='tagsPic' title='".__('-1')."'>" .
+ "<img src=\"plugins/af_sort_bayes/chart_bar.png\"
+ style=\"cursor : pointer\" style=\"cursor : pointer\"
+ onclick=\"bayesShow(".$line["id"].")\"
+ class='tagsPic' title='".__('Show classifier info')."'>";
}
category_id INTEGER NOT NULL,
FOREIGN KEY (category_id) REFERENCES ${prefix}_categories(id) ON DELETE CASCADE,
owner_uid INTEGER NOT NULL,
- FOREIGN KEY (owner_uid) REFERENCES ttrss_users(id) ON DELETE CASCADE,
- content text NOT NULL) ENGINE=InnoDB");
+ FOREIGN KEY (owner_uid) REFERENCES ttrss_users(id) ON DELETE CASCADE) ENGINE=InnoDB");
$this->dbh->query("CREATE TABLE IF NOT EXISTS ${prefix}_wordfreqs (
word varchar(100) NOT NULL DEFAULT '',
$this->dbh->query("CREATE TABLE IF NOT EXISTS ${prefix}_categories (
id SERIAL NOT NULL PRIMARY KEY,
category varchar(100) NOT NULL DEFAULT '',
- probability DOUBLE NOT NULL DEFAULT '0',
+ probability DOUBLE PRECISION NOT NULL DEFAULT '0',
owner_uid INTEGER NOT NULL REFERENCES ttrss_users(id) ON DELETE CASCADE,
word_count BIGINT NOT NULL DEFAULT '0')");
id SERIAL NOT NULL PRIMARY KEY,
document_id VARCHAR(255) NOT NULL,
category_id INTEGER NOT NULL REFERENCES ${prefix}_categories(id) ON DELETE CASCADE,
- owner_uid INTEGER NOT NULL REFERENCES ttrss_users(id) ON DELETE CASCADE,
- content text NOT NULL)");
+ owner_uid INTEGER NOT NULL REFERENCES ttrss_users(id) ON DELETE CASCADE)");
$this->dbh->query("CREATE TABLE IF NOT EXISTS ${prefix}_wordfreqs (
word varchar(100) NOT NULL DEFAULT '',
if ($this->dbh->num_rows($result) == 0) {
$this->dbh->query("INSERT INTO ${prefix}_categories (category, owner_uid) VALUES ('GOOD', $owner_uid)");
- $this->dbh->query("INSERT INTO ${prefix}_categories (category, owner_uid) VALUES ('NEUTRAL', $owner_uid)");
+ $this->dbh->query("INSERT INTO ${prefix}_categories (category, owner_uid) VALUES ('BAD', $owner_uid)");
+ $this->dbh->query("INSERT INTO ${prefix}_categories (category, owner_uid) VALUES ('UGLY', $owner_uid)");
}
}
$this->dbh->query("COMMIT");
}
- function hook_prefs_tab($args) {
- if ($args != "prefPrefs") return;
-
- print "<div dojoType=\"dijit.layout.AccordionPane\" title=\"".__('Bayesian classifier (af_sort_bayes)')."\">";
-
+ function renderPrefsUI() {
$result = $this->dbh->query("SELECT category, probability, word_count,
(SELECT COUNT(id) FROM {$this->sql_prefix}_references WHERE
category_id = {$this->sql_prefix}_categories.id) as doc_count
FROM {$this->sql_prefix}_categories WHERE owner_uid = " . $_SESSION["uid"]);
+ print "<h3>" . __("Statistics") . "</h3>";
+
+ print "<p>".T_sprintf("Required UGLY word count for automatic matching: %d", $this->auto_categorize_threshold)."</p>";
+
print "<table>";
- print "<tr><th>Category</th><th>Probability</th><th>Word count</th><th>Article count</th></tr>";
+ print "<tr><th>Category</th><th>Probability</th><th>Words</th><th>Articles</th></tr>";
while ($line = $this->dbh->fetch_assoc($result)) {
print "<tr>";
print "</table>";
+ print "<h3>" . __("Last matched articles") . "</h3>";
+
+ $result = $this->dbh->query("SELECT te.title, category, tf.title AS feed_title
+ FROM ttrss_entries AS te, ttrss_user_entries AS tu, ttrss_feeds AS tf, {$this->sql_prefix}_references AS tr, {$this->sql_prefix}_categories AS tc
+ WHERE tf.id = tu.feed_id AND tu.ref_id = te.id AND tc.id = tr.category_id AND tr.document_id = te.guid ORDER BY te.id DESC LIMIT 20");
+
+ print "<ul class=\"browseFeedList\" style=\"border-width : 1px\">";
+
+ while ($line = $this->dbh->fetch_assoc($result)) {
+ print "<li>" . $line["category"] . ": " . $line["title"] . " (" . $line["feed_title"] . ")</li>";
+ }
+
+ print "</ul>";
+
+ print "<button dojoType=\"dijit.form.Button\" onclick=\"return bayesUpdateUI()\">".
+ __('Refresh')."</button> ";
+
print "<button dojoType=\"dijit.form.Button\" onclick=\"return bayesClearDatabase()\">".
__('Clear database')."</button> ";
//
+ }
+
+ function hook_prefs_tab($args) {
+ if ($args != "prefPrefs") return;
+
+ print "<div id=\"af_sort_bayes_prefs\" dojoType=\"dijit.layout.AccordionPane\" title=\"".__('Bayesian classifier (af_sort_bayes)')."\">";
+
+ $this->renderPrefsUI();
print "</div>";
}
function hook_article_filter($article) {
$owner_uid = $article["owner_uid"];
+ // guid already includes owner_uid so we don't need to include it
+ $result = $this->dbh->query("SELECT id FROM {$this->sql_prefix}_references WHERE
+ document_id = '" . $this->dbh->escape_string($article['guid_hashed']) . "'");
+
+ if (db_num_rows($result) != 0) {
+ _debug("bayes: article already categorized");
+ return $article;
+ }
+
$nbs = new NaiveBayesianStorage($owner_uid);
$nb = new NaiveBayesian($nbs);
if (count($categories) > 0) {
$count_neutral = 0;
- $count_good = 0;
+
$id_good = 0;
- $id_neutral = 0;
+ $id_ugly = 0;
+ $id_bad = 0;
foreach ($categories as $id => $cat) {
if ($cat["category"] == "GOOD") {
$id_good = $id;
- $count_good += $cat["word_count"];
- } else if ($cat["category"] == "NEUTRAL") {
- $id_neutral = $id;
+ } else if ($cat["category"] == "UGLY") {
+ $id_ugly = $id;
$count_neutral += $cat["word_count"];
+ } else if ($cat["category"] == "BAD") {
+ $id_bad = $id;
}
}
- $dst_category = $id_neutral;
+ $dst_category = $id_ugly;
$bayes_content = mb_strtolower($article["title"] . " " . strip_tags($article["content"]));
- if ($count_neutral >= 3000 && $count_good >= 1000) {
+ if ($count_neutral >= $this->auto_categorize_threshold) {
// enable automatic categorization
$result = $nb->categorize($bayes_content);
- if (count($result) == 2) {
+ //print_r($result);
+
+ if (count($result) == 3) {
$prob_good = $result[$id_good];
- $prob_neutral = $result[$id_neutral];
+ $prob_bad = $result[$id_bad];
- if ($prob_good > 0.90 && $prob_good > $prob_neutral) {
- $dst_category = $id_good; // should we autofile as good or not? idk
+ if ($prob_good > 0.90) {
+ $dst_category = $id_good;
$article["score_modifier"] += $this->score_modifier;
+ } else if ($prob_bad > 0.90) {
+ $dst_category = $id_bad;
+ $article["score_modifier"] -= $this->score_modifier;
}
}
+
+ _debug("bayes, dst category: $dst_category");
}
$nb->train($article["guid_hashed"], $dst_category, $bayes_content);
$nb->updateProbabilities();
}
+ function showArticleStats() {
+ $article_id = (int) $_REQUEST["article_id"];
+
+ $result = $this->dbh->query("SELECT score, guid, title, content FROM ttrss_entries, ttrss_user_entries WHERE ref_id = id AND id = " .
+ $article_id . " AND owner_uid = " . $_SESSION["uid"]);
+
+ if ($this->dbh->num_rows($result) != 0) {
+ $guid = $this->dbh->fetch_result($result, 0, "guid");
+ $title = $this->dbh->fetch_result($result, 0, "title");
+ $content = mb_strtolower($title . " " . strip_tags($this->dbh->fetch_result($result, 0, "content")));
+
+ print "<h2>" . $title . "</h2>";
+
+ $nbs = new NaiveBayesianStorage($_SESSION["uid"]);
+ $nb = new NaiveBayesian($nbs);
+
+ $categories = $nbs->getCategories();
+
+ $ref = $nbs->getReference($guid, false);
+
+ $current_cat = isset($ref["category_id"]) ? $categories[$ref["category_id"]]["category"] : "N/A";
+
+ print "<p>" . T_sprintf("Currently stored as: %s", $current_cat) . "</p>";
+
+ $result = $nb->categorize($content);
+
+ print "<h3>" . __("Classifier result") . "</h3>";
+
+ print "<table>";
+ print "<tr><th>Category</th><th>Probability</th></tr>";
+
+ foreach ($result as $k => $v) {
+ print "<tr>";
+ print "<td>" . $categories[$k]["category"] . "</td>";
+ print "<td>" . $v . "</td>";
+
+ print "</tr>";
+ }
+
+ print "</table>";
+
+ } else {
+ print_error("Article not found");
+ }
+
+ print "<div align='center'>";
+
+ print "<button dojoType=\"dijit.form.Button\" onclick=\"return dijit.byId('bayesShowDlg').hide()\">".
+ __('Close this window')."</button>";
+
+ print "</div>";
+
+ }
+
function api_version() {
return 2;
}