private static void DeleteImageNodes(HtmlNode articleNode)
{
var imageNotes = articleNode.GetElementsByTagName("img");
foreach (var image in imageNotes)
{
image.Remove();
}
}
private static double GetLinkDensity(HtmlNode node)
{
var links = node.GetElementsByTagName("a");
var textLength = GetInnerText(node).Length;
var linkLength = links.Sum(l => GetInnerText(l).Length);
return linkLength * 1.0 / textLength;
}
private static void RemoveScripts(HtmlNode node)
{
foreach (var script in node.GetElementsByTagName("script"))
{
script.Remove();
}
}
private static string GetArticleTitle(HtmlNode htmlNode)
{
var titleNode = htmlNode.GetElementsByTagName("title").FirstOrDefault();
if (titleNode == null) return null;
string currTitle, origTitle;
currTitle = origTitle = GetInnerText(titleNode);
if (Regex.IsMatch(currTitle, @" [\|\-] "))
{
currTitle = Regex.Replace(origTitle, @"(.*)[\|\-] .*", "$1");
if (currTitle.Split(' ').Length < 3)
{
currTitle = origTitle.Replace(@"[^\|\-]*[\|\-](.*)", "$1");
}
}
else if (currTitle.IndexOf(": ") != -1)
{
currTitle = Regex.Replace(origTitle, @".*:(.*)", "$1");
if(currTitle.Split(' ').Length < 3)
{
currTitle = Regex.Replace(origTitle, @"[^:]*[:](.*)", "$1");
}
}
else if (currTitle.Length > 150 || currTitle.Length < 15)
{
var hOnes = htmlNode.GetElementsByTagName("h1");
if (hOnes.Count == 1)
{
currTitle = GetInnerText(hOnes[0]);
}
}
if (currTitle.Split(' ').Length <= 4)
{
currTitle = origTitle;
}
return currTitle.Trim();
}