The easiest way I've found to scrape web sites is to do the scraping in a script (e.g. PHP) on a server, extract the data you're interested in and return it as XML in response to a query from the Roku.
For example, PHP's loadHTML function can parse most HTML (even if not well-formed), returning a DOMDocument object, which you can examine using DOMDocument methods to extract your data. You can create the XML to send back to your Roku using the XMLWriter class, for example.
I like this method because the Roku is very good at parsing XML data, and other scripting languages make it easier to parse the HTML. Also, if you find that the web site you are scraping changes the way its pages are laid out, you can make the corresponding change in your PHP script which takes effect immediately, rather than having to update your Roku channel which takes some time to propagate to your users.
