For most modern Web publishing, this is mostly a matter of finding and extracting the <article> block, as well as metadata (title, byline, dateline).
html-xml-tools is quite useful for this.
I'd created a WaPo extractor that reduced pagesize by about 95%, stripped the nags and paywalls, etc. Endpoint was HTML, but that could just as easily have generated PDF or ePub if I'd wanted.