This would be better titled "a comprehensive guide to aspect-based sentiment analysis using our service." It only abstractly defines how to do it, so I would definitely not call it a "comprehensive guide" (though I'd definitely call it an advertisement. HN has had a lot of these lately.)
Issues with NPS aside, I've had success doing aspect-based sentiment analysis using a summation of TF-IDF and NPS, paired with multinomial naive bayes classification, which is fairly simple and probably doesn't require a 3rd party service. You'll have to find a way to roll up TF-IDF and NPS though, as NPS is a fractional measure - so any given term within a document will have an NPS of -100, 0, or 100. Great for basic "positive", "neutral", or "negative" classification, but not so good for a true NPS score.
Issues with NPS aside, I've had success doing aspect-based sentiment analysis using a summation of TF-IDF and NPS, paired with multinomial naive bayes classification, which is fairly simple and probably doesn't require a 3rd party service. You'll have to find a way to roll up TF-IDF and NPS though, as NPS is a fractional measure - so any given term within a document will have an NPS of -100, 0, or 100. Great for basic "positive", "neutral", or "negative" classification, but not so good for a true NPS score.