I'm in the "I read TA with a skeptical eye and it's never steered me wrong" boat. That said, I think there are some simple things they could do that would make the reviews more useful/helpful beyond some common-sense spam/fake review supervision:

1) Add a filter that would allow you to only show reviews by people with X or more total reviews.

2) Add something like Amazon has that brings up the most helpful/representative positive and most helpful/representative negative review side-by-side.

3) Do some text mining and highlight common themes. For example "Positive reviews note: excellent for families, prompt service, quiet. Negative reviews note: broken elevator, dirty, far from city center."

4) Again like Amazon, work to "verify" reviews. I'm currently on a 3 month trip in India and I've booked several hotels though Cleartrip or Agoda by clicking links on Tripadvisor.com and I think my subsequent reviews of these properties should be highlighted in some way since it can be verified that I in fact booked a room there.

Generally though, the Facebook integration will start to help as certain reviewers can be "automatically trusted" through the social graph. And I have to think that some kind of press item or expose will force their hand on the fake reviews.

Incidentally, the fake reviews "problem", if they were ever to get serious about it, would be a fascinating machine learning problem to put on a crowdsourcing site like Kaggle.com. I would love to get my hands on some of that data, and I know others would as well.