You are absolutely right that asking for registration when the user gets no benefit from it is wrong, and will result in many less people using the features. Therefore you need to consider how the interaction goes and what benefit the user should get from this.
In this case, the user does need to get some benefit, and they do because their reviews can be traced back to them, and so their credibility or reputaion can be assessed by other readers - if I go somewhere that you recommended and enjoyed it, I will be interested in other places you recommend. This credibility, similar to SOs reputation, is something some people will pursue.
OTOH, there will be some who will not want to register, so there is a place for recording reviews without registering. It might be worth having these reviews assessed, which gives another benefit of registering, that ( say ) after the first review, they go up immediately.
Using OpenID or similar should be an option worth looking at, because the registration process needs to be as simple as possible - the benefits are not huge, and the benefit to the paper is far greater, so registering needs to be very simple and straightforward.
You need to balance the benefits achieved - or potential benefits like gifts for regular contributors, as @Dave Sherohman suggested - against the work required to perform the registration. The research is very clear that if the benefits are not clear, then registration will prove just a barrier, and registrations that you get will be poorer quality ( duff information provided, for example ).