I am preparing my PhD in computer science. I want to do a state of the art about the topic I am working on, are there any criteria to consider when choosing papers to review? Thank you in advance :)
There is a systematic way to perform a state of the art in any topic of interest usually referred to as SLR's and I'll be specific to the field of Computer Science here. You should consider reading: Kitchenham and Charters Guidelines on how to perform a systematic literature review and follow the same strategy.
Another strategy is to use the snowballing procedure. However, the two strategies suit better in conjunction to validate on being consistent with your retrieved results on that specific topic.
Or you can use any of them, explicitly.
A very rough criteria is to have a read of the 30/40 most cited papers on the very specific topic. Or find a review paper from the last 3/4 years and read the papers citing it.
In absence/difficulties in doing that (access to publications, topic still too broad, etcetc), try to look for video recordings of award prizes at conferences, usually the awarded person will try to give an overview of the state-of-art plus his/her personal take&contributions on given topic.
Here you can find an example of the ACM A.M. Turing Award: likely it is a bit too popular, try to find a more R&D oriented prize.