I back up what Paul said, but as I had to do something similar just a few weeks back, here's what worked for me: I found most relevant publications (articles) by the same author.
Typically, there's not that much publications related to a thesis, and there's a possibility that most of the papers are just extensions to the first one. Here's what I think from a Computer Science perspective.
- Read the first publication by the author related to the thesis topic
- Some of the following articles are probably application focus for the (novel) technique presented in the first article, or provide a heavy math background - these are not really needed to understand the idea
- There might be an article or two improving the construction algorithm (the concept stays the same, but some implementation improvements)
- In the end, you'll end up reading the first + one or two other articles and that will give you a good idea of what the whole thesis is about
- Now that you understand the concepts presented in the thesis, you understand the Table of Contents fully. You can easily identify chapters interesting to you, and read only the selected ones.
It is still a lengthy process, but I think faster than trying to read the whole thesis, and gives a lot result-wise: you not only understood the concepts you needed, but did literature research as well, and know exactly where to look for every type of extra details you might need.
This all said, this is the process that worked for me when I needed to understand the concept presented, the main idea (but not the details) of implementation, and wanted to be able to apply the concept "by hand" and "on paper" for small mock examples.
I think the process can be adapted for whatever goal you have in reading the thesis: you almost certainly are not interested in absolutely everything presented in the thesis on your first read-out. So, if you are, for example, interested in the application domain, you'll read the application focus articles instead, and not math profs.