How do you decide on a research topic with little research experience?

[Nocturnae]

I am trying to narrow down my research interests for my Master's research, and while I have a good idea about the subfield I would like to work on (natural language understanding and/or machine translation), I find it extremely challenging to identify a specific topic. As a recent computer science graduate with relatively little research experience (especially in terms of breadth), it seems impossible for me to come up with a decent idea. I don't know how anyone else does it, to be honest.

I'm now trying to go through the major and recent papers in the said fields to develop a better sense. It definitely helps to get up-to-date with the field, but most certainly not enough to produce an interesting idea of my own.

How did/would you go about doing this? Any ideas or suggestions?

Thanks in advance!

[Meraki]

I'm relatively new to it all myself, but reading the literature is a great start. Jot down questions you have as you're reading, such as limitations to the study, gaps in the theory that are not addressed, assumptions being made that may or may not be true, etc. This will help you pinpoint areas of opportunity. I also recommend getting out of the office and just taking a walk after doing some reading. Some of my best ideas have come from casual observations on the street, thinking about my prior work experience, or other situations outside of reading the literature. I can then integrate my own observations, thoughts, and current research to see if the idea's been done before, or how I might extend upon the idea with my own research. I also recommend keeping a journal where you jot down your research ideas any time they pop into your head. 

If one or both of the areas you mentioned have a large literature and you don't know where to start, you can look for recent literature reviews to quickly fill in some gaps and identify specific areas of interest, then dig deeper within those areas. You don't need to find a massive gap with an idea that's never been done before - even looking at a currently discussed problem in a different way using different theory or methods, or finding boundary conditions to current theories, etc. can make a big contribution.