One of the most common questions I get from students in the university is about how to do a literature survey. Based on my experience, these are the advice I usually give to those students. If you have never done a literature survey before, I hope this advice can be useful for you too:
- Find papers from your community, not randomly;
- Triage papers before reading them;
- Decide a question you want to learn and practice active reading;
First Hint: Find papers from your community, not randomly.
Usually CS students begin their survey by entering keywords on google scholar and reading papers from there. I think that is a bad idea for a first survey.
Instead, I would recommend that you make a list of papers from the following sources (ask your advisor for suggestions about these, specially the second one): - Papers from people in your laboratory or research group that are related to your field of research. - Papers from conferences and journals that the people in your research group publish to. After that, you should look at the list of references in those papers, and add them to your survey list as well. Finally, read the references of those papers, and add them to your survey, recursively.
This will give you a large list of papers, that are both close to your research topic, and also relevant to your research and those of your colleagues.
Before you start reading those papers, pay attention to the patterns of authors, institutions, and venues (conferences and journals). Are there authors that appear over and over? Common conferences and journals? One important thing that you can learn from literature surveys is what is the community of your research field looks like (you can also go to the web page of those authors to see if there are more papers for you to add to your survey).
Second Hint: Triage papers before reading them;
Ok, now you have a large pile of papers, and you should start reading them, right? No!
This is another common mistake I see in students who are doing a literature survey for the first time. They try to read each paper in their list entirely, and probably give up after the third or fourth, learning very little. You should do a triage first.
To do a triage, first define what you want to learn from your literature review. You are not reading the papers just for reading them -- scientific papers are not Novels! You want to know something: What are the most common used techniques. How many papers use a specific benchmark. What is the state of the art for a specific problem, etc. Your goal is to learn as much as possible about that question from as many papers as possible in your literature review.
So get your huge pile of papers, read the titles for all of them. From the piles, put those papers that look like they can answer your question in a yes pile, and the other papers on a no pile. When in doubt, put the paper on the yes pile.
Next, you get those papers from the yes pile, and read their abstracts. If the abstract looks promising, keep the paper. If not, put it on the no pile.
Next, go through the yes pile again, and read the conclusions of the papers. Put those that don't look promising on the no pile.
By now your yes pile should be much smaller, and you can try to read the papers in that pile. Keep in mind that you don't want to read the entire paper in detail: You want to try and go straight to the part that answer your question, and extract that information.
Third Hint: Decide a question you want to learn and practice active reading;
Now that we are actually reading the papers, we need to make sure that we are getting as much information about them as possible. Get yourself some paper, and take notes as you read. If you don't understand something, try to decide if that will help you to answer your question. If not, take a note to read it again later and skip it for now.
Take notes about how each paper answer your question, maybe making a table about it or something. Taking notes as you read (active reading) is extremely important for your brain to retain information. You might want to consider using reference management software (such as Zotero) to add tags to the papers that you read, or just print them and add plenty of colorful sticky notes. Printing the papers has the advantage that you get a growing, physical reminder of your literature survey, which can actually be motivating for some people (I remember fondly the big folder of papers from the first serious survey I did during my master program).
Summary:
There is, of course, a lot more that could be said about the topic of surveying papers, but I think these three hints can give you a very good start:
- Find papers from your community, not randomly;
- Triage papers before reading them;
- Decide a question you want to learn and practice active reading;