Datathief is another option and is free. You should copy the graph from pdf file and save as a jpg (for example with "Paint" from Windows). Later, open the image.jpg with Data Theft and extract the data. Is very useful to extract data for effect size calculation for meta-analysis:
In my case 99% of the times the link Paolo originally suggested is not able to recognize tables in the papers I work with.
There are many options available so I would recommend to test yourself which one is the best:
If there are a lot of tables, save the pdf as an Excel workbook in your acrobat reader. Then, copy-paste the tables you are interested in. If you are only interested in one table, select it, right-click and export as xls. Personally, it's the one I use after trying several. It's quick and easy although it has got problems, mainly with headers, so you usually have to revise it and iron out for minor changes. It works the same if you were interested in other formats.
Online converters: (quick and easy), (works well but you have to receive an email to download the document).
Offline (free software tools): Bytescout PDF Multitool , Tabula
In my experience, all these options are quicker than the usual copy-paste (even using "Text Import Wizard" or "Paste special"). I consider formatting the table in Word the most time-consuming option.
Hi, if i can be of some help here: for mac users, just use the program called Automator (it comes with your mac, in the folder named applications). It's a 2 step process, no configuration necessary, transforms pdf in text.
If you need to create a dataset, instead, extracting also tables, you will need to use ad hoc programs such as Rapidminer which can create datasets from collections of documents (it is free for students).