I have a large scientific literature dataset (>20000), where references have been inserted in a very rough format (single text string, with authors, year, title, etc.).
We need to convert them into a clean format (BibTeX or similar). Do you know of any automatic system to help us with this work? There are several systems extracting info from the title or from a pdf (e.g. Mendeley) but I haven't found anything working on reference strings
If the references are formatted in some "parseable" way, and each reference can be coreced to a single string, it's just matter of 1) extractimg the various parts (i.e. authors, year, title, journal, pages...) and then 2) rewriting them into a "nice" format (BibTex is almost perfect).
;)
Grazie! ... but you make it too easy Damiano..
We tried, also with the help of IT experts, but the results are quite scarce.
Mendeley allows a matching with the title of the article, and I presume it works by matching the text you type with its database. But it only works if you type it in the system, not providing a long list of titles.
I'm making it easy on two (maybe wrong) assumptions: 1) you have a text file (or something that can be converted into a text file); 2) in that text file a row is a reference; 3) that row has a "constant and consistent" formatting.
If the three assumptions above are met, then is just matter of some rows (R or python, Sir?). That is, is a standard "crosswalking" problem, and usually those peoblems can be solved writing ad hoc scripts in some high-level language.
BTW: once you have all in BibTex, Jabref is a nice tool to handle references.
Yes, you are right and I should have been more precise. The references are all in text files. And the format is not perfectly constant. Not too difficult to arrive to separate authors and year. But after that we have problems in separating title, and journal/book/editors. I wonder if there is any tool that can search for title?
Hi,
If in your bibliographic dataset the paper title have some mark (writte quoted or in italic; or the journal title in italic), text2bib ( http://text2bib.economics.utoronto.ca/index.php/index ) maybe could converted it to BidText styles... if not... it is difficult that this software works ok....
If you bibliographic dataset is comprised mainly by recent (indexed) titles, you could use wizfolio ( http://wizfolio.com ), which extract titles comparing your dataset with Pubmed, CrossRef or other major bibliographic databases...
In other cases.... you need good luck!
Kindest regards,
Antonio.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Graphs