Hi Christian,

I second Iman's answer, it is probably easier to prepare a table like that and gephi is also more likely to import it correctly than a full matrix (although I think this was my first recommendation to you some time ago...). However, I think there are several steps to take between the current arrangement of your file and Iman's version which acquire some excel-knowledge.

First you need to devide the second column by the semicolons so you have each reference in a seperate cell. The next problem is that you have the papers in the first column in a different format than the references (full author names vs. APA reference). I don't know about a way to align both styles automatically, so you probably need to do this by hand. And if you do so, it is probably easier to work with ids as Iman suggested .

Christian, let me know if I you want to meet/skype about that,

best,

Julia

Disclaimer: I'm neither a Gephi specialist, nor a network analysis specialist.

I guess Iman and Julia are right. See:

- https://github.com/gephi/gephi/wiki/Import-CSV-Data

- https://gephi.org/users/supported-graph-formats/spreadsheet/

- https://gephi.org/users/supported-graph-formats/

-> You need to import a) a node table and b) an edge table (see Iman's description for the edge table). if your nodes just have ID and no labels, you don't need to import the node table (Gephi will generate it from the edge talbe, if you select this option; see links above)

What is the best way to generate a node and an edge table?

1) The best option would be if Citavi would offer an export filter in a graph format (format supportet by Gephi: https://gephi.org/users/supported-graph-formats/). To my knowledge Citavi has not an export filter for a graph format (I may be wrong). Maybe they are interested in programming an export filter for a graph format. It might be an interesting feature for Citavi to offer an export for a graph format. You can make your suggestions here: http://www.citavi.com/filter

Or you could ask in the Citavi forum (http://support.citavi.com/forum/) whether Citavi offers an easy way to generate a file (or two files: node table, edge table) like that. In my experience the Citavi support is very active and reponsive in the forum.

2) If option one is not possible: You could write a script/programm code (e.g. Visual Basic Script, Python, maybe code for a Citavi Addon etc.) to convert your Citavi excel file output to the required node/edge table formats. (But you should not use the authors as the Identifiaction for the 'root publication', see below)

3) I guess you could convert it manually (supported by Excel commands and functions) in Excel. This may be faster than option 2, but (in my opinion) it's more prone to errors, (if it's not scripted) it can take a lot of time if you need to convert data again and again (e.g. your bibliography changes etc.), and  it's less reproducible (if it's not scripted). I guess excel offers the required commands/functions to conduct Julia's steps. But (as Julia suggested) you should identifiy your "root publication" in the excel export (first column) not with the author's name, but with the same field Citavi uses for the references (I guess it's the field 'short title'?) (if you need the authors field you can use a second column for them). To conduct Julia's first step (devide the second column), you could use Excel's "Text to Columns" command (see: https://support.office.com/en-us/article/Split-names-by-using-the-Convert-Text-to-Columns-Wizard-2cd989db-2b1f-4d89-b17b-534250ff9905). Several further steps are needed to convert it to the needed node and edge tables from there. I guess Excel offers the required tools (e.g., lookup function, vlookup function, etc. to generate a affiliation matrix; and to convert the affiliation matrix to Iman's format I guess you could use a pivot table; to convert a matrix to a 3 column table see: http://stackoverflow.com/questions/20541905/convert-matrix-to-3-column-table-reverse-pivot-unpivot-flatten-normal).

Please let me know (PM), if you need help with option 3. But I would advice to try option 1 first.

Best,

Joel

Thanks Iman, Julia & Joel!! Your very detailed answers helped me a lot. And thanks for your offer to contact you again with any further questions. First of all, I am going to follow your suggestions and try to feed Gephi in a proper way. Let's keep in touch. Best regards, Christian

hi

Forget about Gephi now. First, try to read your file in any programming langauge to expand your file. Then give each title an ID and build a mathing table between these IDs. Try python it is very easy. Before start save your file in txt format. After you finsh this step save your file in CSV format with source and target as the titels of each column. Open gephi and you can import the file.

If you need any help with Python code to do this let me know

Hey Mohammad, thanks for your feedback. Actually, I am preparing the nodesfile with all IDs at the moment. I am going to have a look at python. It seems to be a pretty easy way to prepare a file for gephi. Thanks for your offer to contact you again!

Best regards, Christian

Hey there,

this is more than interesting to me. Has anyone of you been successful creating a co-citation graph using any of the mentioned techniques? If so, how time-intensive is the process? Any chance of automating parts of it?

Best regards,

Jannis

Hi Jannis,

I went for option 3 (manual preparation using excel functions). I had around 2.000 nodes and 150.000 edges. It took me around 2 days. I think with this amount of data it's doable. If you need concrete guidance please let me know. You can also call me if you want. However, meanwhile I would also highly recommend to spend the time to get familiar with python.

Best

Christian

btw, you only have to prepare a two-mode matrix and can automatically convert this file to a one-mode network (co-citations) by means of UCINET which is freely available.

Dear Violeta,

If you want to transform your two-mode data in excel to a co-occurrence matrix, I recommend to use UCINET (freeware). It has a function to transform your data pretty easily. Afterward you can import this file to GEPHI. However, please have a look again to Joel's post who provided a detailed answer and has much more (IT-) expertise than I do.

Best regards,

Christian

Hello

I work with 2 database Scopus or Web of Science

They give more of the clusters analyzes

Best regards

Ingrid García

I have a related query. If I download a database from Scopus as a .csv file, how can I create nodes and edges tables (or separate .csv files for nodes and edges) from that file, for doing different types of network analysis in Gephi?

I have tried BibExcel which generates a .net file that can be directly used in Gephi for co-citation analysis. But what about doing other kind of analysis, e.g., keywords-based?

Please help.