Co-occurrence network options: words and variables

[Andrealee1993]

Version number of KH Coder

3.02 macOS

Your Qeuestion

Your Operating System
macOS Catalina 10.15.7

What language of text are you trying to analyze?
English

Result:

My question is:
I am quite curious about the accuracy of this graph. I am doing a co-occurrence network through variables: time, with different sizes of corpora(2016, 20 million, 2020, 40 million; 2024 55 million). If the minimum term frequency is set at 200, a smaller corpus might contain fewer than 200 words. Right? Thus, the graph would generate fewer words in 2016. This may cause problems in interpreting the data. How to solve problems like this?

[ko-ichi-h]

Indeed, the co-occurrence network only uses words that appear more than 200 times in the whole data. Therefore, if there are few words that appear more than 200 times in the 2016 corpus, there will be few words that are connected to "2016".

As a possible solution, what would happen if you changed 200 to 175 or 150? Alternatively, it might be a good idea to create a table of characteristic words. The automatically created table shows the top 10 words, but you could manually make it into a table of around the top 20 words.

[Andrealee1993]

Thanks, professor! Got that!

[Andrealee1993]

Many many tanks again! I tried! And it worked! Finally, find some patterns!!!!!

[Andrealee1993]

Dear Professor Koichi,
Regarding this question, may I know the underlying logic of generating characteristic words? And whether is it possible to generate more than ten words like twenty words?
Looking forward to your reply.
Andrea

[Andrealee1993]

Dear Professor Koichi,
Regarding this question, may I know the underlying logic of generating characteristic words? And whether is it possible to generate more than ten words like twenty words?
Looking forward to your reply.
Andrea

[ko-ichi-h]

may I know the underlying logic of generating characteristic words?

Section A.5.7 of the manual may help.

And whether is it possible to generate more than ten words like twenty words?

You need to manually make the table.

On this screen, follow steps (1) to (3).

Then (4) click "2016" in the "Values ​​& Value labels" field, and (5) click [*Words] -> [selected value].

This will display more than 20 words that are characteristic of 2016. Copy the results and create your own table. This will allow you to make it 20 or 30 instead of 10.