Combine nearby words function with the use of predefined codes

[KHCoderuser3000]

Version number of KH Coder

3.Beta.07d

Your Question

Your Operating System
Windows 10

Your Question
Dear Professor Higuchi, Thank you very much in advance for helping me on this issue.
I would like to use KH Coder to identify the frequency of lists of words coded as risk, threat, national security, human security, or planetary security as well as their co-occurrence with words coded as climate change. So far so good. However, it would be splendid if I could also get values on nearby occurrences of risk+national security or threat+human security etc. Following the instructions in the manual, I tried to refer to my previous codes as followed:

*human security threat
near(<*threat>-<*human security>)
*human security risk
near(<*risk>-<*human security>)

However, even if I broaden the search by adding [d0] to look at whole paragraphs there are 0 occurrences of this code which is rather unlikely. This is the case even for documents in which the individual codes for threat or human security each occur over 500 times in the txt file.

Might it be the case that I cannot refer to a whole code when using the nearby words function? If so, how can I solve this issue? There are some 20-40 words associated to each of the codes mentioned above.

Screenshots
If applicable, add screenshots to help explain your question.

What language of text are you trying to analyze?
English

[ko-ichi-h]

Dear user,

Currently, you can only use simple words in "near(A-B)" specification. Sorry for the inconvenience it may cause.

Maybe, you can set the "unit" as "sentences" and use "<*A> and <*B>". This way, you can search for / count sentences which have both *A and *B.

Also please use underbars instead of white spaces in names of codes. "*human_security" would be nice.

[KHCoderuser3000]

Dear Professor Higuchi,

Thank you for your reply!

I will use "<*A> and <*B>" to look for sentences that have both *A and *B. However, using this function only leads to results in which a word coded as *A and a word coded as *B occur directly next to each other, right? There is no other way to look whether or not they occur in the same sentence, but maybe one at the beginning and one at the end of the sentence.

One more question: I tried to read up on the differences between Jaccard, Cosine, and Euclid coefficients, but I still do not know which one is best suited for my data. My txt files include multiple policy documents and/or speeches on climate change (some 5,000-10,000 sentences per file) and I am interested in the occurrences and co-occurrences of a range of words associated with threat, risk, national security, human security, and planetary security. I understand that Jaccard simply depicts the value for sentences in which *A and *B co-occur divided by all sentences in which either *A or *B occur. Cosine checks how often *A and *B occur on average/1000 words in the file, but how does it then define co-occurrence? And how exactly does Euclid differ from Jaccard and Cosine (I got that it uses adjusted frequency to ensure that the distance of the words is calculated based on occurrence patterns, rather than on whether each word appears frequently or not)?

Please excuse my very basic questions and thanks in advance!

[ko-ichi-h]

Dear user,

However, using this function only leads to results in which a word coded as *A and a word coded as *B occur directly next to each other, right?

No. It's not necessary that words occur directly next to each other. A word in *A at the beginning of the sentence, a word in *B at the end, then the sentence will be hit.

Jaccard, Cosine, and Euclid

If you use small units like "sentences" or "paragraphs", I recommend Jaccard.

Cosine is preferable when using speech or writing as a unit. The cosine is similar to the correlation coefficient. It is a measure of correlation rather than co-occurrence.

For small units like sentences, we should look at co-occurrence, and for large units like speech, we should look at correlation.

Best,