Closed AnnikaTjuka closed 8 months ago
I think R is fine, and given that many people are more comfortable with it, it would even be easier for others to interact with the data. I'd re-run the R-code then and give it a check.
Sounds good. Then I'll prepare the analysis in R.
@LinguList I did some preliminary comparisons, but again, the data are very sparse. If we want to compare the directions and the psycholinguistic measures, we are left with only 20-30 colexifications for which we have ratings for both concepts. We could use the 100 body-object colexifications as a starting point to get more b-o colexifications, but then we cannot say anything about the influence of concreteness or valence on directionality. Since the yearbook editors encouraged the submission of ongoing research, I'd suggest we focus on the frequency of directions and networks for now. I added a table for the directionality analysis and network illustrations to the Overleaf doc. I'll include possible further studies using psycholinguistic measures in the discussion/outlook section.
Yes, this is a good idea. And what we can also do is: we add limitations with an outline session, and there we can refer to NoRaRE and even point to data sparsity!
This fits well with our plan to submit some project on a catalog of linguistic diversity (with Robert), where we wanted to systematically evaluate how to address such situations.
An example of addressing them (that would cost resource-wise) is the recent paper by Norcliffe and Majid, who identify their data targets and then collect the datasets. We can also cite their work as an example, but point out that the preliminary character of our study is very useful, as it helps to explore if a topic is worth being investigated...
Sounds good! I'll integrate the study on perception verbs as an example and point out the importance of targeted data collection. I'll finalise the draft over the weekend and send it to you on Monday.
To test whether ratings between two concepts of a body-object colexification correlate, we could use the following NoRaRe datasets:
Scott-2019-Ratings
Warriner-2013-AffectiveRatings
@LinguList I could set up an analysis in R or do you want to set it up in Python?