Barbara to check wording and integrate as needed

I've added a sentence about the Language of mechanisation. In addition, I suggest we explain the terms used, i.e. "workflows", "subjects"

Made minor changes to last paragraph

Changed bicycle back to bike :)

Barbara to liaise with Mia on how to explain concepts such as active/non active

We can put the concepts in a different notebook and then pull in as required.

I agree, we should explain what active and non active means

Barbara to check whether to add more contextual information as done for DAVIZCT notebook: https://observablehq.com/@jmiguelv/lwm-language-of-accidents

Barbara to check labels of axis and alert KDL if they need correcting

Barbara to liaise with Mia on how to explain meaning of concepts such as subjects

Format of year in table of subjects to be corrected if feasible

I've selected 'bike or motorbike' as my workflow and 'the image is too hard to read' as my meaning, but the headline above is 'Subjects from 1849, annotated as a horse-drawn private or public carriage for carrying passengers'

I can't reproduce, did you click on a year data point in the image is too hard to read subplot?

Could we have a little introduction explaining what this section is about?

should we explain what "live" and "visible" mean in Zooniverse?

agreed to link to glossary but need explanation by Mia

could we give a definition of annotation agreement?

we also decided to exclude data points with agreement below 0.65 from the subsequent analyses

Miguel added the description; regarding the agreement in our notes I have "percentage to be determined once other words are analysed so it seems best to wait for all words to be there but in any case threshold added below so left to the user to select it. I think this comment can be resolved now.

very cool!

I'm not sure I understand what "normalised by Observable" means

I was experimenting with the normalisation models available in the Observable Plot library :)

Format of year to be corrected if possible

very clear, thanks!

should we say somewhere that we only retain the annotations with agreement >=0.66 in our subsequent analysis?

done by default but agreement rate can be changed - agreed KDL to add

Added below - please review and resolve if fine.

I don't think this is needed at this point in the notebook anymore as information about which annotations are kept is described earlier in the notebook.

"Normalise...": maybe we should add something like "by dividing by the total number of annotations for that word in that year" at the end? "Display trend line": add "patter of the annotation counts for that word per year" "Display uncertainty...": representation-->representing. Could you also add how this was calculated (linear regression?)

by the way, I really like these plots!

COuld we add an explanation of how the curved lines were created to connect the dots?

Are the years being shrunk according to the amount of text in the classification option?

No, it has to do with the number of annotations per year. We could pin the min/max year, but it would push a lot of the relevant information to the right and most of it would be blank space.

I've emailed a screenshot for context. I'd expect the width to remain consistent as it's all options from the same workflow, but some were smaller than others

Followed up on the email in case others want to comment as well.

could we add links to the sections below about annotators' agreement and Zooniverse data so that the readers are reassured?

and also a link to the glossary

is it possible to control the parameters (trend lines, etc.) for each plot separately

I'd like to see this after the plot with the individual senses

replace email for feedback?

also, do we want to list all contributors with roles somewhere?

agreed at meeting to add Barbara's email

I don't seem to be able to see anything when I click on a year on the timeline. Also, I'm not 100% clear what "on the line of text from an individual article" means.

The line of text refers to the text snippet that shows on the table. If you click on it you can see the original image.

I wonder if it would make sense to group together the counties into broader sets (e.g. North, South) as well, in addition to the individual counties?

explain that this refers to our annotation dataset

This one is actually about all the data in the press directory, the one below is the project specific one.

or rather in our specific annotated dataset?

this text needs changing: we should add "King's College London" to the list of universities.

You should be able to edit at https://observablehq.com/@jmiguelv/lwm-backend-text, please let me know if you are having issues.

it worked!

Discussed with Barbara - she agrees nice to keep but might need to reduce and explain in text

Added text in italic but please refine if needed. Can I confirm what to reduce please? Size, number, speed?

speed please - text changed

I'm not sure what the grey lines represent. Also, what is the y axis?

Expanded the explanation.

this is very cool! But is there a way to display the main change point? In other words, when did one mean take over the other one? Is it possible to explain this? Also, would it be possible to include the changepoint detection method used in Nilo's code?

so if I look at bike, I don't see any changes in meaning, right? would it be possible to connect this table with the plots above so that the users can see a clear (as much as possible!) story? Maybe the table needs to be moved further up to best interpret the plots?

Yes, that is true for bike. Please see the other changes and let me know if you still think this should be moved up.

ok, so bike and coach didn't change, but the other two did, interesting. the text is clearer now so the table can stay where it is, in my opinion

Apologies @barbaramcg, I accidentally removed your comment! Barbara McGillivrayJul 14 08:37 UTC this is very cool! But is there a way to display the main change point? In other words, when did one mean take over the other one? Is it possible to explain this? Also, would it be possible to include the changepoint detection method used in Nilo's code?

With the current method it picks up change points based on the threshold, not necessarily a unique one. The method Nilo used was from a Python package, not sure if there is a version for JavaScript, but will look around to see how feasible it would be to port it. The plot below this one shows when meanings take over.

@barbaramcg, I have implemented a port of the PELT algorithm in JavaScript, would you like the notebook to be updated to use it instead of CUSUM for change point detection?

yes, I think that would make it easier to compare the results of Nilo's code. If possible and if it doesn't take too much space, you could include the results of both PELT and CUSUM. Thanks!