README Overview

[a-jacewicz]

@furkmak I believe that the README for this repo is good to go. I saw that you and Markus have made the majority of the commits on this repo (and believe Sebastian said we shouldn't assign looking over READMEs to Markus?) and as such was hoping you could look over it one last time and either mark it as ready to go or give me feedback on what should be changed. Let me know if you think someone else would be better suited for doing this!

[SebastianZimmeck]

@a-jacewicz (and @furkmak), a few questions. It would be great if you can revise the readme accordingly:

• In addition, mention-based ad tone is also needed (see above). ...

  Do you mean "In addition, the candidates dataset is also needed (see above). ..."? Or, if not, how does ad_tone_constructed interact with mention-based?

• Again, these are assumed to be cloned into the same top-level folder as the entity_linking repo. ...

  So, all the dataset should be cloned into the toplevel of the local ad_tone repo?

  There was not yet any explanation of the entity_linking repo? What does the entity_linking repo have to with ad_tone?

• Some csv files in those repos are too large to be uploaded to GitHub. You can download them through our Figshare page.

  Do we have the Figshare links?

• This repository contains code that generates two variables: ad tone mention-based, which codes ad as 'contrast, 'promote' or 'attack', as well as ad tone constructed, which is based on this flowchart.

  • My comment on the flowchart is the same as my comment on the race_of_focus flowchart. Basically, we need to explain its core points.

• This repo contains eight R scripts, three that deal with ad tone constructed and five that deal with ad tone mention-based. ...

  • Can we add a one- or two-sentence basic explanation of what each mention-based and constructed are and its main difference? I know there was a half-sentence in there. But that was not comprehensible to me reading it.

• The code in this repository creates two variables, ad tone mention-based, and ad tone constructed.

  What does "variables" mean? Two different types of classification results?

• This variable is available for the candidate ads in the 1.4m dataset.

  What is this dataset about? Add link and explanation and how it is relevant in this context.

• When traditional mention-based ad tone is available, we use that; otherwise we sum over ABSA results (also using race of focus). The variable is available for a larger number of ads, and the rest have no ad tone.

  This explanation is probably meaningful for someone who knows how the code works. For me it is unclear.

  • First, why use mention-based first? Is that more reliable.
  • "sum over ABSA results (also using race of focus)" What does that mean? Add links and a few sentences of explanation if important (or if not important, omit).
  • Also, not sure what the last sentence is saying. Add explanation.

• I have the same version questions that I have for race of focus. The revision of the text should be the same.

• In order to use the scripts in this repo, you will need to download the repository into a top level folder.

  What does it meant to download the repository into a top level folder? You mean I just clone (i.e., download the repo) and from the top-level of that cloned/downloaded repo I run the commands later given, load the datasets in that level, etc.?

• All the scripts for ad tone mention-based require ...

  Along the same lines, I am understanding that there are various other resources required. But how do I actually get them to work? Where exactly do I put them on my computer when downloaded? Do I just need to to put them there and that's it or do I need to do anything else. Please describe the details.

(cc'ing @sheoftensaid and @markusneumann)

[furkmak]

I revised this readme. Here is what's left to do:

• Instert Figshare links once ready
• in 3.2 Input Files section, you might want to give the link for input file locations instead of only the repo link. For instance, instead of https://github.com/Wesleyan-Media-Project/ABSA/tree/main, give https://github.com/Wesleyan-Media-Project/ABSA/blob/main/data/google_2022_ABSA_pred.csv.gz for Google 2022 related scripts.
• Can add a bit more detail under inference (what the result looks like, etc.)

Once these are addressed, this issue can be closed.

[a-jacewicz]

* in 3.2 Input Files section, you might want to give the link for input file locations instead of only the repo link. For instance, instead of https://github.com/Wesleyan-Media-Project/ABSA/tree/main, give https://github.com/Wesleyan-Media-Project/ABSA/blob/main/data/google_2022_ABSA_pred.csv.gz for Google 2022 related scripts.

* Can add a bit more detail under inference (what the result looks like, etc.)

I've gone through and done this. Now, we just need to add links once we have them!

[SebastianZimmeck]

Thanks, @a-jacewicz!

[a-jacewicz]

Figshare links have been added! In addition, links to Google 2020 data needs to be updated in READMEs for repos that needed it!