A guide to collecting tweets with Netlytic

spapastamkou commented 1 year ago

Ranke.2 has received a tutorial under the title: A guide to collecting tweets with Netlytic prepared by Frédéric Clavert @inactinique and Hannah Smyth @hasmyth hasmyth. A first version can be read here: https://github.com/C2DH/ranketwo-submissions/blob/master/lessons/tutorials/guide-collecting-tweets-netlytic.md The history of the edits can be viewed here. This issue documents the submission before publication and allows to work with the authors in order to produce a final version.

spapastamkou commented 1 year ago

@inactinique I submit below some comments and suggestions that you can check with your coauthor. Please let me know if she has a GitHub username to give her access to the repo as well. There is no particular deadline, and your is the first lesson that will allow us to test this way of working Ranke.2's lessons via a GitHub repo.

spapastamkou commented 1 year ago

To the authors: I submitted a first (or still zero, if you prefer) version of a text I initially had access to via a Gdoc. The version you find here is a revised one. I list below the main edits I made

Main edits

I reorganized the beginning of the tutorial: parts 1 and 2 of the Gdoc, under titles Goals and Requirement respectively, do no longer exist as such, their contents have been redistributed in two subparts (of equal level) named: Introduction; Overview of Netlytic. For both parts, I removed any mention made to nodegoat, as nowhere in the text this is further developed.
- I added several phrases here and there trying to make the text either more fluent or more tight between different parts or simply to create text. You can get an idea if you check this commit or this one.
- I added captions to the screenshots that are provided. You should however check them and see if you are happy with them or if you wish to edit and modify them.

Images

I downloaded the image files form the Gdoc I had access to and provided file names based on the lesson's slug here. You will find them in this folder of the repo.

Some of them are much too cropped and we miss elements of text at the top. More importantly, the resolution is not very high - or it might be that the image size provided is too small? In any case, when one tries to zoom to see better, the quality deteriorates sensibly. Which is why I am afraid you have to produce them again. I will check in advance with Daniele and Frédéric Reiss if there are no specific requirements on the format or size of the image files.
Image files 2 and 7 (with current numerotation) are exactly the same: this means that the second and the the seventh image of your Gdoc (in the order they appear there) are identical. For the time being I did not integrate a link towards the the 7 in the text available in this repo: please check your initial file and let me know what to do here.

Suggestions (additions, modifications...)

They are organized following the order of the subparts of the current text.

Introduction

Learning goals and requirements: I propose to add a phrase among the requirements some basic notions of text and network analysis (although I propose, in the relevant parts, some additions to explain terms that are used). Especially in the part on the network analysis, specific terms of the graph theory are used without prior introduction of the reader to them.
Mention that text and network analysis are two possible analysis that can be applied to tweets datasets by humanists and social scientists (perhaps this is not clear to everyone, especially beginners).
After the very first phrase of the introduction and before the ones that follow explaining what the tuto is about, I would suggest to add a phrase stating that Twitter data are useful as primary sources for historians and mentioning why and in what kind of analysis. This is not obvious to everyone.

Setting the scene: what is the Twitter API and how does it work?

This part is currently organized around three assignments that the user has to execute alone. But this tutorial was conceived as a teaching aid: so I propose that this part does not take the form of assignments. Instead, use the current organisation of the contents to create text as follows:

provide at least basic definitions and information about: 1) what an API is 2) what it is useful for 3) what possibilities it offers 4) what limitations the user is subject to.
explain the technical terms about the different kinds of APIs one has to deal with as a user, thus guiding the audience to understand this kind of unfriendly stuff:-)
if further reading is proposed (for the REST API and Covid stream) please explain why: which is the purpose? what the user will gain in knowledge to consult these resources? If you wish, at the end of this section you can mention that an advanced user could create assignments based on this part.

The basic idea here is to provide key knowledge to the reader through text instead of sending them elsewhere to let them obtain this knowledge themselves. They can go for further information, if they wish.

What follows in the this part is currently almost in bullets form or a list of phrases and should be developed more like a fluent text. I make some remarks for each point separately that are currently visible and in italics in the file of this repo: https://github.com/C2DH/ranketwo-submissions/blob/master/lessons/tutorials/guide-collecting-tweets-netlytic.md#setting-the-scene-what-is-the-twitter-api-and-how-does-it-work-

Choose your hashtag and build your dataset

What is a "trending topic"? ==> Provide a basic def or explain in a few words
"follow the chain of hashtag links to collect the most relevant ones for your query." => this phrase is not very clear: what do you mean by "chain of hashtag links"? The simultaneous presence of many hashtags in a tweet that reveal more hashtags of interest and allow for a snowball effect?
"This is known as the “snowballing” technique." => provide a link to the concept (for example: https://en.wikipedia.org/wiki/Snowball_sampling )
I added a phrase: "Of course, you may have already a defined object of interest and look for specific keywords independently of the trending topics."
"How big should your corpus be? This will depend on what you want to study and how you want to study it" => Evoke other factors a well, the size of the corpus can also depend from the quantity of tweets produced around a given subject, the duration of the harvesting, the capacity - whether technical or tied to the economic model - of the tool that is used, the limitation of the Twitter API... The researchers can not always master the size of the corpus. Perhaps ask the question: what size is the good size for reliable research conclusions? Also, this question comes in the middle of points that deal with more practical matters of search; perhaps place it either in the beginning or in the end of the section?

Text analysis

Is it possible to add one phrase or two each time to remind the reader:

what is the purpose of creating a word cloud, via Keyword Extractor, of historical data
what is the purpose of obtaining groups of words linked to concepts, via Dictionaries, when analyzing historical data - are we talking about topic modelling here? If yes why not state it openly and link to a wikipedia lemma?
what is the purpose of obtaining tree graphs?
"You may then visualize the Dictionaries as a “tree graph”." => what is a tree graph and what is the purpose of obtaining one
Specify whether it is possible to export an image file of the visualisation of the word cloud? And in what format
data cleaning: explain what this involves roughly or provide a link? The idea is to make a beginner understand in simple words why these things would be useful to them.

Network analysis

"self-loop": explain what this represents
"The ties represent interactions between tweeters, which can be of different types: retweeting, quoting, mentioning, replying." => explain in simple words and briefly what each of these interactions represent. Perhaps a new user is not acquainted enough with Twitter's terminology and built-in functionalities.
In this part, it is mentioned without distinction that the nodes are "tweeters" or "Twitter accounts"; the screenshot taken from netlytic, distinguishes the types of interactions between "users". It all comes to name the same thing in different ways. Perhaps it would be easier for the reader if the authors pick one term (or explain that all terms refer to the same thing, only for this to be clear).
"you can uncheck “layers” to remove noise and focus on larger nodes." => This phrase needs to be reformulated: what are layers, what is noise, why focus on larger nodes (and what are larger nodes): all this can be very cryptic for a beginner.

Reading suggestions

LSE Blog - Using Twitter as a Data Source ==> full reference and link?

Going viral: How a single tweet spawned a COVID-19 conspiracy theory on Twitter ==>full reference and link?

spapastamkou commented 1 year ago

Dear @danieleguido and @frederic-reis As you can see in this commit I added links towards the lesson's image files as per the usual markdown syntax to have the images display in this repo. My first question is: should we continue to have the system described in this wiki page to add images in the text (I would personally be glad if we can avoid the step that includes entering the link in a gsheet) And my second question is: should I include the images folder of this repo in an assets folder (to create here) so that this repo reproduces the structure of the ranketwo repo? Thanks! (Nothing urgent here and we can talk further about his elsewhere if you prefer).

spapastamkou commented 1 year ago

Noting here that I received reply of FReis via email.

spapastamkou commented 1 year ago

With Twitter pre-announcing changes about the possibility to access its API for research purposes, we may have to wait and see what the implications may be for this tutorial. Apparently more information should be available by next week.

jdh-observer commented 1 year ago

Just saw that indeed...

APITwitter

inactinique commented 1 year ago

Oops, wrong account :)

Just saw that indeed...

hasmyth commented 1 year ago

I was wondering if/when something like this might happen :(

spapastamkou commented 1 year ago

Quite unpleasant, but apparently it was a matter of time:(

hasmyth commented 1 year ago

Hi,

I just saw this update from Twitter Developers: https://twitter.com/TwitterDev/status/1641222782594990080

Not good, but it may mean we can at least publish the lesson with some revisions and major caveats about the data that can be collected. I have not seen any updates from Netlytic yet about how they are adapting to the new API restrictions.

The free version is so restrictive as to be unusable for viable research, so perhaps reframe the lesson as very much a ‘sandbox’ or piloting and not something you can truly extrapolate results from at a larger scale (unless the topic they collect really has only a few hundred tweets and is therefore more likely to be a more ‘complete’ dataset).

Perhaps also an opportunity to add some critical note on the volatility of social media research and power dynamics….

What are your thoughts on how to proceed?

Hannah

Dr Hannah Smyth Lecturer in Archives and Records Management Dept. Information Studies University College London Office: G19 Foster Court TEAMS/Tel: +442031083640

Pronouns: She/Her -- Sí/í UCL Staff Page: https://www.ucl.ac.uk/information-studies/hannah-smyth Twitter: @Han_Smythhttps://twitter.com/Han_Smyth?lang=en

From: Sofia Papastamkou @.> Sent: 02 February 2023 16:43 To: C2DH/ranketwo-submissions @.> Cc: Smyth, Hannah @.>; Mention @.> Subject: Re: [C2DH/ranketwo-submissions] A guide to collecting tweets with Netlytic (Issue #1)

⚠ Caution: External sender

Quite unpleasant, but apparently it was a matter of time:(

— Reply to this email directly, view it on GitHubhttps://github.com/C2DH/ranketwo-submissions/issues/1#issuecomment-1414041546, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AHG3GHXYWYRVR63RTN2KQZDWVPPZPANCNFSM6AAAAAASUITGDE. You are receiving this because you were mentioned.Message ID: @.**@.>>

jdh-observer commented 1 year ago

Hi all,

at this stage I would wait -- we'll see end of April if there's something for research and if netlytic maintains its service.

Frédéric

From: Hannah K. Smyth @.***> Sent: 30 March 2023 16:13:13 To: C2DH/ranketwo-submissions Cc: jdh.admin; Comment Subject: Re: [C2DH/ranketwo-submissions] A guide to collecting tweets with Netlytic (Issue #1)

Hi,

I just saw this update from Twitter Developers: https://twitter.com/TwitterDev/status/1641222782594990080

Not good, but it may mean we can at least publish the lesson with some revisions and major caveats about the data that can be collected. I have not seen any updates from Netlytic yet about how they are adapting to the new API restrictions.

The free version is so restrictive as to be unusable for viable research, so perhaps reframe the lesson as very much a ‘sandbox’ or piloting and not something you can truly extrapolate results from at a larger scale (unless the topic they collect really has only a few hundred tweets and is therefore more likely to be a more ‘complete’ dataset).

Perhaps also an opportunity to add some critical note on the volatility of social media research and power dynamics….

What are your thoughts on how to proceed?

Hannah

Dr Hannah Smyth Lecturer in Archives and Records Management Dept. Information Studies University College London Office: G19 Foster Court TEAMS/Tel: +442031083640

Pronouns: She/Her -- Sí/í UCL Staff Page: https://www.ucl.ac.uk/information-studies/hannah-smyth Twitter: @Han_Smythhttps://twitter.com/Han_Smyth?lang=en

From: Sofia Papastamkou @.> Sent: 02 February 2023 16:43 To: C2DH/ranketwo-submissions @.> Cc: Smyth, Hannah @.>; Mention @.> Subject: Re: [C2DH/ranketwo-submissions] A guide to collecting tweets with Netlytic (Issue #1)

⚠ Caution: External sender

Quite unpleasant, but apparently it was a matter of time:(

— Reply to this email directly, view it on GitHubhttps://github.com/C2DH/ranketwo-submissions/issues/1#issuecomment-1414041546, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AHG3GHXYWYRVR63RTN2KQZDWVPPZPANCNFSM6AAAAAASUITGDE. You are receiving this because you were mentioned.Message ID: @.**@.>>

— Reply to this email directly, view it on GitHubhttps://github.com/C2DH/ranketwo-submissions/issues/1#issuecomment-1490381189, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ATM3XRIWQL6S5VWEHKRDASDW6WIHTANCNFSM6AAAAAASUITGDE. You are receiving this because you commented.Message ID: @.***>

spapastamkou commented 1 year ago

Dear both @hasmyth and @inactinique @jdh-observer, let's exchange on this again towards the end of April - as far as we are concerned for Ranke2, we are reworking on the project's website and we set ourselves a deadline for the 1st June for the changes to be effective. With @danieleguido and @frederic-reis we agreed to use your lesson as a test case for applying our new metadata schemas and website structure as it should be the first publication of the renewed project (which is why we opted for not integrating it in the current website). Thanks again for your work!

spapastamkou commented 1 year ago

@inactinique Just note here my final edits in this commit: https://github.com/C2DH/ranketwo-submissions/commit/53af0e86eebef6e6d64e74bcff5b20e95009d421 @danieleguido I also previously informed the yaml metadata according to the new scheme - we should check together this. Thanks!

C2DH / ranketwo-submissions