Extracting Communication Networks - Challenge

[jamesallenevans]

First, describe a social, cultural, or semantic network explicit within, implicit from or underlying your data. This could be the interaction between posters on a social media platform, or comments and reactions on a discussion site, or back-and-forth in a parliamentary debate, or shared stance on an issue (e.g., a stock price, political perspective), or a shared style of speech or focus, or characters within a fanfiction universe, or concepts within a discourse, or constitutions sharing ideas and phrases. Second, state two hunches you have about patterns in this network, with an asterisk (*) after the one about which you are most certain, and a plus (+) after the one that, if true, would be the biggest or most important surprise to others (especially the research community to whom you might communicate it, if robustly supported). Third, describe the dataset from which you will construct or extract this network for exploration and note whether this data could be made available to class this week for evaluation (not required...but if you offer it, you might get some free work done!) If available, place (a) a link, (b) a script (to download and/or clean), (c) a reference to a class dataset, (d) or an invitation for a TA to contact you to get it. Fourth, list in numbered steps what you would do to construct/extract the network from this data. Please do NOT spend time/space explaining the analytical strategy through which you would explore your network and consider your hunches (Then upvote the 5 most interesting, relevant and challenging challenge responses from others).

[romanticmonkey]

Possible network: The diffusion of film criticism concepts (or style of speech) between professional film critics and the general audience.

Hunch: (1) The general audience may mimic the reviews of professional critics when they're making their own. (2) Popular judgement might alter that of professional critics. (e.g. Marvel films might receive less "artistic" reviews and more reviews on their motivational aspects from critics as their popularity rises.)

Datasets: Rotten Tomatoes critics reviews (https://www.kaggle.com/stefanoleone992/rotten-tomatoes-movies-and-critic-reviews-dataset) Amazon Movie & TV reviews (need to filter TV) (https://nijianmo.github.io/amazon/index.html)

[jinfei1125]

Possible network: The Covid-19 related policies posted by governors on Twitter

Hunch: (1) * It is related geographically (for example NY-NJ-CT) (2) It is related by the party (for example democrat governors' may share similar guidelines?)

Dataset: US governor's Twitter posts in 2020

[Raychanan]

Possible network: The interaction of writing style and text content between entertainment stars and their fans on social media

Hunch: (1) Fans may imitate the writing/expression style of their idols on Twitter/Instagram. For example, the frequency of hashtag usage. (2) Entertainment stars may also be influenced by what their fans care about and provide more relevant content. For example, their selfies.

Database: Twitter/Instagram

[RobertoBarrosoLuque]

Network in our data: The interaction between democratically elected representatives (i.e. politicians) and media organizations that cover their actions and behaviour. Two hunches:

1. The behavior, rhetoric and actions of politicians and their coverage in the media have a symbiotic relationship in which each agent depends on the other for relevancy. **

2. The relationship/dynamics between politicians and the media coverage is adversarial. ++

   The datasets to study this relationship are:

   • The corpus of presidential speeches: https://millercenter.org/the-presidency/presidential-speeches
   • The NOW corpus of news media:  https://www.english-corpora.org/now/

[Rui-echo-Pan]

Possible network: The interaction of expression style among males and females in a working group / social media discussion platform.

Hunch: (1) When there are all female in a working group/online discussion group, their expression style tends to be more gentle and euphemistic rather than very straight; the presence of a male in the discussion group may change the style of discussion; (2) Similarly, compared to an all-male group, the discussion with the presence of female tends to be more polite and gentle. (3) It will also be interesting to see the cite/response relation among people.

Data: Reddit, or other online discussion platforms.

[jacyanthis]

Network: Twitter users discussing AI

Hunch 1: AI insiders (i.e. those with many connections to AI-focused Twitter users) tend to be the first to share a new AI idea or news story.*

Hunch 2: The discursive patterns of AI influencers (i.e. those with highest AI-focused follower counts) are more weakly correlated with future AI discourse than AI intelligentsia (i.e. those with high proportions of AI influencer follower counts, but not high total AI-focused follower counts). This is not particularly counter-intuitive, but I don’t have many criticisms of the current study of AI discourse (in part because there is so little of it).+

Dataset: I have access to the Twitter API so could download AI-focused tweets and user info, but I have not yet downloaded and cleaned that data because it is not the focus of my final project. My main dataset of interest, NOW, does not really represent a network because I don’t think I have hyperlink data.

[chiayunc]

1. A semantic network formed by legal terminologies within the UNFCCC that represent how close in senses they are with each other.
2. Procedural obligations like transparency and cooperation have growing importance (becomes more central)(*) these obligation extend to local/national level (domestic legal instrument appear in adjacency)(+)
3. decisinos and resolution from UNFCCC COP.

[MOTOKU666]

Network: China Mainland Twitter users discussing political issues H1: Mainland Twitter users had a greater tendency to behave radically under controversial topics (Chinese citizens are generally "apolitical" in the domestic online network. It may be different if they get a chance or "deliberately" logs in the Twitter) H2: if geographical location identified, the closer to the coastal city, the more frequently they comment on political topics (Since logging to Twitter requires VPN in Mainland and relate understanding of it is often popular in Coastal Mega city like Shanghai, Shenzhen, etc. ) Dataset: Twitter post, country of birth, and geographic location

[Bin-ary-Li]

Possible network: The mention of other artists in the description of another artists' artworks

Hunch: (1) This can form a network of artists connected by how often they are mentioned in each other's works. (2) If performing cluster analysis we might be able to extract some sort of "school of style" or "art style" information.

Dataset: Sotheby's art auction collection.

[theoevans1]

Network: My data concerns the network of fans who participate in fanfiction communities, by writing stories as well as by reacting to and commenting on each other’s stories. This network could also be conceptualized as including the original TV shows these works are based on and their creators.

Hunches: I suspect that fanfiction stories gradually become more linguistically similar to popular stories for the same show/genre.* I suspect that TV shows are influenced by their fans, and become more similar to their fanfiction over time.+

Data: Scraping data from Archive of Our Own (http://archiveofourown.org/) using this script (https://github.com/radiolarian/AO3Scraper), along with the Davies TV Corpus for the second hunch

[hesongrun]

Possible Network: My data consists of company announcements and news. Different companies organized within the same industry may be connected together in the network. Companies in the same chain of production will be closely linked together.

Hunchs: The connectedness of the network will represent the integration of the economy. I suspect the integration of the network will change together with globalization and integration of the world economy.

Data: Company news, company announcements

[dtanoglidis]

Network: A semantic network of the descriptions of qualities of Airbnb listings (e.g. words that are "closer" to the word "room")

Hatches: (+) The structure of the network depends on the specific place, there are different qualities depending on the city described. (*) The structure of the network also relates/and or can be a predictor of the ratings of a listing

Dataset: Inside Airbnb (http://insideairbnb.com/)

[egemenpamukcu]

Network: A network of American actors as the nodes and weighted edges for every movie/TV series they appear together.

Hunches: The communities within the network would correspond to different genres. * The nodes that are at the intersection of different communities (nodes with high betweenness centrality?) could identify elite actors with higher number of awards (Oscar, Golden Globe, Emmy etc. ) +

Dataset: IMDb dataset https://www.imdb.com/interfaces/

[mkjang17]

Network: The network between political discourse about communicating Covid and media coverage of the pandemic across different media outlets

Hunches: Conservative media would be more closely related with the government discourse under the Trump administration * Conservative media are more related amongst themselves, and same for liberal media+

Dataset: Press releases, speech transcripts, news articles

[william-wei-zhu]

Network: a network of the office characters that appear in the same scene.

Hunches: Michael Scott is at the center of the characters network.

https://www.kaggle.com/nasirkhalid24/the-office-us-complete-dialoguetranscript

[jcvotava]

Network: The sharing of ideas across texts and subtexts within the early European socialist movement.

Data: I have nicely clean and available the full text of Marx's Capital, available here: https://drive.google.com/file/d/1FEsftocEUwouZKGWmBtv1xoP0YHU3-7T/view?usp=sharing I have a script I can share which scrapes Marxists.org/archive to extract texts from the entire text corpus, however, I am still working on properly cleaning these texts due to inconsistencies in the underlying structure of the archive. So for this challenge prompt I will just focus on the one Marx text unless someone wants to reach out for the full corpus.

Hunches: 1. The subsections appearing at the end of the text (i.e. Section VIII Chap 26-33 where Marx tries to give a history of the emergence of capitalism) are only weakly related to the theoretical subsections that comprise the majority of the work (compared to how strongly the theoretical subsections are internally related to each other). *

2. Part I of the text (on commodities, chap 1-3) are more weakly related to the rest of the text than Parts 2-7 are to each other. +

[lilygrier]

Possible network: A semantic network of legislative bills related to energy policy that cluster around energy topic (emissions, pipelines, solar, etc.) and positions (regulation vs. anti-regulation). Hunches: (*) The communities created by the network would more greatly reflect topics, as positionality is harder to pick up on. (+) The communities created by the network would reflect pro- vs. anti-regulation more than they would reflect the topic (e.g., emissions vs. nature-preservation) Data: The BIll Summary US Training Data (should filter to include bills with "energy" or "environment" in their text as a crude filter).

[k-partha]

Possible network: Reddit personality forums

Hunches: * Similar personality forums tend to cross-reference each other more often than dissimilar ones.

• Forums for high Openness individuals are over-represented/unusually more active

Data: Reddit MBTI forums.

[xxicheng]

Network: A network in the show Gilmore Girls.

Hunches: Rory Gilmore is the center of the character network.

Data: http://www.crazy-internet-people.com/site/gilmoregirls/scripts.html

[ZhouXing19]

Network: A network of NSF grants projects, and the network of their authors Hunches: The authors in the close network are of close probability to receive the grants Data: NSF