2020 Kaleidoscope

SumUp Analytics
Oct 12 · 4 min read

Deciphering Election News Patterns [Week 2, 10/11/2019]

sumup.ai

This newsletter is a weekly publication showcasing two simple applications of SumUp’s platform: an analysis of news pertaining to the 2020 elections and a technical review of research published on Computational linguistics.

389 days to the elections

What topics are most emphasized across the mainstream media’s coverage of the 2020 presidential election?

To find out, we ran 292 articles from 13 news feeds through the SumUp platform. SumUp analytics is initiating a weekly blog using text analytics technology developed over the last few years to extract key content from a representative sample of the 2020 elections news. All information used is freely and easily available on the web. We didn’t apply any filters. Keep in mind that these results are only representative of the 13 news sources listed at the bottom of this article.

This blog is not expressing an opinion but rather highlights the most relevant topics in the recent news coverage.

This initial analysis was executed using a representative sample of 13 news feeds (in annex), combining all articles available on the first page of these publications on 10/11/2019. Different information sources would yield a different output. We made an effort to be impartial when compiling the input sources but we encourage you to provide feedback. Additional analytics will be provided as this blog evolves.

SumUp 2020 key topics on 10/11/2019, by order of relevance

Topic 1: Joe Biden/Donald Trump
Explanation: Even though mostly not mentioned together, Joe Biden and Donald Trump are the two subjects mentioned in most critical parts of the corpus analyzed. Both Donald Trump and Joe Biden are themselves mentioned in different contexts: impeachment and rallies versus campaigning and fund raising.

Topic 2: Health Care
Explanation: Even though health care appeared last week in the context of Democrats campaigning, health care appears this week again as a campaigning theme heavily tainted by Bernie Sanders health scare episode.

Topic 3: Impeachment
Explanation: Mentioned in three contexts: President Trump, Nancy Pelosi and White House, each with their respective relation to impeachment.

Topic 4: Pete Buttigieg
Explanation: mentioned in various contexts in particular polling, LGBTQ community.

Topic 5: Climate
Explanation: Climate change is mentioned in the context of the democratic agenda

Topic 6: Barack Obama
Explanation: Barack Obama is mentioned in the context of previous elections in Ohio and vice president Biden, directly or indirectly when mentioned by Donald Trump.

Topic 7: Democratic Primary
Explanation: Democratic primaries are mentioned in relation to polling in various states, in particular in Iowa and New Hampshire

Topic 8: Bernie Sanders/Heart attack
Explanation: Bernie Sanders and his health scare episode are also a significant part of the news corpus this week.

These topics are only representative of the key subjects appearing in the sources of information reviewed by Nucleus. A more refined interpretation, in terms of content or order of importance is left to the reader and could easily be pursued with the Nucleus platform. In order to facilitate that analysis, we added the actual Nucleus analysis to the annex, providing further measure of sentiment attached to each of these topics, or summary explanation of each topic.

Annex 1 Election 2020 news sources:
https://www.theguardian.com/us-news/us-elections-2020 https://fivethirtyeight.com/politics/elections/ https://www.chicagotribune.com/politics/elections/ https://www.politico.com/news/2020-elections https://www.cnn.com/specials/politics/2020-election-coverage https://time.com/5675691/2020-election-candidates-website-disability-accessibility/ https://www.newyorker.com/tag/2020-election https://www.foxnews.com/category/politics/2020-presidential-election https://www.usatoday.com/search/?q=2020%20democratic%20elections https://www.washingtonexaminer.com/tag/2020-elections https://www.washingtontimes.com/elections/ https://www.vox.com/2020-presidential-election https://www.breitbart.com/tag/2020-election/


Computational Linguistic Research Week of 10/11/2019

Leveraging its text analytics platform, SumUp has designed a simple method to rank recent research publications, purely based on their relation to the top topics extracted from these publications. Creating a corpus composed of all recent publications (all publications published on arxiv.org over the last week), nucleus extracts 8 key topics representative of that research corpus. Nucleus then identifies the top documents related to these topics. SumUp ranks articles, according to the number of times each article is mentioned in the 8 extracted topics.

Using the corpus composed of articles published during the week of the 11th October 2019, the following 8 key topics are extracted:

[Pretrained Natural Language Models] [RNN/CNN] [labelled training data] [hate speech detection] [transfer learning] [question answering] [fake news][entity recognition]

Using SumUp ranking methodology, the following articles are most representative of the top 8 topics of the week of 10/11/2019:

Capturing Argument Interaction in Semantic Role Labeling with Capsule Networks Xinchi Chen, Chunchuan Lyu, Ivan Titov
This article is mentioned in relation to pretrained natural language models, labelled training data.

Automatic Quality Estimation for Natural Language Generation: Ranting (Jointly Rating and Ranking) Ondrej Dusek, Karin Sevegnani, Ioannis Konstas and Verena Rieser
This article is mentioned in relation to pretrained natural language models, hate speech detection.

Linguistically Informed Relation Extraction and Neural Architectures for Nested Named Entity Recognition in BioNLP-OST 2019 Usama Yaseen, Pankaj Gupta, Hinrich Schutze
This article is mentioned in relation to RNN/CNN, labelled training data.

Contrastive Language Adaptation for Cross-Lingual Stance Detection Mitra Mohtarami , James Glass , Preslav Nakov
This article is mentioned in relation to labelled training data, hate speech detection.

Multilingual Question Answering from Formatted Text applied to Conversational Agents Wissam Siblini, Charlotte Pasqual, Axel Lavielle, Cyril Cauchois
This article is mentioned in relation to labelled training data, hate speech detection.

All of the other articles published during the period under review are individually less relevant to any of the top topics.

SumUp Analytics

Written by

Accelerate Understanding: Explainable AI for Text at Scale

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade