Short stories and novellas revolve around a single sequence of events, while novels contain more subplots. Can we see this quantitatively by examining word usage? We look at power and danger time series of Project Gutenberg books to find out...
A naive approach to querying social media documents to measure public sentiment can introduce irrelevant text that introduce noise or result in misleading measurements. We test a transfer-learning powered classifier with excellent performance that can be fine-tuned with minimal labeled data.
Transitioning to renewable electricity generation is an important component of the US response to climate change. The speed of this transition depends on factors ranging from the rates of technological innovation, market adoption and governmental incentives...
New preprint from our group: https://arxiv.org/abs/2205.15937 We know spending time outside is good for our health, and maybe more so in March 2020 than ever before. Dogs were psyched. But confusing and contradictory public health messaging complicated access to natural spaces...
New on the arXiv: “Ousiometrics and Telegnomics: The essence of meaning conforms to a two-dimensional powerful-weak and dangerous-safe framework with diverse corpora presenting a safety bias” https://arxiv.org/abs/2110.06847 We measure essential meaning using language as the map. Some definitions: 1. Ousiometrics: The quantitative study of the essential meaningful components of an entity, however perceived. 2. Telegnomics: The far sensing of knowledge (~ distant reading) ...
"Sentiment and structure in word co-occurrence networks on Twitter" New preprint from our group led by @uvmcomplexity postdoc @MikaelaFudolig Using tweets containing #imwithher or #crookedhillary, both relating to Hillary Clinton's presidential bid in 2016, we explore the relationship between context and sentiment through word co-occurrence networks. ...
“Augmenting semantic lexicons using word embeddings and transfer learning” New preprint from our group led by Thayer Alshaabi https://arxiv.org/abs/2109.09010 In prior work, we surveyed internet people for their emotional response to words. Surprising no one, we found "hahaha" consistently reflects the good vibes. ...
We have a new paper, interactive visualization, and data platform. Nutshell: we’ve curated 100 billion tweets over 10 years to produce day-scale rank/frequency time series for n-grams in over 100 languages. It’s a whole big thing. A short thread—The paper: “Storywrangler: A massive exploratorium for sociolinguistic, cultural, socioeconomic, and political timelines using Twitter” ...
"Blending search queries with social media data to improve forecasts of economic indicators" New preprint led by Yi Li & Adam Fox from @mmdatascience w/help from our group https://arxiv.org/abs/2107.06096 This work extends a model recently proposed by Dingdong Yi, @ShaoyangNing, Chia-Jung Chang, & S. C. Kou in their paper "Forecasting Unemployment Using Internet Search Data via PRISM" ...
