4-5 pm in Williams 320 (unless otherwise indicated)
| Date | Speaker | Title | Abstract |
|---|---|---|---|
| 9/15/25 | Ian Woods, Department of Biology | Data Talks: Comparative analysis of tardigrade locomotion, and the role of GABA in interleg coordination | Locomotion is a defining feature of animals and is essential to their fitness. In humans, movement disorders are associated with aging and with several child-onset genetic diseases. Effective treatments for most of these disorders are lacking, necessitating an understanding of intermolecular mechanisms regulating locomotion, and how these mechanisms change during dysfunction. As the smallest known animal that walks with coordinated limb movements, tardigrades are potentially a powerful model of limb-driven locomotor function and dysfunction. We developed open-source software to record and analyze tardigrade movement and quantified locomotion at broad spatial and temporal scales, generating a comprehensive dataset (>150 tardigrades, >13000 strides) of tardigrade movement. Comparisons across lifespan, species, and drug treatments indicated that circuits controlling coordination operate independently from those governing high-level aspects of movement such as speed, turns, and walking bout initiation. |
| 9/29/25 | Stephen Sweet, Department of Sociology | Data Talks: Using the General Social Survey to Teach Students Data Analysis Skills | This colloquium shows how students are engaged with the General Social Survey – data representing the attitudes and behaviors of nationally representative sample of adults in the United States. Students identify research questions that fit individual interests. With a goal to create posters suited for presentation, students examine how social experiences vary in accordance with contextual conditions such as gender, race, age, and political orientation. With a flipped classroom pedagogy, students complete 22 deliverables, demonstrating skills in understanding data structures/limitations, data management, hypothesis formation, univariate analyses, variable construction, bivariate analyses, graphing techniques and poster construction/presentation. Students commonly present findings at the Whalen Symposium. |
| 10/13/25 | Scott Erickson, Professor of Marketing | Data Talks: Marketing (and Business) Analytics | This talk will cover some basics of data generation and analysis in business in general and marketing in particular. The rapid development of digital systems for operating environments (ERP) and marketing (CRM) generate huge amounts of data. This “big data” and associated analysis techniques have totally changed how business is done today, especially in marketing. The talk will include numerous examples of how familiar brands are actually data companies, generating and applying their own proprietary data for decision-making. From digital twins (GE Aerospace) to product decisions (Spotify, Netflix), distribution and retail (Amazon, Walmart) to customer relationships and marketing communications (Google, Salesforce), the impact of big data and marketing analytics and its importance to high-performing firms is considerable. Data are typically kept in massive data storage and processing facilities, including those intended for artificial intelligence applications. Programs and technologies for statistical analysis vary, but this presentation will focus on the SAS Viya tool, an easy-to-use, drag-and-drop visualization software that lends itself well to the real-time monitoring and predictive analytics common to marketing (and business) applications. |
| 10/27/25 | Professors Aaron Weinberg & Emilie Wiesner, Department of Mathematics | DATA TALKS: Post-Grunge or Alternative Rock? Teaching a Machine to Decide | Music genre identification is useful not only for commercial applications like recommendation systems but also for deepening cultural engagement with music. Machine learning approaches have traditionally treated this as a “bucket sorting” problem, but music is experientially and culturally much more connected than discrete buckets. In this talk, we'll share process and results from a project to use deep learning techniques to generate a music genre classifier informed by genre theory and inter-genre relationships. We'll start by orienting the audience to the general field of machine learning and some of the particulars of deep learning, then discuss our process of data curation, the use of graph theory to model genre distance, and describe our process of creating and testing neural networks to perform the classification. |
| 11/10/25 | Venkat Govindarajan, Assistant Professor of Computer Science | DATA TALKS: Dark & Stormy: Modeling humor in the worst sentences ever written | Textual humor is a rich and varied phenomenon, but one understudied form is *intentionally bad* humor. In this talk, I present joint work with Prof. Laura Biester (Middlebury College) introducing a new dataset built from the Bulwer-Lytton Fiction Contest (BLFC): an annual challenge that invited participants to craft “an atrocious opening sentence to the worst novel ever written.” We explore two research questions: (1) What’s unique about this dataset of intentionally bad humor, and (2) How does synthetic data (BLFC entries generated from AI Models) compare to the human-written data? After outlining our data curation, preprocessing, and annotation pipeline, I’ll briefly review key NLP tools and methods and how I use them to derive insights into our data. We find that sentences in our data are uniquely different from standard humor datasets: they rely heavily on certain literary devices like metaphor, metafiction and simile, and contain many more novel adjective-noun pairings. Further, we find that AI-generated BLFC sentences mimic the form of the human sentences well but tend to exaggerate the distinguishing characteristics. |