LFCS Seminars: 24 August 2018 - Tomas Petricek

Title: Programming as human data interaction

Abstract:

How do we enable more people to create transparent, reproducible data analyses that encourage further exploration and critical thinking? Today, we are able to prove large systems correct and machine learning keeps gaining new capabilities, but the way we interact with data has changed little over the last 20 years. There remains a gap in data science tooling: spreadsheets are widely used, but work on small data and limit reproducibility; Python or R scripts are reproducible and scalable, but require expert programming skills. In this talk, I will discuss my work on programming tools for data science, first in the context of the functional-first programming language F# and later in the context of TheGamma, an interactive scripting tool for web-based data exploration. On the formal side, I will discuss how the concept of "type providers" make it easy to integrate external data sources into a statically typed programming language.

On the human side, I will discuss how programming environments can learn from spreadsheets to make programming more accessible.

Reference:

The Gamma homepage (https://thegamma.net) and sample projects developed at the Alan Turing Institute (http://gamma.turing.ac.uk) as well as the F# Data library (http://fsharp.github.io/FSharp.Data). An ECOOP paper about TheGamma (http://tomasp.net/academic/papers/pivot) and a PLDI paper about F# Data (http://tomasp.net/academic/papers/fsharp-data).