You can find the main write up of this in this location. This document lists the locations of code used.
The first step to this analysis is getting data on the PD2M program (sesssions, talk title, authors, organization, abstract). The rvest package in R makes it easy to scrape websites. The code for scraping the data is here. The part which was tricky and required some manual cleaning was extracting organization names. I used the Standford Named Entity Recognizer through the R package StanfordCoreNLP for tagging the organization names and then cleaned up names that were incorrect.
Basic summaries by talk, sessions, authors is in this file
The estimation of underlying topics and summaries based on the estimated topics is in this file.
All of this analysis was done in RStudio (version 0.99.465).