From cs448b-fa16-wiki
Jump to: navigation, search


Visualizing Literary Influence in the Paris Review

Final Deliverables

Paper: File:Alec-glassford-final-paper.pdf

Poster: File:alec-glassford-poster.jpg

Code: Available on GitHub

Executable: Available as a web app


Group Members

  • Alec Glassford


For over 50 years, The Paris Review has been publishing a series of in-depth interviews with some of the world's best-known literary writers. One key aspect of these interviews is the writers' discussion of their influences. Graphs are a popular way of representing social networks, and trees are a common way of displaying hierarchies; I would like to build on these methods to make a tool that allows users to visualize the network of influence between different writers who have been interviewed by The Paris Review (and possibly other artists).

This will involve scraping the interviews and mining named entities from them to acquire the data. For visualization, I would like to not only create a directed graph of influence links between the writers, but also allow users to explore the connections of a particular author: easily view their influencers and influences, contextualize influence (e.g. provide the sentence where the influence was mentioned and a link to that interview), find paths of influence between any two given authors. I think this work would have wider applicability to visualizing networks of influence and also the process of gathering data on these networks from webpages.

Project Progress Presentationl


Literature Review

Vizster: Visualizing Online Social Networks This project visualized social network friendship graphs from the website Friendster. It uses node-link diagrams to show the big picture (filtering—by gender for example—and clustering help provide insight on the macro- level), but also provides an interface for more micro- exploration with zooming + panning, search, and focus+context views.

NodeTrix: A Hybrid Visualization of Social Networks This project combines node-link diagrams and adjacency matrices to visualize networks, focusing on the idea that node-link diagrams give a good sense of global structure, while adjacency matrices are better for close analysis. Users can interactively move between the two views as they browse. The software is applied to a network of researchers, with co-authorship of papers represented as the edges.

Social Network Visualization: Can We Go Beyond the Graph? These researchers did a user study on two different visualizations of email networks, Social Network Fragments and PostHistory. The former is a node-link diagram of email contacts, while the latter visualizes patterns of email over time. They noted that the two visualizations were especially effective in tandem: considering the two different perspectives on their emails together gave users the best understanding of their social network.

Mapping the Republic of Letters This Stanford digital humanities project explores the relationships between Enlightenment thinkers. Among other things, the researchers presented visualizations of thinkers' correspondence, their patterns of publishing, their travels, and their membership in different organizations. These visualizations take a number of forms, including node-link diagrams and maps. One key feature of many of the visualizations is interactive access to the original source documents behind the visualization. For example, in a node-link overlaid map of Voltaire's correspondences, users can click on particular edges or nodes to get info about the original letters involved with that part of the visualization and click through to the see the documents.

Project Plan

Already done: First iteration of data-gathering, basic node-link visualization.

By 11/24: Build out the view for an individual writer - including context of original interview available

By 11/28: Build out webapp to allow browsing between individual views/building up of network + basic homepage/search interface.

By 12/2: Refine data gathering/transformation - first to improve accuracy/precision of intra-Paris Review influences (e.g. search by last name only, w/ more sophisticated named entity recognition), then (if time) try using Google Knowledge Graph to pull in outside influences

By 12/5: (If time), work on integrating centrality measures + clustering.

By 12/7: Make poster and send to printer

By 12/11: Write paper