Skip to content

Latest commit

 

History

History
84 lines (67 loc) · 3.85 KB

components.md

File metadata and controls

84 lines (67 loc) · 3.85 KB
Raw

Application components

This file describes the parts which make up the Word Vector Interface (WVI), an R Shiny application created by Jonathan Fitzgerald, Parth Tandel, and the Women Writers Project.

R code

A file called app.R consists of R code which sets up and runs the WVI. The Shiny application pulls all the other components together. The app:

  • reads the JSON catalog to determine which word embedding models to load;
  • loads those models, as well as their titles and descriptions;
  • constructs a skeleton web page, including any links to web assets that the browser should use;
  • lays out the form controls for each WVI tab (“Home”, “Compare”, etc.);
  • defines what happens when someone changes a setting or types in a query;
  • opens up access to the web application at port 3939 on the computer’s local network (localhost:3939); and
  • transforms data from the models into HTML.

Word embedding models

The Shiny application decides which models to load by reading the file catalog.json. This catalog contains basic information on models, such as their “short name” (what to call them in the model selector dropdown) and their location (where the model’s BIN file can be found).

An entry might look like this:

"WWO Corpus": {
    "shortName": "WWO Full Corpus",
    "location": "data/wwo_xquery-nonreg_allTexts.bin",
	    "shortDescription" : "All texts from WWO",
	    "public" : "true",
    "description": "Currently displayed: the entire Women Writers Online corpus of women's writing from 1526 to 1850."
  }

Note that models’ file locations should be relative to app.R, not to the catalog itself.

If an entry is marked as "public" : "true", the corresponding model will be loaded into the Shiny application. This repository contains a lot of models, including some that aren’t available via the Women Writers Vector Toolkit.

All models are currently stored in the data folder, but they could be stored anywhere, as long as the filepath listed in the catalog is correct and the model is accessible. The Shiny app only needs to know where the catalog can be found.

For more on how the word embedding models are created, see the Methodology and Downloads pages of the Women Writers Vector Toolkit.

Web assets

Once the Shiny server puts out a complete web page in HTML, more files are needed to make the page appear and behave as it does. The WVI makes use of Cascading Style Sheets (CSS) for styling, and Javascript for interactivity.

  • Shiny Dashboard is the R library that structures the web application. Installing the library also installs its pre-made CSS and Javascript, which give the web page its dashboard appearance. The R library automatically puts links to the assets in the output webpage.
    • Bootstrap provides CSS for a clean, flexible, and customizable foundation. The Bootstrap CSS is linked in the Shiny app’s output web page, and is hosted by a content delivery network, not a local file.
  • Local assets consist of CSS, JS, and image files specific to the Shiny app and the Women Writers Project. The primary files are:
    • main.css adds WVI-specific styles to the CSS library themes described above. R code places a link to the stylesheet in the output webpage.
    • script.js shows or hides the WVI sidebar on request. The contents of the file are inserted directly into the output webpage by the Shiny app.