VisRecly

What Is This All About?

VisRecly is a task-based visualization-recommendation tool with the goal in mind to allow novice VIS users to create novel visualizations quickly for their dataset, solely by uploading it, then specifying what they would like to achieve with their data (e.g. correlate, find differences, cluster, etc.).

This idea and project were so interesting to me that I decided to make it the project for my Bachelor's Thesis at the CS department of Universität Wien. I enjoy the supervision of Torsten Möller and Manfred Klaffenböck.

You can read the full thesis here.

Decomposing the Project's Name and Ambitions

The first part, VisRec, stands for Visualization Recommendation, while the ly segment is a tribute to Grammarly (the grammar checker) which inspired me to jump into this project. As I exposed myself to formal, academic concepts of visualization and visual data analysis, I realized that a Grammarly-like tool for visualization would be beneficial to VIS novices.

There is an untold story behind every single set of data and visualizations provide great means to tell these stories. Yet, just as people do not enjoy written stories being full of grammar or spelling mistakes or being out of tone, they also do not enjoy looking at visualizations violating fundamental design guidelines or being out of context regarding the task at hand.

On the one hand, as Grammarly dictates grammar and spelling rules for writing, VisRecly aims at enforcing non-negotiable visualization design guidelines in its recommendations. On the other hand, just as Grammarly allows for recommending synonyms and alternative sentence structures based on the communication goals and target audience of the author, VisRecly attempts to help its users find the "tone" of their visualization by ranking recommendations across tasks and highlighting the most suitable ones.

Tech Overview

In an attempt to make this tool as intriguing and accessible as possible, I am building it with modern web technologies, such as:

React, Next.js
Material-UI, TailwindCSS
Pure TypeScript & WebAssembly

This monorepo is made up of the modules below:

libs/data: Houses example data sets and related utilities.
libs/draco: The core of the underlying recommendation engine, defining learning and visualization design guidelines as Answer Set Programming (ASP) problems. It is a custom fork of draco by UW Interactive Data Lab.
libs/draco-web: Custom web-API leveraging the core API introduced in libs/draco and clingo-wasm to solve ASP programs in the browser, eliminating the need for a server component.
libs/ranking: Defines the ranking algorithm and aggregator utilities, considering both data-oriented costs (obtained from libs/draco-web) and VIS-task-relevant preferences.
libs/vis-tasks: Defines the VIS tasks and their associated mark-preferences.
apps/dashboard: The actual client of the above modules, the dashboard that allows users to steer their desired tasks and marvel at the generated vega-lite-based visualizations.

Architectural Overview

(a) Input: accepts user input (apps/dashboard)
(b) Preprocess: prepare data schema (libs/draco)
(c) Generate: Solve ASP problems and produce Vega-Lite specifications (libs/draco-web)
(e) Rank: Compute recommendation costs and rank visualizations (libs/ranking)
(g) Output: Render the ranked visualizations (apps/dashboard)
(d) and (e) are knowledge sources for design guidelines and VIS tasks (libs/data, libs/vis-tasks)

Development

Make sure to install the monorepo dependencies.

npm install

Starting the Project Locally

npm run start

The dashboard should be running on localhost:4200.

Running Tests

~~Running what?~~

Tests are available for libs/draco. They can be executed by running:

npx nx test draco

feat: ranking API

Description

The full pipeline & API for ranking should be created. It should be handled by a separate library, depending on the draco and draco-web modules.

Requirements

Create models to represent VIS tasks and their ranking logic depending on a given encoding
Create default VIS tasks and their ranking logic (at least a stub)
Expose a single ranking function, accepting the following input:
- data as any[]
- encoding preferences of the user (which columns should be included)
- maximum number of models
- --> should output an augmented SolutionSet with $n$ costs per recommendation

Additional Context

The VIS tasks and their associated ranking logic must be declarative and easily modifiable. This is absolutely necessary to be able to enhance the tool from a domain POV without making significant changes to the codebase.

Ranking Approach

The proposed ranking algorithm has two main pillars:

Data-oriented: obtained from draco by reading the cost of a model for an ASP
Task-oriented: "hard-coded" based on relevant literature, by declaring how useful a given encoding is for a specified task

Non-exhaustive list of used literature:

The recommendations are generated as vega-lite, hence, the used marks need to be matched using its schema.

Currently, the following marks are specified in the schema:

"Mark": {
  "description": "All types of primitive marks.",
  "enum": [
    "arc",
    "area",
    "bar",
    "image",
    "line",
    "point",
    "rect",
    "rule",
    "text",
    "tick",
    "trail",
    "circle",
    "square",
    "geoshape"
  ],
  "type": "string"
},

In a very high-level overview, the steps to take are:

Let draco process the raw data with its wisdom collected as .lp files as a part of uwdata/draco#51 and uwdata/draco#52, and save the SolutionSet output.
Iterate through the SolutionSet, and assign a VIS-task-based weight to each solution per VIS task. After this, given that we have $n$ VIS tasks at hand, we can generate $n$ composite costs per recommendation by summing the data-based cost with the VIS-task-based cost.
Clients of the library can further process these details and use them for representation

peter-gy / visrecly Goto Github PK

visrecly's Introduction

VisRecly

What Is This All About?

Decomposing the Project's Name and Ambitions

Tech Overview

Architectural Overview

Development

Starting the Project Locally

Running Tests

visrecly's People

Contributors

Stargazers

Watchers

Forkers

visrecly's Issues

Ranking Approach

Recommend Projects

Recommend Topics

Recommend Org