Ava: From Data to Insights Through Conversation Rogers Jeffrey Leo - - PowerPoint PPT Presentation

▶
ava from data to insights through conversation
SMART_READER_LITE
LIVE PREVIEW

Ava: From Data to Insights Through Conversation Rogers Jeffrey Leo - - PowerPoint PPT Presentation

Sep 03, 2022 93 likes •233 views

Ava: From Data to Insights Through Conversation Rogers Jeffrey Leo John, Navneet Potti, and Jignesh M. Patel University of Wisconsin Madison Rogers Navneet 01/10/2017 University of WisconsinMadison 1 Motivation Why do my customers

slide-1
SLIDE 1

Ava: From Data to Insights Through Conversation

Rogers Jeffrey Leo John, Navneet Potti, and Jignesh M. Patel University of Wisconsin – Madison

01/10/2017 University of Wisconsin–Madison 1

Rogers Navneet

slide-2
SLIDE 2

Motivation

01/10/2017 University of Wisconsin–Madison 2

A legal pyramid scheme (org chart)

Why do my customers churn?

Data Scientist VP

Focus for this talk

slide-3
SLIDE 3

01/10/2017 University of Wisconsin–Madison 3

Issues

  • 1. Lost in translation
  • 2. Long turn-around time
  • 3. Correctness!
  • 4. Reproducibility

…

slide-4
SLIDE 4

Data Science Pipeline

Data Loading

  • from a csv file

Data Cleaning

  • fill missing values

Feature Engineering

  • pick/create appropriate features

Model Selection and Training

  • pick an ML model based on the input and task at hand

Parameter Tuning

  • hyperparameter optimization

Save model for deployment

  • As a UDF, PMML file, …

01/10/2017 University of Wisconsin–Madison 4

slide-5
SLIDE 5

Insights

  • Often the task is …

01/10/2017 University of Wisconsin–Madison 5

Conversation Composition Code

Constrained Natural Language Storyboard In an interactive notebook (e.g. iPython)

Rely on Natural Language Translation rather than Natural Language Understanding Target the lower level of the pyramid and allow the programmer to take control Composable internal architecture and convenient system abstraction

slide-6
SLIDE 6

01/10/2017 University of Wisconsin–Madison 6

slide-7
SLIDE 7

Architecture

slide-8
SLIDE 8

The Ava Storyboard Concept …

01/10/2017 University of Wisconsin–Madison 8

slide-9
SLIDE 9

The right encapsulation for:

  • 1. Technology trend
  • NLT v/s NLU
  • Different ML backends
  • 2. Developer skills
  • Business analysts vs

statistician

What is new?

Previous work

  • Imitation Game: Turning 1950
  • CNL à remove ambiguity
  • Natural Language à Query:

Use feedback to refine ambiguity

Ava

  • Constrain to walk along a pre-

scripted storyboard

  • Storyboard is a Finite State

Machine

  • The user creates their “own” story.

It always ends well J.

01/10/2017 University of Wisconsin–Madison 9

Ci Cj

chatij Controlled Natural Language

slide-10
SLIDE 10

1/12/17 University of Wisconsin–Madison 10

Python Ava 10 20 30 40 50 60

Time (minutes)

Results from a user study with 16 participants. Distribution

  • f the time taken by participants to complete the first model.
slide-11
SLIDE 11

Summary of

The data chatbots are coming Ava democratizes “data science” Benefits: increased human productivity, reproducibility, rapid exploration, a powerful collaboration mechanism, … More automation along every dimension

01/10/2017 University of Wisconsin–Madison 11

The conversation is the code AVA

slide-12
SLIDE 12

Thanks!

12 01/10/2017 University of Wisconsin–Madison

Jean-Michel Ané, Victor Cabrera, Mark Craven, Pamela Herd, David Page, Federico E. Rey, Garret Suen, and Kent Weigel

University of Wisconsin