Artificial Intelligence techniques that could be applied to provide a better search experience.

1. Chulho Jo
August 25, 2017
http://AutoParts911.com

2. Overview
• What is Artificial Intelligence (AI)?
• AI Techniques for Auto Parts Data
• AI Techniques for e-Commerce Solutions
• Current Status and Timeline

3. Artificial Intelligence (AI)?
• AI is the study of how to make software
algorithms do things better than human beings do
in real-time.
• AI includes Machine Learning, Natural Language
Processing, Neural Networks, Expert Systems,
and Big Data.
• AI algorithms are becoming more practical and
feasible in the Auto Parts industry due to
innovative hardware improvement.

4. AI Techniques for Auto Parts Data
• Natural Language Processing
o Create Automata using Lexical Analyzer & CFG (Context-Free Grammar) Parser by defining
a variety of vehicle information patterns in the product or fitment description.
Ex) Incomplete or arbitrary vehicle descriptions are parsed with specific Vehicle Id.
2017 Toyota Camry LE 2.4L
2005-2017 Camry LE
Camry LE 2005-17

Vehicle Id

5. AI Techniques for Auto Parts Data
o Implementation
By defining a proto-type model with 3 makes (Hyundai, Toyota,
Honda) using Lex/Yacc, the vehicle automata successfully returns
Vehicle Id for arbitrary
vehicle descirptions.
o Technical Consideration
• Generating Automata
from ACES VCDB
• Adding Categories and Brands
into Automata Parser
o Goal
Ex) Camry 2017 2.4L K&N Cabin Air Filter

2017 Toyota Camry LE / 2.4L
(Vehicle Id, Engine Id)
Accessories and fluids/Filters/Air Filter
(Category, Subcategory, Terminology)
K&N
(Brand)

6. AI Techniques for Auto Parts Data
• Pattern Analysis & Statistical Reasoning
o Field Mapping Control – Loading data in CSV or Text into Database Table
by selecting the input columns or adding in-line PHP codes.
o SKU Pattern Analysis
Click #1 (the first row) to see what SKUs are there.

7. AI Techniques for Auto Parts Data
• Pattern Analysis & Statistical Reasoning
o Automatic Brand Mapper Using Statistical Data
3rd Party data brands matched against ACES (Our Side) brands
(In this test, 70% of brands automatically detected)

8. AI Techniques for Auto Parts e-Commerce
• Natural Language Processing in Search Text or Chatting Robot
o Create Chatting Parser generated from Automata using Lex/Yacc
o Be integrated with the Vehicle/Category/Brand automata mentioned previously)
o Implementation
Proto-type Syntax Analyzer (Yacc)
worked.
Real dialog parser in progress.
o Goal
Less User Input  More Relevant Search Results

9. AI Techniques for Auto Parts e-Commerce
• Shopping Guide Expert Systems
o Visitor Tracking (Unique Cookie Id or User Id if logged in)
o Recording all Visitor’s behaviors (Navigation, Search Keywords, Selection) into knowledge database
(Knowledge Acquisition)
o Adding more data (Ex, Sales history if exists) into knowledge database
o Running Inference Engine (Reasoning with the knowledge)
Ex) Fact data AB, BC Inference data A C (Knowledge Inference)
o Implementation
Proto-type model in progress
o Goal
Customized User Experience with Customer Knowledge database
Search UI & Results are dynamically changed based on customer
Ex) If customer A has purchased Front Brake Pads a month ago and is now browsing in the Brake
category
 Rear Brake Pad should be displayed with higher priority (Assuming Rear Brake Pads worn out)
 For the same data, the systems respond differently based on customer knowledge (Like human
beings).

10. Current Status and Timeline
• Current Status
Between Proto-Type and Coding
• Timeline
Proto-
Type
Coding
Integra
tion
Launch