1. Overview (Project Background)

This project was promoted for the purpose of improving the accuracy and usability of search experience,
and further providing generative AI-based customer-customized recommendation experience (CX).

In the existing keyword search environment, the relevance between the search terms entered by users and actual products was not sufficiently reflected,
for example, even with clear keywords like 'milk', there was a problem where related products were not accurately exposed.

Additionally, beyond simple keyword input,
there were limitations in responding to complex natural language-based requests such as "Recommend a gift over 100,000 won for my parents"
with existing search and recommendation methods.

Accordingly, this project was promoted with the following directions as the center.

• Improving the accuracy of keyword search results

• Securing natural language-based search and recommendation capabilities

• Establishing a CX foundation that can be expanded to multimodal generative AI-based personalized recommendations in the future

2. Solution (Solution)

In this project, we simultaneously strengthened keyword search and natural language search,
and designed and implemented a structure for automatic recommendation message generation using generative AI based on this.

Key Solutions

• Advanced Keyword Search

  • Strengthening the relevance between search terms and actual product attributes

  • Improving so that products matching the intent are accurately exposed when specific keywords are entered

• Introduction of Natural Language Search and Recommendation Features

  • Understanding requests in natural language form,

  • Providing natural recommendation responses by synthesizing key information such as price, discount rate, and product name

• Automatic Recommendation Message Generation

  • Automatically generating recommendation phrases matching user requests using generative AI

  • Securing a structure that can be expanded to multimodal generative AI (text, image, and context-based recommendations) in the future

Through this,
the goal was to transition from 'search that finds products' to 'recommendations that understand and propose intent'.

3. Result (Results)

Through this project, we have secured the following achievements in the search and recommendation areas.

Improvement of Keyword Search Results

• When searching for clear keywords such as 'milk',
  improved to accurately expose milk-related products compared to before

Securing Natural Language Search Capabilities

• For complex natural language requests such as "Recommend a gift over 100,000 won for my parents"

  • Product name

  • Price

  • Discount rate
    Able to provide natural recommendation responses including key information

Establishing a Foundation for Enhanced Customer Experience (CX)

• Comprehensively strengthening keyword search and natural language search

• Through the automatic recommendation message generation function,
  securing the technical foundation for providing customer-customized recommendation experience

• Completing a structure that can be expanded to multimodal generative AI-based recommendation messages and personalized services in the future