In my U.S. experience, every person asks me, "Oh, this is a little bit spicy, but you came from South Korea; I'm sure you are OK. " Sure, I like spicy food really much. However, we have yet to learn about every South Korean people's taste type. In Yummirific data, we collected 3,397 people's taste preference data with 40 survey questions. Fortunately, we covered that "SPICY." Our question number 8 aligns with "SPICY." "I like spicy food." So, I analyzed 3,397 people's responded data related to basic tasty preferences, Sour, Bitter, Salty, Sweety, and "SPICY. Here are some fun facts about it.
The problem with the existing restaurant recommendation system
01 - Impossibility of personal taste reflection
- The recommendation method is based only on its own information from restaurants.
- The assumption is that all customer's characteristics are universal and identical.
- However, each customer's taste type has a lot of diversity, and so many factors affect the preference of food selection.
- Most services focus on the internet's big data of restaurant information instead of analyzing customer taste types.
02 - Social conflict occurrence caused by services
- Harmful cases arose in Korea caused by malicious reviews or comments.
- Abnormal services occurred in Korea, which is fabricating and deleting reviews.
- Because of deletion reviews, social conflict cases have arisen between customers and restaurant owners.
03 - Limitation of NLP
- Natural language processing is ineffective in analyzing customer reviews to suit customers' intentions.
- Currently, most restaurant recommendation services' natural language processing technology is only at the level of keyword extraction.
How to solve problems?
Using Small Data: Personal Taste Data
Collecting Taste Type Data: Small-Data
- Collecting and observing each customer's taste data.
- Collecting each customer's taste data using a test like MBTI.
Classification
- Estimate customer's latent factors that affect each
- customer's taste type and selection of restaurants
- with collected data.
- Using statistic models to classify people with similar tastes.
Recommend from group
- You can get restaurant recommendations from groups of users with similar tastes.
Patent Information: Our Technology got KR and US Patents!
Korea: 10-1963609 (Granted)
Title of Invention
Determination apparatus and method of user's taste based on a latent variable.
Patentee
MABLIC Corp
Inventor
Yunsik Choung
Registration Date
2019. 3. 25
Google Patents
https://patents.google.com/patent/KR101963609B1/en
US: US10803349 (Granted)
Title of Invention
Apparatus and method for determining the taste of user based on latent variable
Patentee
MABLIC Corp
Inventor
Yunsik Choung
Publication Date
2020. 10. 13
Google Patents
https://patents.google.com/patent/US10803349B2/en
The key technology of Patent
Yummirific Prototype Service introduction
Extensibility of technology
01 - Make use of Small-Data
The customer's taste information can be used as characteristic data of the platform's participating stores.
02 - Cross Machine Learning
Both small data for taste-type tests and extensive data collected according to service usage can be utilized.
03 - Recommendation for Group
Using a personal taste type profile, It can make simultaneous recommendations to multiple customers who want to eat with it.
Applicability of technology
01 - Food delivery service
- It can be used as the most relevant market for this Patent can be used most diversely.
- In particular, while existing services recommend restaurants based on customer actions (search or marketing landing), we can recommend their meal preemptively before mealtime.
- A variety of customer taste-type data collected through patented technology can be provided to restaurants on the platform.
02 - Maps Service(Including Navigation)
- As the food delivery service also uses a map, the map service strongly recommends the restaurant directly.
- The advantage of this patent technology is that you can directly recommend a nearby restaurant based on your location and driving route.
03 - Restaurant Search Service
- Customers can use their taste type data to provide more accurate search results when searching for restaurants on search portal services such as the Naver in Korea.
04 - Messaging Service
- Using this patented technology, we can recommend restaurants to reflect the tastes of each person who wants to have a meal together.
- This approach can be used as an add-on in messaging services, typically meeting two or more people.
- When booking a date course for a couple, you can analyze their food taste types and recommend a restaurant that satisfies both.
Comments
Post a Comment