- 개인화된 추천 (Personalized Recommendation)
- 새로운 와인 발견 (Discover New Wines)
- 시간절약 (Time-Saving)
사진을 클릭하세요.
-
- API 호출하여 JSON 파일 추출
- Columns 총 22 columns x 3675 rows
- wine_ratings_count : 와인 평점 수
- wine_ratings_average : 와인 평점
- foods : 와인에 어울리는 음식
- vintage_price : 가격
- vintage_wine_name : 와인 이름 (이름 + 년도)
- vintage_wine_type : 와인 타입 (id mapping to type)
- varietal_name : 포도 품종
- variety : 와인 품종
- winery : 와이너리 (양조장)
- Row Count
- 중복값 제거 후 df.shape = (2024, 22)
-
- Columns
- description : 리뷰
- title : 와인 이름
- country : 나라 이름
- points : 와인 평점
- price : 와인 가격
- variety : 와인 품종
- winery : 와이너리 (양조장)
- Row Count
- 중복된 행 제거 후 29,971 ⇨ 119,928
- Description words
- 68 words
- 단어 벡터화 제외 5% 이하, 90% 이상
- Columns
-
와린이ver.
와린이는 와인 + 어린이의 합성어로 와인에 대해 모르는 사람입니다. 와린이는 와인의 탄닌이 뭔지, 산도가 뭔지 잘 모르기 때문에 선택하기 보다 쉬운 질문을 해서 와인을 추천해 주는 시스템 입니다.
[예시] Step 1. 더 선호하시는 것을 고르세요.
- 고기 2. 생선 3. 채소
예시로 Step 1을 보면, 각 선택지 마다 와인의 특징을 부여하였고 모델 데이터에서 미리 뽑은 68개의 단어들 중에서 특징들과 어울리는 단어들을 부여했습니다.
-
고기 : 진한 레드와인, 소고기, 양고기, 스테이크, 육즙이 풍부한, 풀바디 레드와인, 탄닌이 풍부, 강한 풍미 => [dry, firm, tannins, red, flavor, dark, cabernet, sauvignon]
-
생선 : 화이트 와인, 산뜻하고 깔끔한 산미, 신선함, 해산물 요리 => [fruit, white, crisp, fresh, bright, touch, sauvignon, light]
-
채소 : 화이트, 라이트레드, 구운 채소, 샐러드, 신선한, 상큼함, 섬세한 => [fruit, white, crisp, fresh, bright, touch, sauvignon, light]
👇와인 추천 시스템을 이용하려면 이걸 눌러주세요.
👇와인 추천 시스템 구현영상을 보시려면 이걸 눌러주세요.
- Personalized recommendations
- Discover new wines
- Save time
Click the image.
[About] A true wine lover and foodie. Can enjoy up to three bottles with great food. Idea contributor.
[About] Used to know only “sparkling / red / white,” but now on the way to becoming a wine knower!
-
- Called the API and extracted JSON
- Columns: 22 columns × 3,675 rows
- wine_ratings_count: number of ratings
- wine_ratings_average: average rating
- foods: food pairings
- vintage_price: price
- vintage_wine_name: wine name (name + vintage year)
- vintage_wine_type: wine type (id mapped to type)
- varietal_name: grape varietal
- variety: wine variety
- winery: winery
- Row Count
- After removing duplicates:
df.shape = (2024, 22)
- After removing duplicates:
-
- Columns
- description: review text
- title: wine name
- country: country
- points: rating
- price: price
- variety: variety
- winery: winery
- Row Count
- After removing duplicates: 29,971 ⇨ 119,928
- Description words
- 68 words used
- Words below 5% or above 90% frequency were excluded from vectorization
- Columns
(Images unchanged)
-
Wine-savvy ver.
(see screenshot) -
Beginner (“Warin-i”) ver.
“Warin-i” is a portmanteau of “wine + beginner.” Since beginners may not know terms like tannin or acidity, the system asks simple preference questions and recommends wines accordingly.
[Example] Step 1. Choose what you prefer.
- Meat 2) Fish 3) Vegetables
For each choice, we assign characteristic tags and map them to a subset of our 68 keywords extracted from the model data.
-
Meat: rich red wines, beef/lamb/steak, juicy, full-bodied, tannic, bold flavors
⇒ [dry, firm, tannins, red, flavor, dark, cabernet, sauvignon] -
Fish: white wines, crisp acidity, freshness, seafood
⇒ [fruit, white, crisp, fresh, bright, touch, sauvignon, light] -
Vegetables: white / light red, roasted vegetables, salads, fresh, zesty, delicate
⇒ [fruit, white, crisp, fresh, bright, touch, sauvignon, light]
(See screenshots)
-
Food-savvy (“Mat-jal-al”) ver.
(see screenshot) -
Recommendation Results
- Best 3 (screenshot)
- Worst 3 (screenshot)
👇 Click to use the wine recommendation system.
👇 Click to watch the implementation demo.










