Detailed description of provided data

Available in VPC

Cloud Data Box provides user behavior data such as NAVER search and shopping, image and annotation data for AI training, and object data for news articles and NLP analysis.
Search and shopping sample data are provided by default when you create a data box. When supplying the sample data, we upload it to Cloud Hadoop HDFS and mount the NAS on the TensorFlow Server as read-only. Request data supply after configuring the analysis environment. Search, shopping, and AI data by the default option (data from the first half of the previous year) will then be provided to you via Cloud Hadoop and NAS mounted on TensorFlow Server as read-only. If you subscribe to Insight Option, you can receive the latest data from the first half of the 2 years back to the last month. On the 15th of each month, you can also receive the data from the previous month.

This section describes the data provided by Cloud Data Box in detail.

Standard data and Insight Option data

Search data (search)

1. Search click data (search/click)

Item	Description
Data introduction	Data showing the keywords NAVER users searched, and the areas they clicked on (Targets search keywords searched by more than 100 users daily)
Data provision period	Standard Data: first half of the previous year (January to June), updated at the beginning of each year Insight Data: from January of 2 years back to the previous month of the current year, updated monthly (For example, on August 2021, subscribers would receive data from January 2019 to July 2021.) (unit: day)
Extraction targets	Logged-in NAVER users with a click history during the set period
Data aggregation criteria	date (base date): search/click date device: device information used by the user for search (mobile/pc) gender: user's gender (f/m) age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): metropolitan city/province based on the user's address loc2 (region 2): city/county/district based on the user's address keyword (search keyword): keyword the user entered in the NAVER integrated search area (converted to lowercase, whitespaces removed) area (click area): service area that the user clicked on in NAVER's integrated search results count (number of clicked users): aggregated by base date, device, gender, age group, region, search keyword, and click area
Sample data location	${DATABOX_HOME_DIR}/sample/search/click
Entire data location	For data from the first half of 2020: ${DATABOX_HOME_DIR}/search20y1h/search/click
Directory structure

2. Search click co-occurrence (search/click_cooccurrence)

Item	Description
Data introduction	Data showing the information NAVER users searched and clicked together in 1 day (Targets search keywords searched by more than 100 users weekly)
Data provision period	Standard Data: first half of the previous year (January to June), updated at the beginning of every year Insight Data: from January of 2 years back to the previous month of the current year, updated monthly (unit: day)
Extraction targets	Logged-in NAVER users with a click history during the set period
Data aggregation criteria	week (base date): search/click date device: device information used by the user for search (mobile/pc) gender: user's gender (f/m) age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): metropolitan city/province based on the user's address loc2 (region 2): city/county/district based on the user's address keyword1 (search keyword1): keyword the user entered in the NAVER integrated search area (converted to lowercase, whitespaces removed) area1 (click area1): service area that the user clicked on in NAVER's integrated search results on keyword1 keyword2 (search keyword2): keyword the user entered in the NAVER integrated search area (converted to lowercase, whitespaces removed) area2 (click area2): service area that the user clicked on in NAVER's integrated search results on keyword2 count (number of users that clicked the data on keywords together): counted by base date, device, gender, age group, region, keyword1, area1, keyword2, and area2 (The search order between keyword1 and keyword2 is not considered.)
Sample data location	${DATABOX_HOME_DIR}/sample/search/click_cooccurrence
Entire data location	For data from the first half of 2020: ${DATABOX_HOME_DIR}/search20y1h/search/click_cooccurrence
Directory structure

3. Search access location (search/click_location)

Item	Description
Data introduction	Data showing the keywords NAVER users searched for by region (As the access area is measured based on IP, there may be errors in accuracy)
Data provision period	Standard Data: first half of the previous year (January to June), updated at the beginning of every year Insight Data: from January of 2 years back to the previous month of the current year (updated monthly) (unit: day)
Extraction targets	Logged-in NAVER users with a click history during the set period
Data aggregation criteria	date (base date): search/click date time (time frame): uses 3-hour time windows (00-02/03-05/06-08/09-11/12-14/15-17/18-20/21-23) device: device information used by the user for search (mobile/pc) gender: user's gender (f/m) age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): access location (metropolitan city/province). Area tracked based on access IP loc2 (region 2): access location (city/county/district). Area tracked based on access IP keyword (search keyword): keyword the user entered in the NAVER integrated search area (converted to lowercase, whitespaces removed). Keywords searched by different users during the day (counted only if they clicked a link after the search) count (number of users that clicked links): aggregated by base date, time frame, device, gender, age group, region, and search keyword
Sample data location	${DATABOX_DIR}/sample/search/click_location
Entire data location	For data from the first half of 2020: ${DATABOX_DIR}/search20y1h/search/click_location
Directory structure

Shopping data (shopping)

1. Product click data (shopping/click)

Item	Description
Data introduction	Data showing the keywords NAVER users searched for and the product categories they clicked on (Targets search keywords/product categories with more than 100 clicks daily)
Data provision period	Standard Data: first half of the previous year (January to June), updated at the beginning of every year Insight Data: from January of 2 years back to the previous month of the current year, updated monthly (unit: day)
Extraction targets	Logged-in NAVER users with product click history during the set period
Data aggregation criteria	date (base date): search/click date device: device information used by the user for search (mobile/pc) gender: user's gender (f/m) age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): metropolitan city/province based on the user's address loc2 (region 2): city/county/district based on the user's address keyword (entry keyword): keyword that the user searched for before clicking on the product cat (product category): section/division/group/class of the product clicked by the user count (number of visitors): counted by date, device, gender, age group, region, product category, and search keyword brand (brand name): presumed name of the brand that the product the user clicked belongs to item (product name): presumed name of the product the user clicked
Sample data location	${DATABOX_HOME_DIR}/sample/shopping/click
Entire data location	For data from the first half of 2020: ${DATABOX_HOME_DIR}/shopping20y1h/shopping/click
Directory structure

2. Product purchase data (shopping/purchase)

Item	Description
Data introduction	Data showing the products logged-in NAVER users purchased (Targets product categories with 10 or more purchases daily)
Data provision period	Standard Data: first half of the previous year (January to June), updated at the beginning of every year Insight Data: from January of 2 years back to the previous month of the current year, updated monthly (unit: day)
Extraction targets	NAVER users with a purchase history during the set period
Data aggregation criteria	date (base date): purchase date device: device information used by the user for search (mobile/pc) gender: user's gender (f/m) age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): metropolitan city/province based on the user's address loc2 (region 2): city/county/district based on the user's address cat (product category): section/division/group/class of the product purchased by the user count (number of buyers): counted by date, device, gender, age group, region, search keyword, and product category brand (brand name): presumed name of the brand that the product the user purchased belongs to item (product name): presumed name of the product the user purchased
Sample data location	${DATABOX_HOME_DIR}/sample/shopping/purchase
Entire data location	For data from the first half of 2020: ${DATABOX_HOME_DIR}/shopping20y1h/shopping/purchase
Directory structure

3. Product click co-occurrence (shopping/click_cooccurrence)

Item	Description
Data introduction	Data showing the products NAVER users searched for and clicked on together during the day (Targets search keywords/product categories with more than 100 clicks daily)
Data provision period	Standard Data: first half of the previous year (January to June), updated at the beginning of every year Insight Data: from January of 2 years back to the previous month of the current year, updated monthly (unit: day)
Extraction targets	Logged-in NAVER users with product click history during the set period
Data aggregation criteria	week (base date): the starting date of the extraction period device: device information used by the user for search (mobile/pc) gender: user's gender (f/m) age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): metropolitan city/province based on the user's address loc2 (region 2): city/county/district based on the user's address keyword1 (entry keyword 1): keyword that the user searched for before clicking on product category 1 cat1 (product category 1): section/division/group/class of the product clicked by the user keyword2 (entry keyword 2): keyword that the user searched for before clicking on product category 2 cat2 (product category 2): section/division/group/class of the product clicked by the user count (number of users that visited web pages on keywords together): counted by base date (weekly), device, gender, age group, region, keyword1, cat1, keyword2, and cat2 (The click order is not considered.)
Sample data location	${DATABOX_HOME_DIR}/sample/shopping/click_cooccurrence
Entire data location	For data from the first half of 2020: ${DATABOX_HOME_DIR}/shopping20y1h/shopping/click_cooccurrence
Directory structure

4. Product purchase co-occurrence (shopping/purchase_cooccurrence)

Item	Description
Data introduction	Data showing the products NAVER users purchased together daily (Targets product categories with 10 or more purchases daily)
Data provision period	Standard Data: first half of the previous year (January to June), updated at the beginning of every year Insight Data: from January of 2 years back to the previous month of the current year, updated monthly (unit: day)
Extraction targets	NAVER users with a purchase history during the set period
Data aggregation criteria	week (base date): the starting date of the extraction period device: device information used by the user for search (mobile/pc) gender: user's gender (f/m) age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): metropolitan city/province based on the user's address loc2 (region 2): city/county/district based on the user's address cat1 (product category 1): section/division/group/class of the product the user purchased cat2 (product category 2): section/division/group/class of the product the user purchased count (number of users that purchased products together): counted by base date (weekly), device, gender, age group, region, cat1, and cat2
Sample data location	${DATABOX_HOME_DIR}/sample/shopping/purchase_cooccurrence
Entire data location	For data from the first half of 2020: {DATABOX_HOME_DIR}/shopping20y1h/shopping/purchase_cooccurrence
Directory structure

AI data (ai)

1. Reused image (ai/clova/iitp_waste_images)

Item	Description
Data introduction	Collect and label recycled images
Data use examples	Developing technology that can be used to solve image classification problems using industrial product data in a real-life environment
Data details	Data format: 3,000 images in JPEG format (1280720, 7201280), iitp_waste_images_3000_result.csv Labeling information 1: General waste (others) 2: Paper 3: Cans and scrap metal 4: Glass bottles 5: Plastics (including PET) 6: Vinyl 7: Styrofoam 8: Food
Data sample (examples)
Entire data location	${DATABOX_HOME_DIR}/ai/clova/iitp_waste_images

2. Food image (ai/clova/naver_food_fixed)

Item	Description
Data introduction	Data that tags food areas in the image with a bounding box
Data location	${DATABOX_HOME_DIR}/ai/clova/naver_food_fixed
Data use examples	Train an AI based on the necessary elements extracted from images to solve problems
Data details	Collected image Number of data: 2,042 images Data format: JPEG image and its JSON file
Data sample (examples)
Entire data location	${DATABOX_HOME_DIR}/ai/clova/naver_food_fixed

3. Restaurant image (ai/clova/externalImageOCR)

Item	Description
Data introduction	Korean text OCR annotation data (including Roman alphabets and numbers) in images of posts, billboards, menus, and restaurant signs
Data use examples	Develop a technology that can extract text from images and convert it to digital data
Data details	Korean, Roman alphabet, and numeric annotation data in collected images of signs Number of data: 1,180 images (signboard 197, restaurant sign 324, menu 614, standing_signboard 45) Data format: JSON, JPEG (original image, result image)
Data sample (JSON examples)
Entire data location	${DATABOX_HOME_DIR}/ai/clova/externalImageOCR

4. News data for NLP experiments (ai/nlp)

Item	Description
Data introduction	Data of an object name (an entity's name) in a news article collected from NAVER News Service linked to the Wikipedia page related to that object
Data aggregation criteria	Consists of news article's title, body, and category. The locations of the object names in the body text are indicated by BIO tags, and IDs are attached to the names.
Data use examples	A technology that links object names (name) to information related to the objects (Wikipedia page corresponding to the names) from text can be developed (entity linking)
Data sample (examples)
Entire data location	${DATABOX_HOME_DIR}/ai/nlp

Insight Pro Option data

NAVER Insight Pro Option data (pro#1: ID by subgroup)

1. Search click data (search/click)

Item	Description
Data introduction	Data showing the keywords NAVER users searched for and the areas they clicked on Not individual-level data but anonymized group-level data (Users were grouped by gender/age group/region.) (Targets search keywords searched by more than 100 users daily)
Data provision period	January of 2 years back to the previous week of the current year, updated weekly (For example, on February 16, 2022, subscribers would receive data from January 2020 to January 2022.) (unit: day)
Extraction targets	Logged-in NAVER users
Data aggregation criteria	date (base date): search/click date time (time frame): uses 3-hour time windows (00-02/03-05/06-08/09-11/12-14/15-17/18-20/21-23) device: device information used by the user for search (mobile/pc) gender: user's gender (f/m) user (user group ID): anonymized user group ID age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): metropolitan city/province based on the user's address loc2 (region 2): city/county/district based on the user's address keyword (search keyword): keyword the user entered in the NAVER integrated search area (converted to lowercase, whitespaces removed) area (click area): service area that the user clicked on in NAVER's integrated search results count (number of clicks): aggregated by base date, time frame, device, gender, user group ID, age group, region, search keyword, and click area
Sample data location	${DATABOX_HOME_DIR}/sample/pro_search/click
Directory structure

2. Product click data (shopping/click)

Item	Description
Data introduction	Data showing the keywords NAVER users searched for and the product categories and products they clicked on Not individual-level data but anonymized group-level data (Users were grouped by gender/age group/region.) (Targets search keywords/product categories with more than 100 clicks daily)
Data provision period	January of 2 years back to the previous week of the current year, updated weekly (For example, on February 16, 2022, subscribers would receive data from January 2020 to January 2022.) (unit: day)
Extraction targets	Logged-in NAVER users with product click history during the set period
Data aggregation criteria	date (base date): search/click date time (time frame): uses 3-hour time windows (00-02/03-05/06-08/09-11/12-14/15-17/18-20/21-23) device: device information used by the user for search (mobile/pc) gender: user's gender (f/m) user (user group ID): anonymized user group ID age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): metropolitan city/province based on the user's address loc2 (region 2): city/county/district based on the user's address keyword (entry keyword): keyword that the user searched for before clicking on the product cat (product category): section/division/group/class of the product clicked by the user count (number of clicks): aggregated by base date, time frame, device, gender, user group ID, age group, region, search keyword, and product category brand (brand name): presumed name of the brand that the product the user clicked belongs to item (product name): presumed name of the product the user clicked
Sample data location	${DATABOX_HOME_DIR}/sample/pro_shopping/click
Entire data location	For data from the first half of 2020: ${DATABOX_HOME_DIR}/pro20y1h/shopping/click
Directory structure

3. Product purchase data (shopping/purchase)

Item	Description
Data introduction	Data showing the products logged-in NAVER users purchased Not individual-level data but anonymized group-level data (Users were grouped by gender/age group/region.) (Targets product categories with 10 or more purchases daily)
Data provision period	January of 2 years back to the previous month of the current year, updated monthly (For example, on February 16, 2022, subscribers would receive data from January 2020 to January 2022.) (unit: day)
Extraction targets	NAVER users with a purchase history during the set period
Data aggregation criteria	date (base date): purchase date time (time frame): uses 3-hour time window (00-02/03-05/06-08/09-11/12-14/15-17/18-20/21-23) device: device information used by the user to make the purchase (mobile/pc) gender: user's gender (f/m) user (user group ID): anonymized user group ID age (age range): uses 5-year age groups (-12/13-18/19-24/25-29/30-34/35-39/40-44/45-49/50-54/55-59/60-64/65-69/70-) loc1 (region 1): metropolitan city/province based on the user's address loc2 (region 2): city/county/district based on the user's address cat (product category): section/division/group/class of the product purchased by the user count (number of purchases): counted by date, time frame, device, gender, age group, user group ID, region, search keyword, and product category brand (brand name): presumed name of the brand that the product the user purchased belongs to item (product name): presumed name of the product the user purchased
Sample data location	${DATABOX_HOME_DIR}/sample/pro_shopping/purchase
Entire data location	For data from the first half of 2020: ${DATABOX_HOME_DIR}/pro20y1h/shopping/purchase
Directory structure

NAVER Insight Pro data (pro#2: combination of NAVER's data and NICE Information Service's data)

1. Combination of NAVER's data and NICE Information Service's data