Text-based user trend analysis system

Text-based user trend analysis system introduction

It is a system that collects and processes text data such as user reviews and comments about the game and provides it as indicators. It provides analyzed indicators to understand user trends and issues in the game.

Location in console

• Hive console (by company) > AI Services > Text-based user trend analysis system

Sub menu

• Dashboard (Regular user permission required)
  • Dashboard
  • Interested/excluded keywords
  • Personal settings
• Content management (Administrator permission required)
  • Dashboard
  • Register game
  • Interested/excluded keywords
  • Data collection management
  • Backoffice log
  • Personal settings

Prerequisites

Request console permission

• Dashboard menu access (Regular user permission)
  • Apply for permission to the console > AI Services > Text-based user trend analysis system > Dashboard menu to the AI Services permission management department for each company to obtain permission to use the menu.
• Access to the content management menu (Administrator permission)
  • GCP employees do not need to apply separately.
  • If you are not a GCP employee, you must additionally apply for GCP company access permission.
• Please Note
  • Permission request and management policy for each company/server (Korean)

Market Reviews Analysis

Dashboard

• This is a screen where you can check summary indicators for the collected data which are accessible from the dashboard or content management menu.
• It provides summary and language-specific indicators based on search criteria.
  1. Select search criteria

• 1. • Select date
       • The default setting is from 2 to 15 days before the current date.
     • Select game
       • You can select one of the registered games using the combo box.
       • Summoners War: Sky Arena is selected by default.
         • We plan to improve it so that you can set the default value individually in the future.
       • If you have access to the content management menu, you can add other games from the Register game tab.
     • Select top N
       • You can select the number of top keywords in units of 10 out of 10-50 using the combo box.
       • 20 is selected by default.
  2. Graph
     • It consists of a summary tab and a tab for each language if two or more languages ​​and two or more countries are specified when registering a game.
       • It consists of a tab of that language if only one language and country is specified.
     • Summary indicators
       
       
     • Indicators by language
       
       
     • Collection status
       • You can check the number of reviews by date.
       • You can check the percentage of market reviews.
     • Keyword
       • You can check the number of the following items for the selected top N keywords/interested keywords/excluded keywords in a graph.
         • You can check the number of appearances by keyword.
         • You can check the ratio of the number of appearances of keywords.
         • You can check the average rating of the reviews that appeared in the keyword.
         • You can check the composition ratio of keywords for all words that appeared in the search period.
         • You can check the number of appearances of keywords by date.
     • Rating
       • You can check the average rating by date.
       • You can check the percentage of each rating from the total ratings registered during the search period.

Sentiment Analysis

• Introduction
  • It is a system that uses sentiment analysis, a natural language processing approach, on user review data of games to assess subjective data, such as user trends and opinions.
  • After gathering the data, we analyze it using an internal universal sentiment analysis model to determine how positive or negative the review was, provide processed indicators, and share the original review text.
• Summary of provided indicators
  • Daily Sentiment Ratio: Daily sentiment ratio and count of the set date
  • Total Sentiment Ratio: Sentiment ratio of all dates
  • No. of Positive/Negative Responses by Top Keyword: Number of positives/negatives of the original text containing the top keywords
  • Positive/Negative Response Ratio for Top Keywords: Ratio of positives/negatives of the original text containing the top keywords
  • Positive/Negative Keywords: Provides keywords with colors based on sentiment scores
    • Word Cloud: Positive/negative keyword cloud
    • Word Table: Positive/negative keyword table
    • Selected keyword: Detailed analysis of keywords selected from word cloud or word table
      • Sentiment score distribution by rating
      • Sentiment original reviews
  • Sentiment score by rating: Indicators that provide ratings and sentiment scores
    • Original review by rating/score
• Detailed indicator
  • Daily/Total Sentiment Ratio
    
    
    • The daily sentiment percentage bar graph displays your app’s daily positive/negative sentiments.
    • The overall sentiment ratio pie graph displays the positive/negative ratio for a specific date.
      • The outer circle represents the overall positive/negative ratio, while the inner circle represents the top keywords’ positive/negative ratio.
  • Positive/Negative Response Ratio for Top Keyword
    
    • The number and percentage of positives and negatives for the top N most frequently written keywords are displayed.
  • Positive/Negative Keywords
    
    
    • Provides a word cloud and color-coded table for the top N keywords based on their positive/negative score.
    • The positive/negative score is organized from 0 to 1, with 0.4 to 0.6 representing neutral. (Negative: 0–0.4; positive: 0.6–1.0)
    • Details appear in the sub-selected keywords panel when you click on a specific keyword in the keyword map or keyword table.
      • It displays the percentage and number of positives and negatives by rating, and the source text for the positives and negatives can be viewed.
        • In the full text, the keywords are colored blue for positive and red for negative.
        • If the source text is lengthy, you can view the entire text by clicking the Read More button.
  • Sentiment score by rating
    
    
    • App review ratings and positive/negative scores are positively correlated after analyzing ratings and positive/negative scores. You can use this to find anomalies and view the full text.
    • To see the full text by rating/score in the right panel, click on the red and blue circles in the positive/negative score by rating indicator.
      • The image above depicts the outcome of clicking on a circle with a high rating of 5, but a positive/negative score close to zero.
• The model employed in sentiment analysis
  • It’s a deep learning-based sentiment analysis model trained on public and private data. (Please contact the AI development team for more information.)
  • The feature will be improved constantly by improving educational data.

Interested/excluded keywords

• It provides the ability to register or delete interested or excluded keywords.
  1. Register/delete interested keywords
     
     

• 1. • Register interested keyword: After entering the interested keywords you want to register, click the Register button to register, and you can check the number of the registered keywords through a graph on the dashboard.
     • Delete interested keywords: You can delete a keyword from the list of keywords you want to delete by clicking the ‘X’ button.
  2. Register/delete excluded keywords
     
     

• 1. • Register excluded keywords: If you click the Register button after entering the excluded keyword you want to register, the registered keyword is excluded from the top N keywords on the dashboard.
       • A graph for excluded keywords is available in a separate tab.
     • Delete excluded keywords: You can delete excluded keywords by clicking the ‘X’ button in the row of excluded keywords you want to delete from the list.

Register game

• It provides the ability to register games for which you want to collect reviews.
• Content management menu access permission is required.
  1. You can check the list of registered games on the left side of the screen.
     
     

• 1. • A new game registration screen opens when you click to register a new game.
     • Select company: You can check the games of the selected company among registered games.
     • Search game: You can check the registered games in the game list through keyword search.
     • a/g/i after the game name means the registered market.
       • a: Amazon App Store
       • g: Google-play
       • i: Apple App Store
  2. New game registration screen
     
     
     • Click the new game button to show an input form.
       • Fill out all the details and click the Save button to save.
       • If you want to collect English reviews, you must select both English and US for language and country respectively.
     • Refer to the following when filling out the form.
       • Game ID: If you have an App ID in the Hive App Center, fill in a part of the App ID in the form of com.company.gameid. Otherwise, make it as similar as possible.
         • Ex) com.com2us.smon
       • data.ai App ID: Refer to the following on the data.ai website to check the App ID for each market and fill in it.
         • data.ai’s App ID is located at the bottom left corner of each app details page or in the URL.
           
       • Game company:  If there is no company name to select, select other and enter the company name in English in the new input box that appears.
         • e.g. ncsoft
  3. Check and edit/delete registered game information
     
     

• 1. • You can view registration information by selecting a game from the game list.
     • After editing the contents, click the Save button to save the modified contents.
     • Click the Delete button to delete the game.
       • Deleted game information and data are internally inactive.

Data Collection Management

• It provides data collection status and the function to request data collection by date and game for the data that has not yet been collected.
• Content management menu access permission is required.
  1. Check for a limited number of API calls per day.
     
     

• 1. • According to data.ai, you can call up to 1,000 calls per day for each game and by date.
  2. You can select search criteria for date, company and game.
     
     
     • The date selection default range is 2 to 15 days before the current date.
     • The default value of company selection is All.
     • The default value of game selection is All.
  3. It shows the current status of data on the number of reviews collected by game and date.
     • Status
       • Distinguish by summary and tab for each language
       • Can check the supported languages for each game in the summary tab
       • Check data collection status by date/game
       • Collection Status
         
     • Request
       • After selecting the checkbox of the game/date that has no collection history, click the process request button to register it in the real-time request status.
         • You can select multiple checkboxes at the same time by selecting by row (date) or column (game).
       • You can check the collection status in the real-time request status, which is updated every 10 seconds.
     • Delete
       • After selecting the checkbox of the game/date that has collection history, click the delete request button to delete collected data.

Personal settings

• A user can set up the initial setting value when entering the dashboard.
  • Select game: Selects a game to view the history of reviews.
  • Top N pick: Selects the number of top keywords.
    
    

System related inquiries

• AI development team(ld@com2us.com)