当前位置：首页 > news >正文

基于大语言模型的推荐系统（2）

news 来源：原创 2025/6/19 5:36:50

本博文继续研究如何利用大语言模型（LLM）来构建播客推荐系统。

使用mongoDB 数据看存储基本数据

播客表（Podcast）
节目表（Episodes)
用户表（User）

用户表(User)

DeepSeek 笔记：推荐的spotify listener 记录格式

1. Basic Account Information

User ID: A unique identifier for the user (e.g., a Spotify-generated UUID).
Username: The display name chosen by the user.
Email Address: The email associated with the account.
Password: Encrypted for security.
Account Type: Free (ad-supported) or Premium (paid subscription).
Country/Region: The user's location, which may affect available content.
Date of Birth: Used for age verification and personalized content.
Account Creation Date: When the user signed up for Spotify.

2. Subscription and Payment Details

Subscription Status: Active, canceled, or trial.
Payment Method: Credit card, PayPal, etc.
Billing History: Records of past payments.
Renewal Date: For Premium users.

3. Usage and Activity Data

Listening History: Tracks, albums, and playlists the user has played.
Recently Played: A list of recently listened-to songs.
Playlists Created: User-generated playlists.
Followed Artists/Users: Artists or other users the user follows.
Liked Songs: Songs saved to the "Liked Songs" library.
Podcasts Subscribed To: Podcasts the user follows.
Device Information: Devices used to access Spotify (e.g., mobile, desktop, smart speaker).

4. Preferences and Settings

Language Preference: The user's chosen language for the app.
Privacy Settings: Whether the user's activity is public or private.
Audio Quality Settings: Streaming quality (e.g., low, normal, high, very high).
Explicit Content Filter: Whether explicit content is allowed.
Social Sharing Settings: Whether the user allows sharing activity on social media.

5. Analytics and Recommendations

Personalized Recommendations: Generated based on listening habits (e.g., Discover Weekly, Daily Mixes).
Top Tracks/Artists: Lists of the user's most-played songs and artists.
Listening Trends: Data on when and how often the user listens to music.

6. Security and Privacy

wo-Factor Authentication (2FA): Whether enabled.
Login History: Records of recent logins and devices used.
Data Sharing Preferences: Whether the user allows Spotify to share data with third parties.

从上面的信息中，截取一部分重要的部分，构建一个用户记录。

用户

UserSchema={
        UserID:String,
        Username:String,
        Email_Address:String,
        Password:String,
        CountryRegion:String,
        Date_of_Birth:String,
        Language:String,
       Account_Creation_Date:String,
}

收听历史(UserListenHistory)

UserListenHistorySchema={
        episodes_id:String,
        listen_time:String,//收听时间
        completion_rate:Numeric, //收听完成率（百分比）

}

关注的播客(UserFllowingPodcast）

UserFllowingPodcastSchema={
Podcast_ID:String,
Fllowing_time:String,//关注的时间

}

History ，Follow，Like 的列表可以数组的方式存储在听众表中

UserSchema={
        UserID:String,
        Username:String,
        Email_Address:String,
        Password:String,
        CountryRegion:String,
        Date_of_Birth:String,
        Language:String,
        Account_Creation_Date:String,
        History：HistorySchema，
        Follows：followsSchema，
        Likes：likesSchema
}

播客表（Podcast）

podcastSchema={
              podcast:String，
              uuid:String，
              title:String,
              image:String,
              description:String
              language:String
              categories:String,
              website:String,
              itunes_id:String，
              follows:Numer               
              }

节目表（Epicodes）

   epicodeSchema={
   audio:String,
   audio_length:String,
   description:String,
   pub_date:String,
   uuid:String,
   podcast_uuid:String,
   likes:number
}