Personalized Article Recommendation

Personalized Article Recommendation

What are Recommender Systems?

Recommender systems are a subset of decision support systems that can analyze past behaviors and provide recommendations for current issues. Simply put, in recommender systems, we attempt to guess the user's thinking pattern using information about their behavior, the behavior of similar users, and their opinions, to suggest the most suitable and closest item to the user's taste. These systems essentially simulate and automatically execute the same process we use in our daily lives.

Outcomes of Recommendations

Recommendations provided by recommender systems generally have two outcomes:

1. Assisting the User in Decision Making: For example, suggesting which of several available options is better and should be chosen

2. Increasing User Awareness: Increasing the user's awareness in their field of interest. For example, when providing recommendations, the user becomes familiar with new items they previously did not know about.

Considerations for Recommender Systems

When users face a massive amount of information and items, the recommender system must provide recommendations considering:

• The context and environment the user is in
• The user's need
• The system's knowledge about the user
• The user's activity history

Methods and Algorithms

Recommender systems use various methods and algorithms:

1. Collaborative Filtering
2. Content-Based Filtering
3. Social Network-Based Filtering
4. Knowledge-Based Filtering

Recommender Systems in Noor Specialized Journals Database (Noormags)

Noormags uses three main methods to recommend related and personalized articles to its users:

1. Collaborative Filtering Algorithm ("Others' Suggestions")

By clicking on "Others' Suggestions", a list of articles will be recommended to the user. This option works based on the collaborative filtering algorithm, with recommendations based on other users' views.

Process:
1. First, users who have viewed this article are identified
2. Then, articles that those users have viewed are recommended to the target user
3. But that the target user has not yet viewed

2. Content-Based Filtering Algorithm ("Related Articles")

By clicking on "Related Articles", a list of articles will be recommended to the user. This option works based on the content-based filtering algorithm.

Process:
- Provides recommendations that are content-wise related to the target article
- First, the article selected by the user is analyzed
- Then, articles with greater content similarity to the selected article are recommended

3. Hybrid Approach (Collaborative + Content-Based Filtering) - "Personalized Recommendations"

To see this system in action, the user should:
1. After logging into the database, click on their name in the top bar of the homepage
2. Select the "Recommended Articles" option

After selecting this option, a page is displayed showing recommendations based on the user's behavior.

Important Note:
- Since no history exists for a new user, this system does not provide recommendations at the beginning of membership
- However, over time, as the user interacts with the database and creates a history of their visits, recommendations will gradually be generated for the user

Applications

• Recommending scientifically relevant articles based on user taste and history
• Discovering new and relevant articles in the researcher's field of expertise
• Saving search time through automatic personalized recommendations
• Increasing awareness of aligned research in the specialized field

Benefits

• Personalized recommendations based on actual user behavior
• Use of two main algorithms (collaborative and content-based)
• Learning capability and gradual improvement as user history grows
• Complementary recommendations through "Others' Suggestions" and "Related Articles"