LaSER: Language-specific event recommendation.

Area of Science:

• Information Science
• Computer Science
• Human-Computer Interaction

Background:

• Societal events often have local impacts affecting specific language communities.
• Current entity recommendation systems neglect the crucial language context.
• This gap limits effective information retrieval for language-specific user needs.

Purpose of the Study:

• Introduce the novel task of language-specific event recommendation.
• Develop and evaluate an approach, LaSER, to address this task.
• Enhance the relevance of recommended events by incorporating language context.

Main Methods:

• Propose LaSER, a novel approach for language-specific event recommendation.
• Blend language-specific latent representations (embeddings) of entities and events.
• Incorporate spatio-temporal event features within a learning to rank model.
• Train the model using publicly available Wikipedia Clickstream data.

Main Results:

• LaSER significantly outperforms existing state-of-the-art recommendation baselines.
• Performance improvement reaches up to 33 percentage points in MAP@5.
• Demonstrated superior language-specific relevance of recommended events.

Conclusions:

• Language-specific event recommendation is a crucial task for information retrieval.
• LaSER effectively addresses the challenge by integrating language and event features.
• The proposed method enhances user experience in web navigation and exploratory search.