Enhancing enterprise knowledge retrieval via cross-domain deep recommendation: a sparse data approach.

Area of Science:

• Computer Science
• Information Retrieval
• Machine Learning

Background:

• Enterprise knowledge retrieval is hampered by data sparsity and inefficient cross-domain knowledge transfer.
• Traditional recommendation methods struggle with these challenges in enterprise settings.

Purpose of the Study:

• To develop and evaluate a novel cross-domain recommendation model (CDR-VAE) for enterprise knowledge management.
• To address data sparsity and improve cross-domain knowledge transfer using deep learning.

Main Methods:

• Developed a hybrid autoencoder with domain alignment, termed CDR-VAE.
• Tested the model on an enterprise dataset and the Movies&Books benchmark.
• Evaluated recommendation performance using metrics like HR, Recall, and NDCG.

Main Results:

• CDR-VAE achieved superior performance (HR=0.642, Recall=0.432, NDCG=0.715) compared to existing models.
• Shared latent representations were crucial for effective cross-domain learning.
• The model demonstrated robustness in sparse enterprise scenarios, with users favoring relevant content types.

Conclusions:

• CDR-VAE effectively mitigates sparsity and cross-domain barriers in enterprise knowledge management.
• The model offers theoretical and practical insights for deep learning-based recommendation systems in data-scarce environments.
• This research advances the field of knowledge retrieval and recommendation systems.