Domain-specific Topic Model for Knowledge Discovery in Computational and Data-Intensive Scientific Communities.

Area of Science:

• Computational biology
• Neuroscience
• Data science

Background:

• Accelerating knowledge discovery is crucial for data-intensive fields like bioinformatics and neuroscience.
• Domain scientists face challenges querying vast, diverse information for new methods, tools, or datasets.
• Existing methods struggle to extract specific, relevant knowledge from large scientific corpora.

Purpose of the Study:

• To propose a novel domain-specific topic model (DSTM) for uncovering latent knowledge patterns.
• To enhance the discovery of relationships among research topics, tools, and datasets within specific scientific domains.
• To improve the efficiency and accuracy of knowledge extraction for domain scientists.

Main Methods:

• Developed a novel domain-specific topic model (DSTM) as a generative model.
• Extended the Latent Dirichlet Allocation (LDA) model.
• Employed the Markov chain Monte Carlo (MCMC) algorithm for unsupervised latent pattern inference.
• Applied the DSTM to large datasets from bioinformatics and neuroscience (over 25,000 papers).

Main Results:

• The DSTM demonstrated superior performance compared to state-of-the-art baseline models.
• Achieved better results in discovering highly-specific latent topics within domains.
• Evaluation metrics for generalization and information retrieval confirmed model effectiveness.

Conclusions:

• The DSTM significantly improves knowledge discovery in specialized scientific domains.
• Facilitates the identification of intra-domain, cross-domain, and emerging knowledge trends.
• Offers a powerful tool for researchers in computational and data-intensive fields.