Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Technical Report on classification of literature related to children speech disorder

Ziang Wang, Amir Aryani

TL;DR

This work tackles scalable synthesis of biomedical literature on childhood speech disorders by deploying a reproducible NLP pipeline that combines LDA and BERTopic on a PubMed-derived corpus. It shows that 14 clinically meaningful topics emerge, with LDA achieving a coherence of 0.42 and perplexity of -7.495, and BERTopic yielding a low outlier rate (<20%) while enabling hierarchical topic exploration. The study provides open data and code to facilitate replication and further exploration, and discusses trade-offs between interpretability and semantic richness, offering directions such as ontological integration and dynamic topic modeling. Overall, the approach supports automated literature reviews in speech-language pathology and lays groundwork for ongoing, cross-disciplinary knowledge synthesis.

Abstract

This technical report presents a natural language processing (NLP)-based approach for systematically classifying scientific literature on childhood speech disorders. We retrieved and filtered 4,804 relevant articles published after 2015 from the PubMed database using domain-specific keywords. After cleaning and pre-processing the abstracts, we applied two topic modeling techniques - Latent Dirichlet Allocation (LDA) and BERTopic - to identify latent thematic structures in the corpus. Our models uncovered 14 clinically meaningful clusters, such as infantile hyperactivity and abnormal epileptic behavior. To improve relevance and precision, we incorporated a custom stop word list tailored to speech pathology. Evaluation results showed that the LDA model achieved a coherence score of 0.42 and a perplexity of -7.5, indicating strong topic coherence and predictive performance. The BERTopic model exhibited a low proportion of outlier topics (less than 20%), demonstrating its capacity to classify heterogeneous literature effectively. These results provide a foundation for automating literature reviews in speech-language pathology.

Technical Report on classification of literature related to children speech disorder

TL;DR

This work tackles scalable synthesis of biomedical literature on childhood speech disorders by deploying a reproducible NLP pipeline that combines LDA and BERTopic on a PubMed-derived corpus. It shows that 14 clinically meaningful topics emerge, with LDA achieving a coherence of 0.42 and perplexity of -7.495, and BERTopic yielding a low outlier rate (<20%) while enabling hierarchical topic exploration. The study provides open data and code to facilitate replication and further exploration, and discusses trade-offs between interpretability and semantic richness, offering directions such as ontological integration and dynamic topic modeling. Overall, the approach supports automated literature reviews in speech-language pathology and lays groundwork for ongoing, cross-disciplinary knowledge synthesis.

Abstract

This technical report presents a natural language processing (NLP)-based approach for systematically classifying scientific literature on childhood speech disorders. We retrieved and filtered 4,804 relevant articles published after 2015 from the PubMed database using domain-specific keywords. After cleaning and pre-processing the abstracts, we applied two topic modeling techniques - Latent Dirichlet Allocation (LDA) and BERTopic - to identify latent thematic structures in the corpus. Our models uncovered 14 clinically meaningful clusters, such as infantile hyperactivity and abnormal epileptic behavior. To improve relevance and precision, we incorporated a custom stop word list tailored to speech pathology. Evaluation results showed that the LDA model achieved a coherence score of 0.42 and a perplexity of -7.5, indicating strong topic coherence and predictive performance. The BERTopic model exhibited a low proportion of outlier topics (less than 20%), demonstrating its capacity to classify heterogeneous literature effectively. These results provide a foundation for automating literature reviews in speech-language pathology.

Paper Structure

This paper contains 14 sections, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Data Collection
  3. Clustering Methods
  4. Latent Dirichlet Allocation (LDA)
  5. BERTopic
  6. Discussion
  7. Topic Modeling with LDA
  8. Topic Modeling with BERTopic
  9. Comparison and Interpretation
  10. Limitations and Future Work
  11. Data and Code
  12. Data
  13. Code
  14. Conclusion

Figures (5)

  • Figure 1: Overview of the topic modeling workflow using LDA and BERTopic
  • Figure 2: Optimal Number of Topics Evaluation for LDA
  • Figure 3: Intertopic Distance Map for LDA
  • Figure 4: Hierarchical Clustering for BERTopic
  • Figure 5: Heat Map for BERTopic