This study conducts a targeted literature review on human-machine collaboration (HMC) within computer science from 2019 to 2023, utilizing topic modeling to address challenges such as outdated research and manual curation costs. We compare Latent Dirichlet Allocation (LDA) and Non-Negative Matrix Factorization (NMF) to determine their effectiveness in analyzing HMC literature. Our results show that NMF, which leverages TF-IDF weighting, produces more distinct and interpretable topics compared to the overlapping clusters often generated by LDA. Specifically, NMF identifies four key themes in HMC: natural language processing applications, the integration of human strengths with algorithms, advancements in image-related tasks, and the application and evaluation of AI agents. These findings suggest that NMF is better suited for capturing nuanced research trends in HMC. This study provides valuable insights for improving literature review methodologies and advancing the understanding of human-machine collaboration in computer science.