The automation of title extraction and topic modeling for academic texts represents a significant opportunity to bring efficiency and organization to research. The paper will explore two main methodologies for meaningful title generation and topic mining with respect to research papers and related academic data in this paper. The first approach utilizes Latent Dirichlet Allocation (LDA) for topic modeling, applied to two datasets, achieving coherence scores of 0.785 for Dataset 1 and 0.4724 for Dataset 2 (a larger and more diverse dataset from the arXiv repository). To improve the title extraction task, authors present a new method based on word importance ranking with TF-IDF and sentence construction with nearby word occurrences, which then gets refined syntactically with Part-of-Speech tagging. This proved to be quite promising in extracting correct titles relevant to the context of academic papers as the cosine similarity score came out to be 0.7891 while evaluated on BERT embeddings. Additionally, authors explored a deep learning-based topic modeling methodology using BERTopic, which combines BERT embeddings, UMAP dimensionality reduction, and HDBSCAN clustering for future, providing a more nuanced and context-aware analysis of research topics. This research paper provides a comparative analysis of these approaches and demonstrates their potential for enhancing research paper classification and summarization tasks.