Systematic Literature Review Langchain Proposed
Rakha Asyrofi, Mutia Rahmi Dewi, Muhammad Irfan Lutfhi, Prasetyo Adi Wibowo
Abstract
While systematic literature reviews are frequently carried out within software engineering research, performing them in a rigorous and reproducible manner can be difficult. This paper proposes some new methods for evaluating and validating systematic literature reviews. Our approach consists of several steps, such as: Selecting a set of relevant scientific papers to analyze, Developing a list of questions and criteria to evaluate each literature review, and Determining what types of functionality and performance should be evaluated. We tested our method by having multiple experts evaluate the literature reviews based on our questions and criteria. We measured the similarity in scores between each expert to determine the reliability of the evaluations. The average similarity index between experts was 0.58 to 0.83, indicating a reasonable level of agreement in their assessments. This shows our evaluation method can produce fairly consistent results, even when different experts are involved. The relatively high level of agreement is notable considering each expert brings their perspectives and opinions in analyzing literature reviews. By providing concrete questions, criteria, and evaluation methods, we aimed to guide the experts toward more uniform evaluations. In summary, we developed and tested a new approach for evaluating and validating systematic literature reviews in software engineering. By assessing reliability via interrater agreement, we showed that consistent and reproducible results are possible using our evaluation framework and methodology. Our methods could help researchers gain more insight into what makes for an effective and high-quality literature review.