# Searching for Relevant Lessons Learned Using Hybrid Information Retrieval Classifiers: A Case Study in Software Engineering

arXiv: https://arxiv.org/abs/1812.05168 · HTML: https://mlsanity.com/paper/1812.05168/

## Key facts

- Cited by 4 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 12 Dec 2018.

## Built on by (most cited first, 4 of 4)

- [Can Open-Source LLMs Compete with Commercial Models? Exploring the Few-Shot Performance of Current GPT Models in Biomedical Tasks](https://mlsanity.com/paper/2407.13511/) (2024)
- [Is ChatGPT a Biomedical Expert? -- Exploring the Zero-Shot Performance of Current GPT Models in Biomedical Tasks](https://mlsanity.com/paper/2306.16108/) (2023)
- [News Article Retrieval in Context for Event-centric Narrative Creation](https://mlsanity.com/paper/2106.16053/) (2021)
- [Can Language Models Critique Themselves? Investigating Self-Feedback for Retrieval Augmented Generation at BioASQ 2025](https://mlsanity.com/paper/2508.05366/) (2025)

## Abstract (arXiv)

The lessons learned (LL) repository is one of the most valuable sources of knowledge for a software organization. It can provide distinctive guidance regarding previous working solutions for historical software management problems, or former success stories to be followed. However, the unstructured format of the LL repository makes it difficult to search using general queries, which are manually inputted by project managers (PMs). For this reason, this repository may often be overlooked despite the valuable information it provides. Since the LL repository targets PMs, the search method should be domain specific rather than generic as in the case of general web searching. In previous work, we provided an automatic information retrieval based LL classifier solution. In our solution, we relied on existing project management artifacts in constructing the search query on-the-fly. In this paper, we extend our previous work by examining the impact of the hybridization of multiple LL classifiers, from our previous study, on performance. We employ two of the hybridization techniques from the literature to construct the hybrid classifiers. An industrial dataset of 212 LL records is used for validation. The results show the superiority of the hybrid classifier over the top achieving individual classifier, which reached 25%.