Escrever um comentário

Title: Automating Business Intelligence Requirements with Generative AI and Semantic Search

Summary Decision: Accept, with minor revisions.

This manuscript introduces AUTOBIR, a generative-AI and semantic-search architecture designed to automate Business Intelligence requirement gathering, query formulation, ontology discovery, and visual analytics. It is a timely contribution, given the increasing demand for intuitive, low-code analytical intelligence solutions in industry. The authors frame their motivation clearly, explaining why BI elicitation remains labour-intensive and prone to technical friction. The goal of the work is evident, which is to provide a system that can convert natural language into executable analytical queries and visualizations.

Methodologically, the paper was evaluated as a qualitative system-implementation study, rather than a quantitative performance benchmark. In this context, the authors provide adequate evidence of functionality through system architecture description, ontology-to-query workflow, screenshots, and Subject Matter Expert testing across multiple domains. Instead of metrics, the evaluation is supported through usage demonstration, tool behaviour, and real-world applicability. This is appropriate, but the work would benefit from presenting SME feedback as recognisable thematic categories (for example, usability impressions, semantic alignment, learning curve, task efficiency). Even without numerical scoring, a clear qualitative theme map would give the evaluation more structure and communicative power.

The visual components included are helpful.

Particularly Figure 3, which illustrates the system architecture clearly. One additional schematic showing the full interaction pipeline (i.e., natural language → ontology selection → query generation → visualization) would make the system’s workflow instantly digestible. A brief summary figure or paragraph capturing SME-observed strengths and limitations would also strengthen interpretation for readers outside the NLP/semantic systems community.

The writing quality is strong overall. Only minor grammatical corrections and readability touch-ups are recommended. For example, on page 2, the line “structured to promote a comprehensive un­derstanding” carries a hyphen artifact that should be corrected to “structured to promote a comprehensive understanding.”

On page 3, the explanatory sentence under Table I would read more smoothly if rewritten to indicate plural usage and clearer flow. For instance: “The results display the total earnings of each product, specifically for transactions converted to Euros, calculated by multiplying the line total by the average exchange rate.”

On page 4, the term “Setup tools” should appear in lowercase within the sentence context (“setup tools”).

On page 6, the sentence “A pre-trained or fine-tuned language model transform these strings…” contains subject-verb disagreement; it should read “transforms these strings…”

These grammatical errors do not undermine comprehension, but refining them will make the final manuscript more coherent to non-expert readers.

In conclusion, this is a high-quality, forward-relevant work that demonstrates meaningful progress in automating BI requirement workflows. With modest improvements the manuscript will be strong and publication-ready.

Competing interests

The authors declare that they have no competing interests.

Use of Artificial Intelligence (AI)

The authors declare that they did not use generative AI to come up with new ideas for their review.