[Paper Review] The BLue Amazon Brain (BLAB): A Modular Architecture of Services about the Brazilian Maritime Territory

Paulo Pirozelli, Ais B. R. Castro+20 more|arXiv (Cornell University)|Sep 6, 2022

Topic Modeling4 citations

TL;DR

The BLue Amazon Brain (BLAB) is a modular AI architecture designed to enhance awareness and provide intelligent services about Brazil's maritime territory, the Blue Amazon. It integrates a conversational agent (BLAB-Chat) with natural language processing modules, a data lake, and specialized services like BLAB-Reporter and BLAB-Wiki, enabling complex question answering and information dissemination despite challenges in data scarcity and domain-specific language modeling in Portuguese.

ABSTRACT

We describe the first steps in the development of an artificial agent focused on the Brazilian maritime territory, a large region within the South Atlantic also known as the Blue Amazon. The "BLue Amazon Brain" (BLAB) integrates a number of services aimed at disseminating information about this region and its importance, functioning as a tool for environmental awareness. The main service provided by BLAB is a conversational facility that deals with complex questions about the Blue Amazon, called BLAB-Chat; its central component is a controller that manages several task-oriented natural language processing modules (e.g., question answering and summarizer systems). These modules have access to an internal data lake as well as to third-party databases. A news reporter (BLAB-Reporter) and a purposely-developed wiki (BLAB-Wiki) are also part of the BLAB service architecture. In this paper, we describe our current version of BLAB's architecture (interface, backend, web services, NLP modules, and resources) and comment on the challenges we have faced so far, such as the lack of training data and the scattered state of domain information. Solving these issues presents a considerable challenge in the development of artificial intelligence for technical domains.

Motivation & Objective

• To develop an integrated, modular AI architecture focused on the Brazilian maritime territory, known as the Blue Amazon, to improve public and scientific awareness of its ecological, economic, and strategic importance.
• To address the challenge of fragmented and scattered oceanographic data by building a centralized, service-oriented system with natural language understanding and generation capabilities.
• To create and deploy domain-specific NLP tools in Portuguese—such as question answering, summarization, and knowledge graph construction—despite limited training data and low-resource language constraints.
• To establish a scalable infrastructure combining a data lake, conversational agent, automated news generation, and a thematic wiki to support environmental education and decision-making.
• To lay the foundation for future AI services, including predictive modeling for oceanic phenomena and ontology-based reasoning, by curating domain-specific datasets and resources.

Proposed method

• BLAB employs a controller-based architecture that orchestrates multiple NLP modules, including a question-answering system, multi-document summarizer, natural language-to-SQL converter, and knowledge graph generator.
• The system integrates a data lake with internal and third-party databases, enabling access to diverse sources of oceanographic, environmental, and economic information about the Blue Amazon.
• BLAB-Chat uses a pipeline of NLP components—such as a paraphraser, topic model builder, and natural language generator—to process and respond to complex, user-generated questions in natural language.
• Specialized services include BLAB-Reporter, an automated news generator, and BLAB-Wiki, a curated thematic knowledge base for the Blue Amazon domain.
• The system leverages both pre-trained multilingual models and custom-trained models in Portuguese, with a focus on adapting them to low-resource, technical domains.
• The architecture is designed to support future expansion with reasoning modules (e.g., debaters, planners) and numeric prediction systems (e.g., tide and current forecasting) based on metocean data.

Experimental results

Research questions

• RQ1 How can a modular, service-oriented AI architecture be designed to integrate diverse NLP capabilities for a complex, data-scarce domain like the Blue Amazon?
• RQ2 What are the key challenges in training effective NLP models for Portuguese in a specialized, low-resource domain such as marine science?
• RQ3 How can scattered and heterogeneous oceanographic data be unified and made accessible through a conversational AI interface?
• RQ4 To what extent can a conversational agent provide accurate, context-aware responses to complex, technical questions about marine ecosystems and maritime policy?
• RQ5 What role can curated datasets and domain-specific knowledge resources play in improving the performance and reliability of AI systems in environmental domains?

Key findings

• The BLAB-Chat system successfully integrates multiple NLP modules—such as question answering, summarization, and knowledge graph generation—into a single conversational interface for the Blue Amazon domain.
• Despite limited training data, the system demonstrates the feasibility of building domain-specific NLP services in Portuguese, a low-resource language, through targeted data curation and model adaptation.
• The development of two new QA datasets and a large document corpus for the Blue Amazon has significantly supported NLP model training and evaluation.
• The system’s architecture enables extensibility, with plans to integrate future modules such as argumentation reasoners, sequential planners, and metocean forecasters.
• The creation of BLAB-Wiki and BLAB-Reporter provides structured, reusable knowledge and automated content generation, enhancing public access to oceanographic information.
• The project highlights critical limitations in current AI models when applied to specialized domains, particularly the need for better techniques to handle scattered, non-curated knowledge sources.