.png)
MALENA at a glance
Trained by IFC on unique emerging markets ESG data, the MALENA AI solution uses natural language processing to analyze unstructured text such as annual and sustainability reports, news, and impact assessments, and rapidly identifies over 1,000 ESG risk terms and predicts sentiments based on context.
MALEMALENA: Machine Learning ESG Analyst
ESG Sentiment Analysis
About MALENA API
Model Training Data Summary
MALENA has been developed by retraining a large language model using a dataset of over 150,000 manual annotations. With 91% accuracy in ESG sentiment analysis, MALENA outperforms out-of-the-box sentiment analysis models with lower accuracy levels of 68%.
Responsible AI
MALENA is underpinned by a Responsible AI framework that incorporates a grounding feature, a model evaluation dashboard, and follows a machine learning operations process to provide transparency, enable users to assess reliability and fairness, and ensure traceability and auditability, promoting trust and transparency in the system.
MALENA API Input/Output
The MALENA API comprises two endpoints depending on the type of API input.
The `text-job` endpoint enables API users to execute one or more API requests using plain text as input. The `text-job` response endpoint provides API output in both JSON and CSV formats.
The `binary-job` endpoint allows users to process single or multiple binary files and supports various file types and extensions such as .PDF, .DOCX, and .TXT. The `binary-job` endpoint returns API output in both JSON and CSV formats.
API Data Security FAQ
As an IFC application, MALENA API is built on the World Bank Group’s cloud infrastructure and adheres to institutional security controls and systems for identifying and managing vulnerabilities.
.png)
.png)
.png)
.png)
