Intelligent assistant for editorial archive consultation

The challenge

Editorial archives accumulate a large amount of knowledge over time, but traditional keyword search mechanisms are not always sufficient for complex questions, comparisons, or queries that require relating information distributed across different articles. This reduces the ability to explore and reuse the value of already published content.

The solution developed

CISP1 developed an AI assistant connected to the editorial archive. The application interprets questions in natural language, performs semantic search on the available content, and retrieves the most relevant articles and excerpts to build the answer.

Answers are generated from the retrieved content and feature references to the articles used, allowing the user to validate and deepen the information directly at the original source. The assistant also maintains the context of the conversation, enabling subsequent questions and more complex analysis without the user needing to restart the search at each interaction.

How the application works

• Interprets the question and the accumulated context of the conversation.
• Performs semantic search in the indexed editorial archive.
• Retrieves articles and excerpts relevant to the query.
• Generates a contextualized answer based on the retrieved content.
• Presents references to the articles used and maintains context for subsequent questions.

Value for the operation

• Facilitates information discovery in large volumes of editorial content.
• Accelerates research that requires relating multiple articles and contexts.
• Increases the use of the archive as a knowledge base.
• Improves verifiability of answers through indication of the sources used.

Capabilities applied