Enhanced Consultation of the Italian Encyclopedia

ADVANCED FEATURES

NEW WAYS OF ACCESSING ENCYCLOPEDIC KNOWLEDGE

Generative artificial intelligence transforms access to the content of the Enciclopedia Italiana (Italian Encyclopedia). Through four innovative features, the first volume of the Enciclopedia Italiana opens up to new methods of consultation.

Automatic synthesis in four types of summaries, English translation, vocalization, and conversion into synoptic tables: each tool is designed to make consultation more dynamic, intuitive, and accessible to all.

The aim is to make encyclopedic heritage accessible; not only through traditional methods but also through new interactive experiences that simplify access to knowledge.

insights

A TECHNOLOGICAL INNOVATION LABORATORY

The project tested several latest-generation linguistic models to process the contents of the first volume of the Grande Enciclopedia, developing specific solutions to manage complex texts in HTML format, large entries and adapt the output to different formats.

Six Large Language Models (LLMs) and multiple voice technologies were evaluated, optimising the systems for each functionality. Each solution was subjected to thorough testing and evaluation by the Treccani editorial team.

SUMMARY

Four types of automatic summaries (short, medium, simplified, bullet points) and translation of entries into English, with full preservation of the original meaning and structure. Text-to-speech synthesis for listening to content, aiming to make the encyclopedia accessible through speech.

TRANSLATION

Automatic translation of entries, preservation of HTML tags, and management of abbreviations and specialized terms.

QUIZ

Multiple-choice quizzes automatically generated from encyclopedic entries through LLM models.

Questions and Answers

A NEW WAY OF CONSULTATION

No longer searching by entries, but questions asked directly in natural language. The question-and-answer system retrieves the most relevant information from encyclopedic content and generates responses that consider the context.

Thanks to the integration of a knowledge graph, the system searches and connects information distributed across multiple entries, offering a broader and more articulated view. A more immediate, intuitive access to knowledge capable of addressing the complexity of users’ questions.