Avogatto

The problem

Legislative texts are written to be complete, not to be read quickly: articles referring to other articles, definitions that apply to a whole chapter, amendments overlapping in time. Whoever has to understand what a rule says about one precise case spends more time finding their way than reading. I wanted to find out whether natural language processing could do that orientation work: read a legislative text, understand its structure and return what is needed.

The solution

Avogatto was born as an artificial intelligence project dedicated to processing and understanding legislative texts. I conceived and developed it as an experiment: natural language processing applied to a domain, the legal one, where precision matters more than fluency. Along the way the project became a product, with the questions a product brings with it: what it does, for whom, how it is maintained. It was later transferred to a company in Ticino.

Technologies

No product list: these are the project’s technical fields.

• Artificial intelligence
• Natural language processing (NLP)
• Legal tech
• Product development

The result

Avogatto is, for me, an example in four fields: AI experimentation, natural language processing, legal tech and product development. The most concrete result is the transfer: a project born as an experiment became something a company in Ticino took over. The company is not named on this page. The path is the one told here: an idea, an experiment, a product, a transfer.