Our platform utilises state-of-the-art Artificial Intelligence tools and Machine Learning solutions to navigate, clean and connect tens of millions of historical, unstructured and disparate data records to provide clarity.
We then apply Process Automation to navigate the complex legal and administrative tasks of establishing clean title ahead of sale.

Legal Process Automation

We have worked with experts to create a legal manual detailing every conceivable obstacle to establishing clean title to land ahead of sale. To streamline this activity, we are automating vital elements of the process and employing a range of techniques such as: user document classification, legal paperwork generation, smart forms and automated communication between all parties.

In doing this, we will save thousands of hours of manual legal work, instead of providing our legal team with all the information they need to effect and manage the process through the courts as efficiently as possible.

Deep Neural Networks

As the current, most sophisticated of Machine Learning methods, the use of Neural Networks underpins our approach to solving extreme complexity.

From organising and cleaning unstructured textual data on real estate, identifying and linking records, to developing predictive models for business processes, our proprietary Deep Learning algorithms are what makes Terra Adriatica’s mission possible.

Natural Language Processing

By building upon the latest advances in Natural Language Processing, such as Facebook's FastText and Google's BERT models for vectorised representation of text, we have succeeded in extracting highly structured information buried in noisy ("dirty") free text.

We do this by extracting information from unstructured records with our Fine-Grained Named Entity Recognition Module and categorising land parcel related records with a multi-class neural network classifier.

Semantic Search

Historical and geographic variations of Croatian names, as well as simple typos present a significant challenge when searching for credible links between records. Using our core NLP modules and an embedded representation of semantic similarity of records, we have developed a proprietary Semantic Search Engine for retrieving information on possible land titles in Croatia from the data our users provide us.

Do you have
3 minutes
to spare?
Be a friend!