I have always been very fond of maintaining healthy eating habits. This habit of mine has gained more foundation and greater interest in seeking detail in the composition of food and its interdependence in causal relationships with health problems in the last 8 years in which I became a vegetarian. NuHealthyPath stands for: Nutrition Healthy Path
What it does
This solution leverages the Semantic Search and explainability power of a Knowledge Graph based on ontologies in order to bring accurate answer for questions related to Malnutrition diseases, Fittest recipe and dish composition given some preferences or restrictions (can you imagine how many food restrictions Vegans have to face, from regular comercial feeding?).
How we built it
I load several ontologies and non-structured datasets, applying over that Natural Entity Recognition and Relation Extractions processing to build uppon that a Knowledge Graph. Having Neo4j for traversal queries through the graph representation of the concepts and other knowledge assets extracted from datasets and combining that resultset as enrichment for a inverted index search into Elastic from the Search Engine context - the entry point for the user query interpretation -, then the list of entities is displayed to the user in the Ract frontend application.
Challenges we ran into
To develop data loader for each specific dataset and also fine tunning the Knowledge Graph data model in terms of normalization given different vocabularies to handle from different datasets.
Accomplishments that we're proud of
Improve my knowledge regarding nutrition, and how impactant in our health a better understanding of that can be.
What we learned
Was a pleasure to improve my knowledge regarding search engine and how Neo4j can improve the relevancy and explainability on that perspective, specially when in comes into the Knowledge Graph and Ontologies world. Also was very interesting to play with Neo4j Graph Data Science Algorithms, specially those ones for similarity calculation.
What's next for NuHealthyPath
For the future, I can create web-scrappers to retrieve data from Supermarket/Grosseries websites and also perform some recommendations to the user from that. Also would be interesting to onboard more complex datasets to expand the relevancy or broaden the knowledge domains for the search queries