Isagog Knowledge Graph service API specification
The Isagog platform for Knowledge Graphs consists in the following logical modules:
- interaction: user interaction interfaces
- knowledge: concept-level methods and structures
- language: natural language processing methods and structures
- data: instance-level manipulation, search and query methods
Each module is tagged and features a specific root path.
The supplied maven pom file produces a java client and may be modified to generate any resource by openapi tools
The supplied maven pom file produces a java server stub and may be modified to generate any resource by openapi tools
Here is a high-level sketch of the platform's architecture featuring two main use cases:
This picture shows how a user utterance should be processed.
- The utterance (raw text) is received by the interaction service and forwarded to the language service.
- The language service gives back an annotation structure over the given sentence, or a rebuttal message.
- The interaction service sends the annotated sentence to the knowledge service, which tries to build conceptual frames upon it (here is most of the magic!).
- The interaction service selects the best frame candidate ans sends it to the user to review.
- The interaction service dispatches the curated frame to the data service, based on its type, i.e. Query or Update. Users' frame curation may feed a continuous learning process.
- The data service result flows to the interaction service and then to the user.
Here is a sketch of the data ingestion, i.e. texts such as documents or mails, or rdf files previously prepared.
- The ingestion service discriminates the file type: it sends texts to the language service for content analysis, and structured data (e.g. csv) to the knowledge service for conceptualization.
- The ingestion service pulls the textual content in the document store, and possibly the extracted triples in the data store.