- From: Sebastian Samaruga <ssamarug@gmail.com>
- Date: Mon, 15 Oct 2018 16:47:04 -0300
- To: W3C Semantic Web IG <semantic-web@w3.org>, public-lod <public-lod@w3.org>, public-rww <public-rww@w3.org>, public-aikr@w3.org
- Message-ID: <CAOLUXBtp2=6xN=c9q1xTG65fUrkwTwSw_D38PJXjWxuVCkN5Qg@mail.gmail.com>
Draft: could a framework of patterns leverage SW / ML adoption for end to end business applications / BI use cases: (sending this again because download link in blog is not working) Semantic Web / RDF as the 'glue' of / for ML dataflow encoded input 'features' / output 'tensors' and ontology aligments. Mappings for knowledge input / augmented (learning) output rows (RDBMS example) processed by ML models in semantic alignments. SW contexts encodes 'meaning' into translated input features / obtained learning output tensors via RDF CSPO quads Resource ID creation / assignment algorithm. Tensor shapes rendered as algorithmically Resource ID enabled (ANNs activation functions) operating over and preserving Resource IDs / Statements integrity (validation). Dataflow semantic 'forms' application language: encode code and data functionally as Context RDF quad statements. Context activations performs functional intra / inter Context transforms across layers statements. Encoded 'form' statement resolves to getters / setters applications by means of algorithm to obtain resulting inferred 'forms' (templates: system resource encoded quads). Tools: *TensorFlow / ML:* I/O: Tensors (features / classes, discrete values in 'shapes'). Learning: Classification: class / instance identification. "Messi: Player : 10" Learning: Clustering: similarity (common attributes / links resolution). "Messi player of Barcelona". Learning: Regression (discrete value in function of input features, roles in contexts: value / event for x when y was z in w). "Messi captain of Barcelona in last tournament". Semantic Microservices (proposed component): I/O: RDF encoded features / outputs CSPO quad statements (reactive stream events bus). Resource ID creation / assignation algorithm (Semantic IDs: operable, tensor embeddings). Augmentation: Aggregations: data, schema, behavior statement layers dimensional aggregation. Type inference by attributes / values aggregation. Alignments: ID resolution: class / instance identity discovery (ontology / schema matching) ML models. Attributes / links resolution: clustering ML models Roles in contexts resolution: regression ML models *Distribution / Dataflow:* Integration / Discovery / Activations. Contexts / Layers: Dimensional upper ontology layers alignments between Contexts (data, domain, application levels). Reactive Extensions (RX). Dataflow 'forms' enabled 'templates' inter context levels. Activation: Resources Context's streams as observers / observables (RX) of Context / upper layers events. Event ('form') fires node augmentation (learning) / resolves to nodes emmiting knowledge 'forms' events related to their knowledge of the source event. *Use Cases:* Semantic Microservices Adapters (endpoints, integration / transforms). GraphQL: adapters schema / tenmplate transforms. Forms functional language translation (I/O: integration). Adapters: Workflows / API Rendering (OData, REST: Spring HATEOAS / HAL). Refine / ETL (Adapters I/O). Big Data Deployments (Adapters I/O). BI / Dashboards (Adapters I/O). Declarative Business Applications Framework (Adapters I/O). *Links:* OData: https://www.odata.org Spring HATEOAS: https://spring.io/projects/spring-hateoas HAL: http://stateless.co/hal_specification.html http://openrefine.org/ http://www.opencalais.com https://solid.mit.edu Sebastian Samaruga. http://exampledotorg.blogspot.com
Attachments
- application/pdf attachment: Schema.pdf
Received on Monday, 15 October 2018 19:48:53 UTC