A paradigm shift where Information is no longer subsidiary to applications but is the application by itself.
Knowledge based solutions
bee4sense platform founding vision is that classical SQL/BI approaches cannot be successfully applied to Big Data. Our DNA is to converge structured and unstructured data into a unified storage and processing knowledge model inspired by Semantic Web
Corporations have access to an increasing variety of information via machine sensors, applications, open and linked data initiatives, and social media – the so-called Big Data. Any corporation is aware that its ability to convert this information into opportunities is critical.
These same information remains siloed and poorly accessible due to:
Multiplicity of internal and external sources
Applications, document management systems, databases, datalakes, professional information providers, internet, and social media.
Security and licensing constraints
Applied to information
Increasing variety of formats
Text, images, satellite imaging, videos, sounds, SQL and NoSQL, indices, feeds, IOT, Logs and more.
Variety of languages
The challenge is: “How to provide users with a unified access to all the information they need?”.
bee4sense solution, deployed on premise, private cloud and public cloud.
BEE4SENSE DESIGN PRINCIPLES
Enabling Information Into Knowledge
Data Quality and Integrity
Data integration is very often addressed through a large variety of technologies especially when it comes to non-structured data (text, images, sounds, videos…). As each technology will not mutualize processes and methods, data integrity and consistency across formats will not be ensured.
bee4sense’ Unified integration framework mutualizes structured and unstructured data acquisition and processing processes across formats:
- Data acquisition and transformation
- Format conversion
- Data and process lineage
Information Security and Governance
Ensuring that information is protected at all stage of its lifecycle.
Capturing and updating security defined in the various source systems and applying it at all stage of the data lifecycle from acquisition, through enrichment to publishing to end user.
Enabling security management throughout transformation processes by applying additional restrictions from source.
Information traceability and auditability
bee4sense makes sure that change history is properly recorded at any stage of the data lifecycle by keeping history of:
- Creation, deletion, and changes on structured data (field value before and after) and unstructured data (document plain text and metadata change).
- Models and process versions applied to data at any stage of its lifecycle.
- Data access by users and administrators and systematically logged.
Openness and interoperability
Big Data and AI is a very vast domain where to achieve the best a platform shall ease the encapsulation of third-party initiatives and technologies.
API first architecture to maximize interoperability across the bee4sense modules and third-party technologies. any bee4sense module is integrated through the same APIs used for third parties.
Connector framework SDK enabling third parties to develop new sources although mutualizing 80% of the code.
Multi-model storage and publishing to enable data to be interoperable with external requirements.
ML/AI unified annotation framework to encapsulate third party initiatives.
Unified Knowledge Model
Converging information into one unified Knowledge base is critical to make sure that users can contribute to capitalizing knowledge by using existing data, enriching data to create information and knowledge.
Semantic Web design principles have been systematically applied at any stage of the information lifecycle management to maximize convergence and cross fertilization of data assets..
Unified access to knowledge
End users are not interested in the technical representation of information, files, tables, fields, joins, spreadsheets, they only care about being able to create the information they need by their own with no technical skill.
Based on bee4sense Knowledge Graph, Search and Geospatial technologies, users can search, navigate from various representations of same information (Map, Graph, Search, Dashboard, Folders), enrich and customize information, share information.
BEE4SENSE FEATURES SUMMARY
3 Steps to Empower Users with Knowledge
Acquiring information and attached security
Enriching information through rules, AI and ML
Providing users with autonomous access to all the information they need
Infohub: Unified information access
Part of the bee4sense suite, Infohub, is an information federation platform aiming to acquire raw data and attached security whichever format, application and location and staging without any alteration in one unified multimodel storage for further enrichment. Unifying storage enables to avoid complex synchronization processes and guarantees data integrity across required storage models.
Acquire raw data whichever format, application and location from 100+ application connectors and format converters
Update continuously detecting changes and updating to make sure that data is up to date
Store without any alteration in one unified multi-model storage for further enrichment.
Secure information by acquiring and broadcasting security rules defined in source systems and enrichment processes downstream
Govern information through data and process lineage
Garant access to AI and ML processes downstream and to users through search.
Learn, Enrich and Link
Sense builder leverages a modern Semantic Web Information Processing architecture enabling to handle the Big Data variety of information formats and volume:
Elastic infrastructure enabling to handle massive volumes of data and processes.
80%[i] of data scientists’ workload is allocated to accessing and preparing data, a single point of access to any data is critical to maximize ROI.
AI processes require data to be accessed and publish in different formats which is addressed by bee4sense’s Infohub multi model storage.
[i] Cleaning Big Data: Most Time-Consuming, Least Enjoyable Data Science Task, Survey Says – Forbes – March 23, 2016 by Gil Press
Generic model deployment for enrichments
Industrializing model deployment and lifecycle management is also a challenge that Sense Builder addresses by providing a generic industrialization framework to Data scientists and third-party AI solutions providers:
- Data transformation and rules engine
- Data conversion image to text, speech to text
- Natural Language processing and text mining
- Computer vision via selected partners
- Satellite detection via selected partners
- Machine learning and AI framework
The information enriched is converged to:
- A unified Knowledge base enabling to cross fertilize information and highlight meaningful information directly to users.
- Infohub to update data lineage and ensuring that all information is converged.
oppScience focuses on its core competencies, Natural Language Processing, Record Linkage, Machine learning and AI and rigorously selects and integrate third party best of breed initiatives inside the bee4sense Fablab.
bee4sense Fablab is an API first architecture enabling partners to be fully integrated into bee4sense to converge all enrichments into required use cases.
Join the bee4sense Fablab
Publish & Use
Infohub : Unified information access
bee4sense SARA provides users a single point of access to search, navigate, view and enrich information through different representations as map, timeline, graph, list or file representation.
Information relation is unified through a graph representation enabling users to access more details to organize or enrich information in a meaningful way, to upload their own information into the system and share representations with others.
Successful User Experience is all about providing users with a pleasant usage experience that beyond User Interface (UI) considers, user profile, context of use, information access and navigation in the most effective manner.
Platform Detailed Features
Acquiring raw data and attached security whichever format, application and location and staging without any alteration in one unified multimodel storage for further enrichment. Unifying storage enables to avoid complex synchronization processes and guarantees data integrity across required storage models.
Connectors library - Over 100
- Structured (Applications, DBMS, NoSQL)
- Semi structured (Emails, Feeds, Conversations)
- Unstructured (Documents, Web and text formats, Sounds, Videos, images and Satellite imaging)
- Massive crawling (Web,Darkweb)
- Additional sources through Connectors SDK mutualizing 80% of required code
Security acquisition, update and propagation
- Connectors to Corporate and Application directories
- Security on columns
- Security on lines
Data lineage from raw data acquired
- Logging change history at schema and field value levels including security
- Data lineage audit console
Multi model unified staging
Connectors administration & monitoring
- Connectors administration and orchestration console : batch and streaming from sources and user upload
- Multimodel staging administration and monitoring
- Fully distributed although enabling for centralized administration
Unlike classical data enrichment heritating from structured data techniques, Sense builder leverages a modern Semantic Web Information Processing architecture enabling to handle the Big Data variety of information formats and volume:
Data transformation and format conversion
- Structured data enrichers library
- Rules engine
- Text conversion (plain text, HTML, XML, Tables and document structure)
- Image to text (OCR)
- Speech to text
Natural language processing and semantics
- Non configurable named entity extraction from text
- Advanced named entity extraction
- Relationship extraction
- Opinion polarity tagging from text
- Knowledge graph
- Natural language processing SDK
Image and video computer vision
- Face and character recognition and tagging
- Object recognition and tagging
- Image and video classification
- Computer vision SDK
Satellite image pattern and object detection
- Environmental characteristics (forest, building, water, biodiversity)
- Shape detection (vehicle, building, parking, field, infrastructure)
- Economic intelligence (monitoring of production site activity, traffic, transportation)
- Teledetection SDK
Machine learning and AI modeling framework
- Data science and NLP unified annotation framework
- Encapsulation of third-party and proprietary algorithms in a generic framework
- Schema management
- Administration console
- Packaging & life cycle management of models
- ML/AI standard libraries
- Supervised / unsupervised classification, clustering, sequential models, text mining
- SVM, HMM, CRF, RNN, graph mining
- Third party API framework
Administration and orchestration console
- Process and model orchestration
- Transformation and enrichment data lineage and versioning (back to infohub)
- Monitoring and exception handling
Unlike classical BI representation, Big Data requires UX to dynamically adapt to context, data, volume and density of relationships. A unified and interoperable UX enabling users to access and enrich information without any technical skill from one entry point.
Configurable unified information access UX & APIs
- On the fly user information upload and processing
- Reference data management
- Ontological data management
- Case management system
Let's talk about your project !
14 Avenue Trudaine
M-F: 9am - 7pm
+33 (0)1 77 75 73 00