ElinarAI – our own product
ElinarAI takes data and information management to the next level. It is possible because of advanced Artificial Intelligence (AI). It really provides comprehensive tooling for applying AI on content-centric problems.
Elinar Artificial Intelligence
Elinar has been heavily involved with advanced topics on text analytics for over a decade. We have also deep expertise in text analytics and inherent problems associated with linguistics, word dictionaries, and regular expressions. These things, just named a few, have enabled us to take unstructured data analytics into the next level.
Elinar has a strong history on augmenting IBM Watson NLU®, IBM Watson Explorer®, IBM StoredIQ®, and IBM Datacap®. These work with AI provide AI-Augmented Enterprise Content Analytics and Management solutions. Our own R&D on Unstructured Data AI started 5 years ago. In the end, this work has resulted in significant Elinar IP: ElinarAI
ElinarAI core capabilities are:
Complex Feature Extraction from unstructured data. For example, ElinarAI is able to:
- Extract Order or Invoice lines.
- Deduct from Field Service Men report which parts were damaged or what was the reason for the damage.
- Tell why customer got angry from a customer complaint.
- Create a structure on Medical Record; extract a reason for a doctor visit and tell what medication was ordered.
- Extract complex privacy information from a document.
- Help with redaction tasks; ElinarAI can discover what needs to be redacted from a record.
Advanced Classification of a document or business record.
So the ElinarAI can:
- Learn and create virtually any type of classification, for example, security classification.
- Use multi-valued complex classifications, for example, which body part was injured.
- Mix complex classification schemas together; for example, there can be tens of classification categories (like MRI, Rtg, Specialist Visit, Generalist Visit, Nurse Visit, Damaged Body Part and so on).
ElinarAI uses supervised learning. It learns from samples that can be either obtained automatically from existing business process/system or generated manually by humans.
ElinarAI contains all tooling for Training Data Preparation, AI Training, and Scalable Inferencing. Our solution is not a framework that requires deep involvement by corporate IT developers. Instead, it allows the business to create AI training data in an intuitive environment. And once the training data has been created (or provided) ElinarAI provides inferencing using the simple but secure REST interface to be used by any corporate process.
ElinarAI is an AI tool that serves virtually any industry using our generic text analytics (ElinarNER or IBM Watson NLU®). Elinar has specialized text analytics modules available for several industries and use-cases:
ElinarAI is available for our technology partners for reselling and distribution. We offer full development and deployment support for your AI projects. We also support “Black Box”; you can take Elinar technology and brand it as your own.
Elinar has currently a large number of “Product Seeds” that we encourage our partners to build into full industry/topic specific repeatable products. We also have proven the technology we use. So now we are looking for partners that possess deep industry knowledge and a vision on how to capitalize it in large scale.
Examples of current partner products/offerings:
- Privacy data mining (Cross-Industry)
- Legal resolution analysis (Legal)
- Medical records redaction (Medical)
- Senior Citizen home care report analytics (Social Services)
- Medical diagnostic support (Medical)
- Invoice → Accounting (Financial Services)
Partner channel is a strategic growth area for Elinar for next years, so do not hesitate to contact us for an in-depth discussion!
ElinarAI consists of two major components. There are ElinarAI Miner (Docker Container) and actual AI instances that are run on Tensorflow Serving containers. Commonly Customer Data Center hosts ElinarAI Miner instance. AI Instances are running at Elinar data center on IBM Power8/9 HPC appliances. We also support fully customer hosted option with additional cost. We do AI model training at Elinar data center using IBM Power8/9 HPC appliances.
With ElinarAI, you will get full REST API for business processes and external systems to query AI and provide training data. It also provides highly optimized user experience for manual training data entry with an intuitive approach for both tagging and advanced document classification.
ElinarNER (Elinar Named Entity Recognizer) is one major part of ElinarAI Miner. In practice, it prepares and anonymizes data for both AI training and inferencing. ElinarNER has several custom models available from Privacy Data to Medical Vocabulary.
Model training happens 100% anonymized (pseudonymized) data with significant portions of contents heavily changed. We are training models to recognize meaningful patterns in unstructured data. It’s important because with that we can remove all identifying information and business secrets from the training data. This ensures that even the strictest information security and privacy requirements can be covered with ease. In practice, it means that when we are training the model on Sales Orders (SO) we can’t deduct from training data following things: For example, which product was ordered, how many of them or by which individual.
Deep Learning Models
The TensorFlow models of ElinarAI vary based on the complexity of the business problem and the volume of training data. They range from simple classification to sophisticated deep models for data extraction. And in this case, simple classification means a fairly narrow construct with a few Convoluted layers and 3-5 recurrent layers.
We have also developed a large “cookbook” of deep learning models for various text and content-centric problems. Elinar specialists do naturally have a long experience with these models. They will fine-tune the models with the client architects for optimum performance and accuracy based on business requirements. For example, AI that creates a security classification uses quite much different topology from AI that needs to extract 40 different topics (like a reason for complaint) from a document. These classification levels are Secret, Confidential, Company Confidential and Public.
Training data management and Workflows
All customer data is managed by ElinarAI. Of course, you can manage training data as per project/topic basis. We’ve secured training data, so individual trainers will only see data assigned to them. This ensures secure training data creation. Administrative users can assign training data to individuals in bulk. They can also access training data created by each individual to ensure the quality and correctness of data. To use ElinarAI, it allows the business to leverage the crowdsourcing and other innovative ways to mitigate the cost of AI implementation.
- ElinarAI is a comprehensive solution for understanding unstructured data
- Leverages State-of-the-Art Text Analytics (Watson NLU or ElinarNER) together with Deep Learning
- It includes IBM Watson and Elinar-developed AI solutions.
- Runs on top of IBM Watson Machine Learning Accelerator (formerly known as IBM PowerAI); leverages Tensorflow, Tensorflow Serving and TensorRT