SPi’s flagship, intelligent, end-to-end technology platform, SPARK brings together modular AI-based components to address the diverse data management challenges faced by enterprises across multiple verticals.

SPARK is a one-stop solution, addressing data acquisition, data quality, and the transformation of analytics-ready data. The platform and its modules are especially powerful when processing unstructured data, an ongoing challenge for enterprises of all kinds.

The platform features three principal modules.

Extraction Module
Built on an ML engine, SPARK’s extraction module leverages cognitive models to achieve industry-leading accuracy in extracting unstructured data from PDFs, documents, maps, emails and images.

The extraction module also uses SPi’s next-generation web harvesting platform, powered by AI and NLP accelerators, to extract information and monitor websites for raw, standardized data.

• Technology-assisted source acquisition for a more systematic and thorough approach to acquiring the right data.
• Nimble approach allows a single script to extract similar data from multiple sites.
• Data cleansing, structuring and noise removal at the acquisition stage save significant storage costs downstream.
• Web page monitoring tracks and captures website changes as they happen.
• Server-based PDF extraction and OCR reduces load on user devices.
• Automated identification of potential issues such as font details, uncertain spacing and soft hyphens within the extracted file, reduces the need for human curation.
• Table, image and diagram extraction included as standard.
• Small sample sizes for training high levels of automation through ML.

Transformation Module
Leveraging a quality framework built on our custom AI and ML models, and algorithms customized for each industry, SPARK’s transformation module creates and maintains a master dataset from all data fed into it.

• Proprietary knowledge repositories and key de-duplication and standardization suites allow SPARK’s transformation module to scrub reference data such as company and product names, geographies, phone numbers and subject matter expert (SME) terminologies.
• SPARK’s multi-level ML-based model helps disambiguate and link people, company and other datasets to create a single source of truth.

Enrichment Module
Leveraging SME knowledge across multiple domains, SPARK’s Enrichment Module offers information providers and other enterprises a unique opportunity to leverage existing datasets to their full potential.

This module includes entity extraction, taxonomy and classification, summarization, metadata creation and more.

• Customized rules and knowledge repositories for a wide range of domains including finance, legal, real estate, science, engineering, medicine, social sciences and humanities, and inter-disciplinary information.
• Software and SMEs working together ensure scalability and accuracy, supporting stringent SLAs.
• Keyword extraction, indexing and concept identification, along with SEO-optimized summaries.
• In-built NLP modules facilitate concept extraction, editorial and readability checks.

A unique conceptualization and development hub to foster innovation and experimentation, SPi Labs leverages next-generation technologies and open source capabilities. The facility brings together software engineers, SMEs, researchers, data analyst and technology specialists to expand SPi’s capabilities, accelerating the development of innovative solutions that enable organizations and society to make better use of big data and domain knowledge.

SPi Labs success stories include solutions developed for concept extraction and the automated business relevancy discovery in the health and science domain, and semantic analysis and NLP research to understand terminology in legal documents.

As part of this initiative SPi continues to invest in proprietary cognitive technology to target the complex problems commonly encountered in invoice processing and handwritten text recognition.

The essential first step to good data is identifying the right data. We apply proven technology and processes in our Source Discovery and Analysis, Data Acquisition, Cleansing and Normalization, De-duplication and Disambiguation, and Metadata Management services.

If data is oil, enrichment is the refinery. SPi has wide experience in enriching data across multiple industries, our enrichment services including Mapping and Linking, Entity Extraction (indexing), Summarization, Knowledge Modeling and Annotation.

Database Design & Maintenance

Our content support team handles the design of content and data structures and schemas, as well as workflows to implement new capabilities on your hosting platform.

We have rich experience in a wide range of proprietary and open source CMS platforms. Along with our in-depth knowledge of data products, this expertise enables us to provide you with a one-stop solution to optimize all your data processes.

Content Migration

SPi has teams highly experienced in migrating content from legacy platforms to new ones, with a detailed understand of how schemas need to be set up for unstructured data. Our full suite solution includes requirements analysis, content structure mapping, ETL process management, ingestion QA and testing. We have migrated millions of records for leading information providers in the risk and compliance, health and science, and legal domains.

Platform Build and Deploy

With SPARK, our workflow management system, MAGNUS, and custom platforms, SPi builds, customizes and deploys various platforms to help optimize data workflows. All our platforms are built with APIs which can be customized for any CMS and deployed in the cloud or on-premise.

SPi’s team of data experts, including data engineers and data scientists, helps customers with visualization design and implementation, integration of visualization tools with existing workflows, and report building, scheduling and distribution.

Focusing on an outcome-based, analytic approach, SPi helps companies use data to solve business problems. Using an iterative process in the creation of analytic models, we work with customer analytics, marketing analytics, product and sales analytics, operations and logistics analytics, and content analytics.

The client helps retailers and brands launch products, increase sales, and drive traffic by offering innovative customer rewards programs through its shopping rewards app. The geo-location technology within the app guides shoppers to the nearest store location where a partner’s products are available. With its pay-for-performance model and customizable product campaigns, retailers and brands that partner with the client have seen a significant return on investment.

SOLUTION
SPi Global provided end-to-end data collection and maintenance, including geo-tagging the location information of retailer storefronts. SPi Global leveraged cutting-edge technology to achieve scale as well as the expertise of in-house data analysts for maximum accuracy.

For this project, we developed a custom scraping script to collect and store location information from the official websites of the retailers. To improve data quality, our team then created an address-matching algorithm to identify new, relocated, and closed stores as well as any potential duplicates in the collected data.

After the initial build out, we periodically checked stores to ensure the latest data, taking into account any new or closed stores.

Ultimately SPi Global delivered a scalable and customizable solution with significant flexibility for the client to adapt data based on their product and promotion strategy, streamlining the onboarding of new retailers to the platform.

As the world’s most comprehensive patent information database, the client serves researchers, students, engineers, designers, technicians, scientists, and research and development (R&D) professionals who need comprehensive information on patents filed in their field.

The client wanted a scalable and accurate solution that would create value-added abstracts for improved patent discoverability within their intellectual property database.

SOLUTION
SPi Global deployed ConSCIse™, its proprietary platform-enabled content abstracting solution for the client. ConSCIse blends proprietary software with Subject Matter Expert (SME) curation to deliver high-quality abstracts.

As part of this solution, SPi Global developed a ConSCiSe module for engineering with NLP and ML models, facilitating contextual translation from 16 languages with integrated translation memory tools.

To ensure an optimal QA layer, we employed SMEs from diverse engineering domains (computers to mechanical).

The scaled solution was designed to cope with managing data from over 31 countries and reduced TAT from 21 days to 5 days.

A leading STM publisher was seeking a mechanism to scale journal transfers within their portfolio. This was offered when authors had listed alternate journals at the time of submission, but only a small percent of authors followed up on transfers. The use of multiple systems also created a challenge in building a system-independent workflow.

SOLUTION
To support the client, SPi Global set up an initial set of pilot journals on the Transfer Desk application, then built a custom journal recommendation model based on AI that uses the content in the chapter combined with business rules to recommend potential transfer journals.

A function established within the team to facilitate author follow-through, by managing the transfers once an author has accepted a transfer.

SPi Global launched this system in 12 weeks and went live with 10 journals. The project scaled to the current management of 200+ journals.

A leading global publisher, featuring over 100K publications across various content types including News, Business, Finance, and Legal.

This publisher wanted to publish new content quickly without sacrificing quality. It was important to monitor and ensure that all publications coming into their collection and conversion system were updated with high-quality content and on time.

SOLUTION
SPi Global had end-to-end responsibility for this offer — from securing the content to posting it on the internet.
Additionally, we owned the data analysis, design, and creation of conversion programs for new products included into the client’s online platform. Monitoring support was run 24/7 for troubleshooting conversion, display, and functionality issues for online products under maintenance.

We addressed content licensing compliance issues and work with source providers to resolve relationship issues, while also maintaining direct contact with client’s licensing unit and publishers.

Finally, we worked with an IT vendor to transition additional content management tasks previously part of the enterprise platform.

An integrated drug database with the most current, accurate, and technologically advanced drug data and drug decision support wanted to ensure the accuracy of their information.

A typical user relies on the client’s drug database to understand how much a drug costs across various states in the US. Therefore, having the most current drug pricing is essential for the client and their customers.

SOLUTION
SPi Global delivered an all-encompassing automated website monitoring, data extraction, and time-saving mapping tool. The solution involved development of customized workflow applications leveraging ACQUIRE and SPiZone.

This offer actively monitored changes in websites owned by state authorities for State Maximum Allowable Costs (SMAC) and Average Actual Cost (AAC) of drugs, then downloaded the latest pricing data. Additionally, we extracted drug and price information from downloaded PDFs using entity recognition models.

With this tool, we achieved a high level of automated transformation and load process with matching algorithms that mapped the drug names extracted from the PDFs to drug databases owned by the client.

Finally, to optimize client utility and savings, we developed a monitoring schedule based on predictive modelling that ensures data update frequency without burning a hole into infrastructure costs.

The client has a leading platform that streamlines and secures the entire accident reporting process from data capture, storage, and access to analysis and distribution. The platform primarily serves law enforcement agencies, individuals, insurers, and authorized parties.

The client was looking to reduce costs by optimizing the current process through automation without sacrificing high data quality.

SOLUTION

SPi Global worked to create an offer that captured critical data elements from police auto-accident report images – sourced from 64 state agencies, each containing 100 to 400 data fields – and optimized the reporting workflow.

By redesigning the traditional double keying method with a 2-phase iterative automation approach, we were able to deliver a functional technology solution to manage the variable workload. Our approach included:

• Phase 1: Implementation of an auto-extraction service powered by OCR+ business rules, then design an intuitive UI layer with a workflow, dashboard, and reporting mechanism to create a golden set for training
• Phase 2: Building on Phase 1 automation, increasing efficiency and accuracy of both extraction and QC through state-specific ML models

A leading provider of brand protection services to IP firms, brand managers, and legal firms needed a scalable and cost-effective solution to enable seamless discovery and management of customer searches

SOLUTION

Combining technology and SME knowledge, SPi Global created a tailor-made trademark database watch and management solution to deliver efficient and effective brand monitoring.

The offer utilized the latest computer vision technology to index and annotate the trademarks using Vienna code classifications, and created a knowledge repository of frequently requested trademarks for effortless and quick processing of trademark monitoring. It also included an intelligent workflow management module to direct routing based on SME specialization.

The system monitored nearly 400,000 trademarks annually for potential infringements with 99.985% accuracy, and progressively improved turnaround time from 14 to 3 days.