Solutions

KaaS

By 2025, almost 80 percent of the data at the enterprise level will be unstructured. Usually, a lot of time and money is spent on organizing the data into structured forms like data warehouses. Given the exponential growth in generation of new content, we need tools to directly access and process the unstructured data rather than try to curate them first.

KaaS (Knowledge as a Service) is our suite of services to mine unstructured data. Using Artificial Intelligence and Machine Learning, KaaS extracts and organizes relevant data while hiding the complexity of AI models and frameworks. Developers can build and deploy enterprise AI solutions around these data points quickly – even if they have limited machine learning expertise.

“Many of our enterprise customers are struggling to unlock the hidden value in their data. One reason it’s hard is because much of the data is unstructured. KaaS allows them to see the value quicker than any other off-the-shelf solution on the market,” said Sudharsan Rangarajan, vice president of engineering (AI & Data CoE) at Publicis Sapient.

KaaS services are designed in such a way that you can use only the relevant services for your business problem. These services can be orchestrated to build scalable, more complex enterprise-grade solutions like:

Chatbots – Support queries from customers, internal staff and research analysts – with minimal training/tuning effort
Virtual assistants – Improve customer service while lowering costs
Research platforms – Help Analysts make sense of a large number of documents
Search engines – Quickly find documents based on relevance and summarize key points
Company profiles – Provide market intelligence (customers, problems, competition, growth potential, etc.)
Opinion synopses – Explain public opinion trends and drivers behind them

From business standpoint, our customers see benefits on multiple fronts:

Time to market: A typical unstructured data mining project will save between 25 and 60 percent in development time and effort.
Data privacy and security: Since the platform can be hosted on-premise, our clients can retain full-control of the raw data.
Reduced training time: Typical mining solutions require extensive training to be useful. KaaS solutions use transfer learning to get reasonable accuracy with minimal effort.
Scalability: The state-of-the-art KaaS platform is containerized using Docker and can be deployed on-premises or on different cloud platforms.

The KaaS portfolio of services makes sense of natural and unstructured language so it can be structured and operationalized. Our tools can resolve ambiguous terms, identify intended meanings (when they’re different from the literal meanings), evaluate the emotional state of the speaker or author, categorize phrases and concepts and recognize semantic relationships.

With this thorough understanding of a document or website’s contents, KaaS trains chatbots and search engines to engage with this language as close as possible to the way a human would – with an ear for nuance and context. This facilitates a more natural and fluid interaction with artificial intelligence solutions and improves workflow accuracy.

“Any time you need to extract anything meaningful from the text, KaaS is useful. It consists of various services that you can use to build an application”, said Mayank Saini, data scientist and engineering lead of KaaS at Publicis Sapient.

Few of the KaaS services that are part of the offering are:

Machine comprehension
Document tagger
Entity disambiguation
Paraphrase detection
Section extraction
Relations extraction
Natural language to structured query language
Coreference resolution
Semantic ranking of answers
Active learning
Text classification
Question understanding
Sentiment analysis

KaaS removes the complexity of machine learning so its services can be integrated into larger application architectures easily – accelerating the time to value. Since the services can be deployed over the cloud or locally, businesses can protect their data and ensure they comply with regulations or policies.