4 Steps to Building Intelligent AI ModelsBy Innodata Inc. on August 1, 2019
As commercial enterprises embrace technology to become more agile and competitive, many are turning to artificial intelligence (AI) and machine learning to increase operational efficiency, accelerate innovation and make faster, more informed business decisions. Not surprisingly, the market for AI is expected to exceed more than $191 billion by 2024 at a CAGR of 37%. Despite the promise of AI to completely transform business, many challenges remain.
According to a report by MIT Technology Review, insufficient data quality was one of the biggest challenges to employing AI. What’s more, 85% of AI projects will “not deliver” for organizations, according to Gartner. Ironically, data itself is often the biggest obstacles to data transformation.
“You can’t feed the algorithms if you don’t have data. Solid, clean data in large volumes, well-tagged and well organized is crucial,” that according to comments from the Chief Data Officer at the Department of Defense, Michael Conlin.
Without access to clean, accurate and useable data, machine learning models don’t have a very good foundation to learn from. After all, AI is only as smart as the data it consumes.
While AI and machine learning models can be trained to classify input patterns or predict tasks, objects, or functions, the scarcity of semantically enriched data – structured or unstructured – presents a serious challenge for data scientists and enterprises leaning on AI and machine learning to achieve any real value. In order to produce actionable and effective insights, AI and machine learning models require clean, accurate and useable data. While data is often referred to as the “new oil”, expertly labelled and annotated data is actually an organization’s most precious commodity.
4 Fundamental Requirements for Building AI Applications
Over the past decade, Innodata has spent significant time helping our customers in this area. As a result, we’ve had the opportunity to uncover some of the key requirements needed in building an effective AI application.
1. Raw Data
Having access to the right raw data set has proven to be critical factor in piloting an AI project. Raw data is information that has typically not been processed or analyzed and is routinely considered inoperable. But deeper analysis can yield opportunities to turn raw data into useful insight. For example, one of our clients was looking to understand key challenges associated with their customers self-serve system and looking to improve the customer experience. After a thorough introspection of all the shared data, we honed in customer call center transcripts as a way to understand the trends and train their AI models.
Ontologies play a critical role in machine learning. According to the Wikipedia definition, ontologies are “formal naming and definition of the types, properties, and interrelationships of the entities that really or fundamentally exist for a particular domain of discourse.” In other words, ontologies give meaning to things.
Think of this as teaching your AI to communicate using a common language. It is critical to identify the problem statement and understand how AI can interpret data to semantically solve a certain use case. The need for out-of-box ontologies or availability of client ontologies that can be used as the basis to form the data labeling is critical.
Annotation (also known as data labeling) is quite critical to ensuring your AI and machine learning projects can scale. It provides that initial setup for training a machine learning model with what it needs to understand and how to discriminate against various inputs to come up with accurate outputs. There are many different types of data annotation, depending on what kind of form the data is in.
It can range from image and video annotation, text categorization, semantic annotation, and content categorization. Humans are needed to identify and annotate specific data so machines can learn to identify and classy information. Without these labels, the machine learning algorithm will have a difficult time computing the necessary attributes. How data is annotated and labeled brings us to our next and most crucial requirement: subject matter expertise.
4. Subject Matter Expertise and Supervised Learning
Our clients have learned how important it is to have subject matter experts (SME) that understand their specific industry and complex needs. This goes back to the need for annotated data. If there are even slight errors in the data or in the training sets used to create predictive models, the consequences can be potentially catastrophic. That’s why the need for specific domain expertise is so crucial, and why human knowledge still plays a pivotal role in artificial intelligence.
For example, being able to interpret complex legal obligations and agreements from ISDA contracts require legal specialists that can identify and label the most appropriate information. The same goes for other fields like science and medicine where deep understanding and fluency of the content cannot be taken for granted.