Chatterbox Cognitive Computing – Machine Learning Platform
‘Current’ IT systems are structured and binary in nature; the interaction with these systems is bound by the user interface and inherent process to understand meaning from the data. The application is often siloed and limited in its inherent usefulness by the data sources it has access to. New cognitive technologies; based on machine learning and natural language processing; augment, enhance and ultimately over-shadow these systems by being more ‘human-like”.
Around 80% of all data is unstructured; and often unavailable to these ‘current’ IT systems. Within an organisation this unstructured data can live in email, collaboration solutions, social data and beyond. Cognitive technologies can ‘teach’ an understanding and meaning in the natural language expressed within a document, email, metadata, short-form, long-form – machines can be taught meaning, context and therefore produce insight from a lack of structure. A statistical natural language processing approach sees a meaning being assigned to data in a manual annotation phase (in hours / days) and then the machine with its inherent intelligence continues to evolve and enhance its accuracy of information classification. When multiple smart classifiers are trained and aligned to filter out multiple levels of understanding a highly disruptive product emerges; with a highly reliable level of accuracy and with a speed of execution no human could deliver.
The machine classifier actually starts to ‘understand’ natural language, the emphasis of language, intent and sentiment expressed, and ways of reasoning and problem-solving. Cognitive technologies have widespread potential for adoption across multiple industries and enterprises automating tasks inherently ‘owned’ and delivered by humans. Activities and work can be mapped into a cognitive application that can work 24/7; processing data at speed and scale and with an incredible level of accuracy.
On a recent ‘use’ case at Chatterbox Labs we set about a large data set to train our machine classifiers to recognise patterns and provide detailed matching across varied descriptions of data. Through using our machine learning platform no coding was required to create specific classifiers seeking to understand the data before it. A manual annotation process managed within our platform gave the machine classifiers a guide as to what to look for and what ‘meaning’ there was to be found. The data was then introduced through the assembled product at scale and in minutes results with f-scores (that is, the combination of precision and recall) performing at a minimum of 0.8, through to a one off peak of 0.99.
Cognitive technologies are being deployed across healthcare to improve patient diagnostics; in the insurance industry to automate the claims and assessment processes; for car manufacturers to guide their engineers in complex activities; and we are only at the beginning of what can be automated. Unstructured data is being given meaning; and a ‘structure’ that provides clear insight for a data driven industry. Cognitive technologies and especially statistical driven natural language processing (NLP) is going to revolutionise and disrupt every industry sector.
If you have a ‘use case’ for a cognitive computing product to automate ‘activities’ in your enterprise and don’t want to build a product from scratch then please get in touch with our VP Strategic Alliances, Andrew Watson, Andrew@chatterbox.co or check out our web site www.chatterbox.co