AI, Machine Learning, NLP, & Extract
During a recent phone call, I was asked if Extract does NLP (natural language processing). My answer was that Extract’s methods, rules and machine learning, and NLP are similar in that both are about finding information in data and then doing something with it.
NLP VS EXTRACT
The important difference is that Extract is good at finding exactly what our customers are looking for and NLP is good at finding things you may not have known existed in the data set. Extract is good at finding the very specific kinds of data & details, like names, that are required for a task and NLP looks for trends & concepts in order to make predictions… like whether there is a geography / cancer relationship. Both are good at learning as they process documents but the learning is done differently.
artificial intelligence and machine learning
Major factors involved in NLP and in Extract’s process is AI (artificial intelligence) and ML (machine learning). As with NLP, Extract uses machine learning to improve the time it takes for Extract to write rules that will produce great results. Extract relies on rules because we have found it is the only way to get the accuracy rates that our customers demand to protect social security numbers or to find discrete data in a patient’s laboratory test result. Extract makes very good use of ML to achieve its customer’s accuracy requirements more effectively.
NLP has the image of being something new. Early experiments started in 1954 and involved the fully automatic translation of sixty sentences from Russian into English. The authors of this experiment predicted machine translation would be solved in five years. Well, not so fast.
The goal of NLP is to do away with computer programming languages, and what would be left is simply the “human language.” Fact is, we are sixty years into the NLP effort and while we are seeing great benefits from AI, we are still a long, long way from a ubiquitous human language.
On March 7, 2107 the Wall Street Journal printed an interview with Andrew Ng (chief scientist at Baidu) and Neil Jacobstein (chair of AI and robotics department at Singularity University), and both gentlemen predict great things for AI. Ng suggests the changes will be as dramatic as the changes electrification brought to industry a hundred years ago. Giant companies like Baidu are investing heavily in speech recognition and are seeing regular use doubling year over year. Jacobstein says AI is already outpacing human thinking in many areas – as an example the diagnosis of pulmonary hypertension (80% accuracy vs 60%).
