NLP: Natural Language Processing

Natural Language Processing was one of the earliest hallmarks of intelligence that computer scientists were grappling with even in the 1950’s. While there were some hopeful early somewhat successful research projects, such as for machine translation between two languages, the challenge proved to be immense; and along with AI research in general, NLP development slowed down.

Part of the early hurdle was that complex sets of manual hand written rules were developed for each given NLP system- a painstaking process when you consider that we spend our whole lives learning the rules of languages yet still can’t figure out if its there, their or they’re! To make matters worse, languages are living rules; while most of them stay the same, periodically they change, add new words, etc. How can we create NLP models which learn from language as it is, and also adopt new rules on the fly?

In the late 1980’s, along with the development of AI/Neural Networks in general, machine learning algorithms started to get used for NLP. Moore’s law had also progressed enough to provide for significantly greater computation power; as the Internet grew it was easier to get larger datasets of text to learn from; and the computational statistical component of machine learning enabled a NLP model to derive statistical rules rather than hardcoded facts.

Soon entered a proliferation of chatbots that could do NLP well enough, even to serve commercial purposes. The recent advance of LLM’s seemingly has put the nail in the coffin of the rules based NLP models, but as anyone who has played with ChatGPT or the like for a few minutes knows, there is still a long way to go to achieving NLP models that can not only learn statistical patterns of communication, but understand the meaning or nuance of what they are saying.

All resources are chosen by Data Mine staff to be of decent quality, and most if not all content is free.