The emergence of Large Language Models has led to many breakthroughs in processing text and producing new and innovative products. Generative text is one of many applications in the Natural Language Understanding space. Text analysis, vectorization, and word-embeddings provide another way to analyze text to help drive business decisions in scientifically repeatable ways. The vectorization approach can be used to explore datasets like product reviews, open ended survey questions and trouble tickets to understand the context of a large set of text data in seconds. A language model is created for each dataset (a targeted language model) which can then be used to explore a specific set of text data. Datasets might include a hundred thousand or million product reviews or an open-ended survey question answered by fifty thousand employees to understand what is going on with customers or employees at a very deep level. Large language models and vectorization use different approaches to understand and interpret text. While large language models like GPT-3 transform massive amounts of text into a single neural network for generating text, the vectorization or word embeddings approach can be used on small or large datasets in a repeatable and consistent manner, allowing for more reliable data analysis of a specific set of data. Come and learn about natural language analysis and ways to process text in new and innovative ways.

This meeting will be held in person at the Knobbe Martens' Irvine offices (2040 Main St # 14 Irvine, CA). The doors will open at 6:30 PM for networking. Announcements and the presentation will begin at 7:00 PM. Parking validation will be provided to attendees.

Speaker:


Wood Harter

William Wood Harter is a Chapman University Alumnus and Adjunct professor who has worked in Aerospace, Security, Gaming, and now Artificial Intelligence and Natural Language Analysis. He currently works as a Solutions Architect for Luminoso, an award-winning leader in tools for understanding language around product reviews, open-ended survey questions, trouble tickets, and help desk transcriptions.