A.I. and Machine Learning Terms for Quality and Compliance Teams

If you are a quality and compliance professional, you certainly know the ins and outs of all the GXPs. But as new technologies become available, do you speak the language of artificial intelligence (A.I.) and machine learning?

On Aug. 3, Redica Systems Senior GMP Quality Expert Jerry Chapman discussed the various tools he uses when he works on structuring Redica Systems data in the webinar, “A.I. for Quality and Compliance Teams.” Chapman has used his extensive GMP expertise to add additional intelligence to Redica Systems datasets since 2019.

[Related: For more from Jerry Chapman, read his compilation of five FDA GMP inspection case studies presented by FDA ORA’s Ileana Barreto-Pettit at a recent industry conference.]

The following are some of the machine learning terms he presented in the webinar. 

• Text Parsing – A computer algorithm cannot truly “read” human sentences but it can read words and groups of words. Parsing splits sequences of words based on set rules. For instance, some of the richest content in a Warning Letter often follows the words “specifically” or “for example.” A parser can pull the information that follows these words, potentially gathering some critical enforcement information.
• Tokenization – Refers to a sentence split up into individual “tokens,” the simplest of which are the words that make up a sentence.
• “Parts of speech” (POS) tagging – For anyone who has diagrammed sentences, this concept will be familiar. POS tagging trains a computer to understand the structure of human speech (e.g., “subject, verb, object” in English).
• Stemming – This is the process of reducing each word in a document to its base, or “stem.” The words “consigned,” “consigning,” and “consignment” have a stem of “consign.”
• N-grams – This refers to a contiguous sequence of n items from a given text sample and are created by subject matter experts based on their experience and then tested over time in iterations.

To learn how Chapman uses these various aspects of machine learning to build up Redica Systems data, view the webinar. He also provided two case studies showing how quality and compliance teams can use data, one addressed recent 503B outsourcing facility FDA Warning Letter citations, and the other covered data integrity trends “hiding” in FDA Warning Letters.