Natural Language Processing (NLP) and Semantic Analysis
A Different Hypothesis In Numerical Modeling and Machine Learning Systems
In Natural language processing (NLP), Semantic Analysis tries to address the meaning of natural language. It attempts to capture a given text’s meaning while including context, entities, logical structuring of sentences, and governing grammar roles.
The current state-of-the-art NLP systems have come a long way and are getting more complex with astounding results. Still, they all stop short of truly understanding the meaning of a given text like the human brain.
So the question is, can we ever get close to the way our brain works using current techniques and ML models? or do we need a fundamentally different approach to ML altogether?
Here is my hypothesis! Be it not a fundamentally new approach but somewhat different. Let’s assume that an AI system is an equivalent algorithm or process to find a solution to an equation numerically. In our case, the equation is Natural Language, the X-axis is semantic definitions, and Y-axis is syntactic rules and concepts.
Before solving this equation, let’s plot it to find the boundaries and initial values. Looking closer at X-axis interceptions, we can see that these boundaries are, in fact, the semantic definitions of a given text applied through this equation! And that is the most interesting, high-level pre-processing of the data before committing any ML models.
Once these boundaries are defined and solved, the rest is business as usual but focused, targeted, and reduced in scope. A simple, straightforward approach in pre-processing before calling the NLP models, right?
The good news is that we at PDRVC have found a way to define these boundaries computationally, hence the semantic definition of a given text, following the hypothesis mentioned above and introducing a new set of algorithms that we call Semantic Boundaries Identifiers (SBI in short).
For us, this has changed quite a few things in setting up, processing, and analyzing textual data through the NLP pipeline. Not exactly how our brain works but at least following some higher-level top-down concepts here.
If Semantic NLP is your focus and anything above even remotely makes sense, I love to hear your thoughts and experience, so please send me a message and get the conversation going…
