Announcing NLX Boost™

Today, NLX is announcing NLX Boost™, a Large Language Model (LLM)-powered feature that can improve traditional NLP (like Amazon Lex or Google Dialogflow) performance by up to 90%.

Andrei Papancea


Today, NLX is announcing NLX Boost™, a Large Language Model (LLM)-powered feature that can improve traditional NLP (like Amazon Lex or Google Dialogflow) performance by up to 90%.

NLX Boost evaluates and prioritizes the semantics of a customer’s utterance to improve intent detection accuracy. By doing so it “boosts” the performance of the underlying NLP model by leveraging the patterns and commonalities between different ways customers may express the same intent. 

Example: I want to cancel my flight / I have a canceled flight

In real-world applications, customer input can often be noisy. Utterances contain typos, abbreviations, informal and nuanced language, and so on. Unless a customer’s training data is specifically structured to pick up on the differences between the two utterances in the above example, the NLP will more often than not fail to route the request to the correct intent. These small differences in language are incredibly common and are one of the big reasons so much effort is applied to tuning training data. Unlike traditional NLPs, NLX Boost focuses on the underlying meaning to pick up on these nuances rather than matching of words/phrases with minor deviations to classify intent more accurately and effectively.

“With NLX Boost, our customers will spend much less time fine-tuning their training data without having to sacrifice performance. In fact, they’re likely to see gains. This is a particularly big deal for our enterprise customers that are operating at scale, managing 100s of intents, across many languages, as many of them do. We’re giving them a bunch of time back to focus on building and automation instead of tuning training data.” - Andrei Papancea, CEO & Chief Product Officer, NLX

Managing training data to drive the most performance out of traditional NLP is a laborious and never ending task. Not only does training data for one intent impact performance for the rest of the model, you are constantly re-training the NLP based on out-of-date data to try and keep pace with customer preferences and trends. NLX Boost dramatically reduces this headache without adding any burden to the customer. For larger enterprise teams, this can have a material positive impact on costs associated with training and tuning the NLP model. 

“We’ve made it really easy for customers to use NLX Boost. It’s enabled with the flip of a switch and runs in the background with no meaningful impact to response time during live conversations. We’ve been piloting it with a couple of our largest customers and they have seen significant intent detection accuracy gains.” - Sam Trost, Director of Engineering, NLX

NLX Boost is available starting today and it works across all channels (chat, voice, multimodal) and languages.

Interested in learning more? Get in touch.

Andrei Papancea

Andrei is our CEO and swiss-army knife for all things natural language-related.

He built the Natural Language Understanding platform for American Express, processing millions of conversations across AmEx’s main servicing channels.

As Director of Engineering, he deployed AWS across the business units of Argo Group, a publicly traded US company, and successfully passed the implementation through a technical audit (30+ AWS accounts managed).

He teaches graduate lectures on Cloud Computing and Big Data at Columbia University.

He holds a M.S. in Computer Science from Columbia University.