Add the intersection of NLU, LLMs and natural language data

RAG Evaluation

Retrieval Augmented Generation (RAG) is a very popular framework or class of LLM Application. The basic principle of RAG is to leverage external data sources to give LLMs contextual reference. In the recent past, I wrote much on different RAG approaches and pipelines. But how can we evaluate, measure and quantify the performance of a RAG pipeline?

COBUS GREYLING
5 min read

Follow-Up Interview With Cobus

The following is a redacted interview between Greg (CEO @ HumanFirst) and Cobus Greyling (www.cobusgreyling.me); in this interview, we dig deep into the intersection of NLU and voice / ASR technologies.

GREGORY WHITESIDE
9 MIN READ

Practical semantic search and clustering

Semantic similarity is quickly becoming a mainstream technique for working with natural language data; embedding new examples into a latent space uncovers the relationship between them. PineCone.io (among others) are doing a great job of producing content that describes how semantic similarity works under the hood.

GREGORY WHITESIDE
6 MIN READ

Interview with Cobus Greyling

[Greg] HumanFirst is a big fan of yours Cobus :) I've personally been reading your blog for the last two years, and have learned a lot about everything happening in Conversational AI, NLU and Call Center AI thanks to your deep-dives on existing solutions and prolific articles covering a wide range of topics.

GREGORY WHITESIDE
7 MIN READ

Are we focusing too much on the labeling problem?

Programatic labeling has been a hot area of development in the last 2 years, with Snorkel Flow leading the way, and other commercial and open-source projects appearing in this space.

GREGORY WHITESIDE
6 MIN READ

Voice of the Customer Data is 🚀

Most businesses are sitting on high-quality Voice of the Customer Data , generated across a growing number of channels in the context of sales, marketing, product and operations.

GREGORY WHITESIDE
4 MIN READ

GPT-3 in Conversational AI

The first iteration of HumanFirst was a browser extension that ingested customer support live chat conversations, trained a black-box AI model that predicted what the chat agent’s next response should be, and provided auto-complete suggestions to the agents to speed up their replies.

GREGORY WHITESIDE
5 MIN READ

Help Centers are Underrated in Conversational AI.

> In this article, I use the term "Help Center" to refer to all types of Help Center, Knowledge-base and FAQ solutions. Help Centers are rarely a core part of the

GREGORY WHITESIDE
4 MIN READ

Labeling isn't hard if you know what to label.

Companies like Scale.ai [scale.com] (recently valued at $7B+ dollars [https://fortune.comscale-ai-valuation-new-funding-fundraising-data-labeling-company-startups-

GREGORY WHITESIDE
2 MIN READ

Conversational AI should start with search.

Let’s take an airline example: “I want to change my flight from the 3rd to the 4th”.

GREGORY WHITESIDE
3 MIN READ

Walking before running in conversational AI.

Conversational AI is often equated with chatbots. Yet chatbots rarely do the “AI” part correctly (i.e: understanding user intents).

GREGORY WHITESIDE
2 MIN READ
Articles

Building a Semantic Emoji Prediction NLU

Learn one way to approach semantic Emoji prediction with cascading intents.

GREGORY WHITESIDE
6 MIN READ
Articles

The crux of NLU is managing the long-tail of data.

> Until now, building and maintaining the training data needed to support a long-tail of intents has been a difficult problem — yet this is what is needed to go from a Gregory Whiteside.

GREGORY WHITESIDE
2 MIN READ