Messaging Analytics

Developing AI with EQ

To build more human-like and engaging AI products, developers will need tools not available in the public domain. The current generation of chatbots and LLMs may have 200+ IQ, but 0 EQ or emotional intelligence. Most have two major shortfalls: (1) They speak the same way to everyone, lacking the insights and adaptability of humans during a conversation. (2) They are robotic, because most current-generation foundation models trained from material published on the internet or customer service chats, or worse, forums & social media. This is not how people, especially those with an existing relationship, actually chat with each other.

We have the analytics and data to address both these problems to improve EQ and sophistication of the next generation of AI.

EQ: Understanding The Human On The Other Side

Having insight about the user is critical for your bot to use language that can resonate and create an emotional connection with users . Humans have an advantage over current chatbots because they can, for example, know when they’re talking to a 15-year-old vs. a 50-year-old and adapt their language accordingly. We can equip your bots with the only known AI models that can predict insights about your chatbot users, like demographics, personality, preferences, etc. based on their words.

Insights & Analytics

Chat and messaging conversations are more than just a string of words. Every word choice, typo, phrase, emoji or pause in a conversation reveals insight about the person that could be used by your chatbot to modify its behavior. As people, we’re limited in our ability to simultaneously listen, remember and analyze every datapoint in a way machines and AI can. This limits our ability to fully capture everything that’s communicated in a conversation.

Our analytics engine can generate every datapoint imaginable from a messaging conversation. Coupled with the millions of conversations and labels from users, we built AI to predict hundreds of insights just from a small snippet of a given conversation. These insights include demographics (e.g. age/sex), relationship type (e.g. romantic/ friendly/professional), relationship (e.g. a parent vs. a sibling), moods, interests and personalities (trusting vs. guarded).

These insights will be important in building chatbots and conversational AI with human-level intelligence. Most models have never trained from actual human-to-human chat data, so they have no idea how people ACTUALLY talk to each other. Anyone looking to work on this should contact us at info@textmei.com. There’s no point in reinventing the wheel. Here’s just a small sample of analytics/AI engine that underlies our products.

Each message has metadata that can be more revealing than the words themselves. This can be studied over time to establish a baseline pattern, identify trends and even pinpoint the day when things changed.