Building a chatbot in 2018 and building one today
From Alexa intents and flows to agents built on language models: what has changed in building a chatbot or voice assistant, and what hasn't.
Nieves Ábalos
Co-founder at Monoceros Labs
Nieves Ábalos is co-founder of Monoceros Labs, where she leads conversation design, voice assistant and Spanish synthetic voice projects. She holds a degree in Computer Engineering and a master's in Software Development from the University of Granada, where she began researching in 2009 with the spoken and multimodal dialogue systems group, publishing on dialogue system evaluation and ambient intelligence. Since 2023 she has been doing a PhD in Artificial Intelligence at the University of Granada on language models and theories of human communication.
Before Monoceros she was a researcher and product manager at BEEVA (BBVA group), where she led natural language processing and conversational interface research and the company's intelligent agent platform. Amazon recognized her as an Alexa Champion. In 2018 she founded the Spanish chapter of Women in Voice, the first in Spanish, and in 2021 received its Chapter Pioneer Award.
She has been teaching since 2011. Today she teaches artificial intelligence on UNIR's master's program and speech synthesis on the Language and Artificial Intelligence master's at the Autonomous University of Madrid. She co-directed KSchool's Deep Learning and AI master's and, with Carlos Muñoz-Romero, designed the first voice interface design workshop taught in Spain (2019). She co-authored the RTVE-UGR Chair paper on speech synthesis and news verification (Applied Sciences, 2024).
Posts by Nieves Ábalos
Evaluating an agent before your customers use it
How to evaluate an AI assistant or agent before launch: what to measure, test conversations, a model as judge, attacks and testing with real people.
An assistant repeats the prejudices in its data
Where the biases in AI assistants and agents come from, how they show up in a conversation or a voice, and what you can measure to reduce them.
In a newsroom with agents, who checks the work?
Newsrooms are automating more and more and moving toward AI agents. Why checking everything by hand doesn't scale, and why you need an expert in the loop.
What it takes to call a chatbot an agent
What an AI agent is and how it differs from a chatbot: tools, memory, planning and autonomy, plus questions to find out what you're actually being offered.
Is it my turn to speak?
Why voice agents talk over people or go quiet: how humans manage turn-taking and how machines try to detect it.
What changes when your app learns to talk
Adding an AI chat to an app looks easy. The ways to build a conversational app and the challenges that remain: context, errors, testing and cost.
What to ask yourself before building an assistant
Before choosing technology for a chatbot or voice assistant: which conversations you want to have, where, through which channel, how you'll measure success and what to prototype.
What reads well doesn't always sound good
Text written for reading usually sounds wrong in a voice assistant. Six adjustments for writing for the ear: sentences, options, punctuation, numbers and pauses.
Technology for turning text into speech
Want to know which tools to use to bring your written content to life? Here are a few keys.
12 books to learn about voice interfaces
There's no better way to learn about voice interfaces than a good read. So we've put together a first list of reference books on speech technology and conversation design, grouped by key topic so you can choose according to your profile or role in building VUIs.
8 tips for building good Alexa skills in Spanish
Our design tips for building good Alexa skills (voice apps for Amazon Alexa) in Spanish… and getting them certified quickly.
Reflections: the problem of loneliness
Loneliness is a serious problem in aging, technologically hyperconnected societies. But technology can also help us feel less isolated and transport us to other places.