Email-based Intelligent Virtual Assistant (EIVA)

Email-based Intelligent Virtual Assistant for Scheduling Presented by Anand Chowdhary
Supervised by Dr. Job Zwiers Creative Technology BSc University of Twente Enschede, the Netherlands

is a creative technologist and entrepreneur. Apart from studying CreaTe,
he also co-founded Oswald Labs, an accessibility technology startup, for which he was featured in Forbes and Het Financieele Dagblad lists. He also studied at Santa Clara University as part of the Global Engineering Exchange program, and Bachelor Honors in Science at the UT. Email-based Intelligent Virtual Assistant for Scheduling Presented by Anand Chowdhary Supervised by Dr. Job Zwiers Creative Technology BSc University of Twente Enschede, the Netherlands

Supervised by Dr. Job Zwiers Creative Technology BSc University of Twente Enschede, the Netherlands EIVA

Abstract Manually setting up appointments by email wastes tens of
hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and 8 out of 10 USE EMAIL FOR SCHEDULING

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and 8 out of 10 USE EMAIL FOR SCHEDULING 6 out of 10 MEETINGS ARE RESCHEDULED

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and 8 out of 10 USE EMAIL FOR SCHEDULING 6 out of 10 MEETINGS ARE RESCHEDULED €150 per month IN LOST TIME WHILE SCHEDULING

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 9 out of 10 ASK ASSISTANT TO SCHEDULE

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 9 out of 10 ASK ASSISTANT TO SCHEDULE 8 out of 10 AI ASSISTANT FOR SCHEDULING

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 9 out of 10 ASK ASSISTANT TO SCHEDULE 8 out of 10 AI ASSISTANT FOR SCHEDULING THE SOLUTION?

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with

every month for professionals, because several email exchanges are required
before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood

before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood Florian Overkamp Founder of Speakup B.V.

before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood Idea Research Development External testing Analysis

before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood

before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood SCREEN SHARING

and industry best-practices with the support of Speakup. Users can
simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Natural Language Processing

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Natural Language Processing Language Detection & Variable Extraction

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Natural Language Processing Language Detection & Variable Extraction Intent Classiﬁcation

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Natural Language Processing Language Detection & Variable Extraction Intent Classiﬁcation Date Parsing

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Natural Language Processing Language Detection & Variable Extraction Intent Classiﬁcation Date Parsing Google Cloud Natural Language API

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Natural Language Processing Language Detection & Variable Extraction Intent Classiﬁcation Date Parsing Google Cloud Natural Language API Node Natural with Naive Bayes Classiﬁer natural

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Natural Language Processing Language Detection & Variable Extraction Intent Classification Date Parsing Google Cloud Natural Language API Node Natural with Naive Bayes Classifier natural Chrono Node with custom filters Chrono

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Natural Language Processing Intent Classiﬁcation Node Natural with Naive Bayes Classiﬁer natural

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Intent Classiﬁcation import { BayesClassifier } from "natural"; const classifier = new BayesClassifier(); classifier.addDocument( [ "set up an appointment", "schedule a call", "meet me for dinner", ], "setupNewAppointment" ); classifier.addDocument( [ "i can’t make it", "reschedule this call", "find another time", ], "rescheduleAppointment" ); classifier.train();

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Intent Classiﬁcation classifier.getClassifications( "i don’t think I’ll be able to make it to this time" ); // Classifications: // // setupNewAppointment: 0.1666 // rescheduleAppointment: 0.3333 // This never-before-seen sentence is correctly classiﬁed as ‘Reschedule Appointment’ This is an example with a tiny training dataset

simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared Conﬁguration Management Transactional Emails Authentication Rate Limits MariaDB Redis ElasticSearch Simple Storage Service Public RESTful API Scheduled Jobs Third-party APIs Vue.js Nuxt.js Authorization in Fetching Unit Tests Continuous Integration Initialization Tests Uptime & Error Monitoring Sentry Resource Monitoring Caprover GitHub Actions Netlify Deployments

scheduling preferences. A companion web application is also developed to
manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

preferences. A companion web application is also developed to manage
meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. PROFESSIONALS OF PARTICIPANTS 18 employees at companies or organizations 9 students at schools or universities 3 self-employed professionals or freelancers

meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. 19,177 API requests 98 Emails processed 2,244 Pageviews

Germany Finland France Russia 12 Countries India Nederland Poland Kenya
United States United Kingdom Iran Indonesia

7 Provinces Flevoland Overijssel Gelderland Utrecht S. Holland N. Holland
Friesland

meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. Firefox Chrome Safari Safari Mobile Edge Chrome-related macOS Linux Windows Android iOS Ubuntu 992 529 377 217 83 44 772 729 637 59 3 34

meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. WEB APP EXPERIENCE 4.4 5.0 5.0 MEDIAN MODE MEAN WEB APP DESIGN 4.6 5.0 5.0 MEDIAN MODE MEAN

meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. WEB APP EXPERIENCE 4.4 5.0 5.0 MEDIAN MODE MEAN 4.6 5.0 5.0 MEDIAN MODE MEAN PRIVACY FEATURES 4.7 5.0 5.0 MEDIAN MODE MEAN WEB APP DESIGN

meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. ASSISTANT EXPERIENCE 4.5 5.0 5.0 MEDIAN MODE MEAN UNDERSTANDING 4.4 5.0 5.0 MEDIAN MODE MEAN

meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. TIME RECOMMENDATIONS 3.9 4.0 4.0 MEDIAN MODE MEAN TRUST 3.6 4.0 4.0 MEDIAN MODE MEAN

meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

the average rating of the overall assistant and app experience
of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. 29 out of 30 EIVA MET EXPECTATIONS

of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. 29 out of 30 EIVA MET EXPECTATIONS 25 out of 30 WOULD USE EIVA SERVICE

of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. 29 out of 30 EIVA MET EXPECTATIONS 25 out of 30 WOULD USE EIVA SERVICE 24 out of 30 WOULD PAY AT LEAST UP TO €5

of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

tracking. All but one participants said that the EIVA met
their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. WHAT DO YOU LIKE THE MOST? 5 × Ease of use 3 × Fast response time 1 × Overall concept 1 × Security features 1 × Conversational interface 1 × Web app user interface

their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. WHAT DO YOU LIKE THE MOST? 5 × Ease of use 3 × Fast response time 1 × Overall concept 1 × Security features 1 × Conversational interface 1 × Web app user interface WHAT DO YOU LIKE THE LEAST? 1 × Pretending to be human 1 × Lengthy app onboarding 1 × Mobile user interface

their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers. WHAT DO YOU LIKE THE MOST? 5 × Ease of use 3 × Fast response time 1 × Overall concept 1 × Security features 1 × Conversational interface 1 × Web app user interface WHAT DO YOU LIKE THE LEAST? 1 × Pretending to be human 1 × Lengthy app onboarding 1 × Mobile user interface FEATURE RECOMMENDATIONS Meeting without emails SMS and WhatsApp support Multiple guests in a meeting Manual time slot selection Cancel or reschedule UI

their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

and 4.5 out of 5 respectively. Users’ behavior was also
understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

A user experience evaluation of both the EIVA and its
app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

hours every month for professionals, because several email exchanges are required before receiving conﬁrmation. Since not everyone can afford to hire full-time assistants, Email-based Intelligent Virtual Assistants (EIVA) can help by automating this task. In this paper, a functional EIVA is developed based on research and industry best-practices with the support of Speakup. Users can simply add their assistant’s address as ‘CC’ in an email, and EIVA will share the recommended location and time slots with guests, based on the user’s availability (using their calendar) and scheduling preferences. A companion web application is also developed to manage meetings and settings. The code is open source and written in TypeScript with Node.js and Vue.js and deployed on Amazon Web Services architecture in Europe. The product is built with focus on data privacy and user personalization. A user experience evaluation of both the EIVA and its app with 30 participants was conducted that found positive reception, with the average rating of the overall assistant and app experience of 4.4 and 4.5 out of 5 respectively. Users’ behavior was also understood with heatmaps and visualizations using pageview and mouse clicks tracking. All but one participants said that the EIVA met their expectations, and 25 out of 30 would use it in the future if it launches as a service. Most would also be willing to pay for it, with an average amount up to €6.16 per month. Participants also shared their frustrations and recommendations for future work. In the future, NLP classiﬁcation should be improved and user recommendations should be implemented before launching the EIVA service for consumers.

Supervised by Dr. Job Zwiers Creative Technology BSc University of Twente Enschede, the Netherlands NOMINEE CREATIVE TECHNOLOGY BACHELOR AWARD UNIVERSITY OF TWENTE.

How does the Naive Bayes Classifier actually work? What’s the
author’s previous work, Ara? How did you collect data? How did you design the questions? How did you analyze the collected data? How did you capture stakeholders’ requirements? What questions did you explore using personas? How do you deal with gender bias in assistants? Do email recipients know EIVA is not human? How do you parse dates in emails? What authentications standards does the app use? Do you implement rate limits for security? How do you use scheduled jobs? How do you send client-side fetch requests? How do you manage state in the frontend app? Does EIVA work in both English and Dutch? How do you test the source code? How does the CI/CD pipeline work? How does you test external services? How much was the infrastructure cost? Do you monitor the service’s uptime? Do you monitor server resource usage? How do you know if an error occurs? Do people actually check their email or send emails? What calendaring services do people use? How many API requests or emails were logged? What were people’s first-impressions? How did people like the web app? How did people like the assistant over email? How much would people be willing to pay for this? What were common user frustrations? What were common feature recommendations? How would you improve this in the future? What about AI-driven job loss? What’s the business case for this service? How are the database tables structured? Where can I find the code for this?

How does the Naive Bayes Classiﬁer actually work? ← Back

What’s the author’s previous work, Ara? ← Back

How did you collect data? ← Back

How did you design the questions? ← Back

How did you analyze the collected data? ← Back

How did you capture stakeholders’ requirements? ← Back

What questions did you explore using personas? ← Back

How do you deal with gender bias in assistants? ←
Back

Do email recipients know EIVA is not human? ← Back

How do you parse dates in emails? ← Back

What authentications standards does the app use? ← Back

Do you implement rate limits for security? ← Back

How do you use scheduled jobs? ← Back

How do you send client-side fetch requests? ← Back

How do you manage state in the frontend app? ←
Back

Does EIVA work in both English and Dutch? ← Back

How do you test the source code? ← Back

How does the CI/CD pipeline work? ← Back

How does you test external services? ← Back

How much was the infrastructure cost? ← Back

Do you monitor the service’s uptime? ← Back

Do you monitor server resource usage? ← Back

How do you know if an error occurs? ← Back

Do people actually check their email or send emails? ←
Back

What calendaring services do people use? ← Back

How many API requests or emails were logged? ← Back

What were people’s ﬁrst-impressions? ← Back

How did people like the web app? ← Back

How did people like the assistant over email? ← Back

How much would people be willing to pay for this?
← Back

What were common user frustrations? ← Back

What were common feature recommendations? ← Back

How would you improve this in the future? ← Back

What about AI-driven job loss? ← Back

What’s the business case for this service? ← Back

How are the database tables structured? ← Back

Where can I ﬁnd the code for this? ← Back

Email-based Intelligent Virtual Assistant (EIVA)

Email-based Intelligent Virtual Assistant (EIVA)

More Decks by Anand Chowdhary

Other Decks in Technology

Featured

Transcript