Conversational experiences are interactions between users and devices that simulate human conversation. They can be done through text or voice and are processed using natural language interpretation mechanisms, rather than a fixed hyphen.
Some use cases include:
- Connecting users to business : e-commerce and customer service.
- Internet of things : home entertainment and automotive devices.
- Connecting the business with its employees : research and automation of tasks
In this article we will cover the concept of Chatbots and Voicebots: what they are, implementation tips and usage scenarios.
What are bots and why use them?
The primary objective of the Bots is to provide an access point to a service through of a conversational interface.
They may be responsible for d e the initial stage of an interaction, such as the first stage of customer service, or of complementing the experience of another service, such as notifying boarding updates at the airport.
The adoption of mobile platforms by users allowed companies to make their services available through proprietary applications, which ended up saturating the market.
Allied to this event, the need to install and Uninstalling the applications meant that only a small part prevailed, most were messaging applications.
Such applications have provided a new way to expose products and services to users that, along with advances made in the field of processing of natural language, have allowed to automate in an intelligent way the channels of access to these products and services, thus promoting the creation of experiences with nversacionales.
Types of bot
- Bots personal: bots that interact with only one user at a time.
- Bots from team: bots that interact with multiple users on a single channel.
Household robots such as Alexa or Google Home respond to everyone in the environment as if they were a single user.
- Business bots : bots intended to perform tasks such as automating processes and notifying users (Ex: Slack) .
- Consumer bots: bots intended to stimulate interactions with users, such as sending coupons and news on e-commerce platforms (eg: Facebook Messenger).
Chatbots are bots that interact with the user through of a text format. They are popularly used to automate tasks such as ordering food or buying movie tickets.
It works more independently, allowing us to integrate it with other platforms such as Facebook Messenger, Slack, Microsoft Teams, among others.
- Personality : the personal characteristics of the bot, such as the sense of humor and tone of voice, depending on the brand, the type of end user and the tasks to be carried out.
- Rich interactions : allows the processing of files, images, audios, maps, among others.
- Notifications : allows sending notifications to motivate the user to carry out new interactions, avoiding any form of spam.
- Management dialogue: Manage dialogues with multiple intentions and make appropriate decisions It’s throughout the conversation.
- Comments : The comments let you improve the bot over time, thus improving error handling and user experience.
Voicebots are bots that interact with the user through voice. Such interactions are carried out through a voice user interface or VUI .
Currently, 72% of Google users who have an intelligent device interact daily using their voice.
This is due to the fact that voice communication is faster, more intuitive and empathetic, since features such as tone of voice, intonation and speed of speech are relevant for better interpretation of information.
Furthermore, the preference for voice communication has increased with the inclusion of smaller display devices on the market , such as smart watches or without displays, such as Alexa and Google Home.
The choice of a voice recognition system depends on two crucial factors :
- Dataset robustness and precision : recognition and correct interpretation of speech.
- Performance of end point of speech detection : how fast the device Identify the beginning and end of an interaction.
These factors are responsible for defining the success rate of your conversational experience. In addition, it is recommended to consider:
- The need for an intrusion system ( Barge-in ): determines if users can stop the bot while responding .
- Waiting times : what will be done when the user finishes his shift in the conversation.
- Lists N-Best : a list of possible speech recognition results. It helps prevent “not found” results and treat cases where the user corrects the information.
In VUIs, voice biometrics can be used to Two purposes:
- Authenticate a user : Dialog boxes can be used to perform user access control. Such functionality requires the license of specialized systems and its use is the subject of several security-related discussions. You can read about an interesting case, in this link.
- Identify users : the voice can be used to identify the user who is interacting with the bot. This way of use can be found in bots that transcribe the content covered in meetings, for example.
Wake up words
The activation words are words that they are used to initiate an interaction with a VUI, such as “Alexa” or “Ok Google.”
Some considerations for choosing this type of command:
- They cannot be short words as it makes processing difficult.
- They should not be frequent words in everyday life.
It is important to find words that represent a balance between poor recognition (having to repeat the word many times to start the interaction) or excessive, where the device misinterprets similar words (confusing “Alaska” with “Alexa”, for example).
Bots require a Wide range of data to provide a personalized user experience, which can be collected in the following stages:
- When initiating an interaction : IP address, provider.
- During an interaction : information required by the bot, such as names, phone numbers, and email addresses.
- At the end of an interaction : comments collected through user-provided surveys.
Therefore, it is necessary to ensure the security of data and the privacy of users. Consider the following points of attention:
- Avoid storing biometric data : for example, when starting an interaction with the word for activation “Alexa”, the Processing of this audio is done on the device itself. Then, the entire stream is based on the cloud and the processed audio is discarded at the end of the interaction.
- Avoid capturing audio permanently : Process the captured audio only after pressing a button ( push to talk) or by means of a activation word thus avoiding listening to the user’s conversation.
- Processes data anonymously: avoids associating sensitive user data, such as names, birthdays or financial data, with interactions made with the bot. It is also recommended to remove such information from the records generated by the voice recognition mechanisms.
More information on how to guarantee the privacy of users can be found in the General Data Protection Regulation.
Bots are powerful tools that can help users and businesses in their daily activities, automating and simplifying tasks and processes, from notifications to sentimental analysis and image interpretation.
In the words of Emmet Connol:
Bots should be used to enhance the end-user experience, not just to make life easier for customer service teams.