About Chatbots and Agents

According to the Chatbot Report from Salesforce, 15% of customers used chatbots as a communication channel with companies within a year. Chatbots are particularly prevalent in customer support. 24/7 availability and efficient operation are just two reasons for the growing adoption.

From answering frequently asked questions to complex application processes, many levels of sophistication are on the market. In the USA, for example, a company offers a chatbot service for class action lawsuits, thus even replacing the lawyer (Source). They don’t even stop at the medical field and help with serious conditions such as depression.

In addition to pure text input, voice command input also has great value. It has long had a fixed place in navigation and multimedia systems in cars, and at home, Alexa turns on the lights and heating. In the background, the logic usually works similarly to text input. “Only” speech recognition is added.

But with all the many use cases that can be covered with chatbots, Alexa skills, Siri, and Co., you wonder all the more why, for example, Alexa exits early even in smaller conversations.

The answer becomes apparent after a scientific consideration of the topic. According to a definition from Stanford, personal assistants and chatbots have the following characteristics:

• Task oriented
• Focused on handling thematically delimited areas
• Consists of an expandable set of ‘domains’ and ‘frames’

In this context, chatbots are referred to as “Frame based Dialog Agents.” A frame is an area or a task. When the agent receives an instruction, the domain and frame are determined as follows:

Wake me up tomorrow at 6 o'clock

The time is a variable, a so-called slot. This is interpreted from the instruction: “tomorrow at 6 o’clock” is converted by the agent into an exact date.

This information that the agent has extracted from the instruction can now be processed further. With the domain ‘ALARM-CLOCK’, a clock app can set the alarm on the smartphone in the example case. This is currently how most assistants and chatbot systems work: Alexa, Google Assistant, Siri, Cortana, and many more. At innFactory, we use “Dialogflow” from Google and “AWS Lex” from Amazon as frame-based dialog agents.