Feeling Lonely? Depressed? A New AI Chatbot Can Get You Out of the Rut

Category: AI Insights

Published date: 01.10.2021

Read time: 6 min

Even during the early days of artificial intelligence, researchers always dreamed about a system that could engage in empathetic conversations with humans. For example, we can go as far back as 1950 when the Turing test was first introduced by Alan Turing, which determines whether or not a computer is capable of thinking like a human being. Over the years there have been many innovations designed to mimic human behavior in a text-based conversation, to pass the Turing Test within a controlled scope. Even though a lot of these systems did have some successes, were mostly based on hand-crafted rules and worked well only in constrained environments. An open-domain social chatbot had remained an elusive goal until recently.

Today we will tell you about Xiaolce (literally “Little Ice” in Chinese) which has become one of the most popular social chatbots in the world. It has 660 million registered users and more than 5.3 million followers on Weibo, the Chinese equivalent of Twitter. Let’s take a closer look at Xiolce to understand its capabilities and how it was created. 

Why Makes Xiaolce So Impressive?

When we look at previous generation chatbots, they had a lack of understanding of the context of a conversation, which prevented them from going deeper than a single call-and-response. In other words, the systems had no real memory of what’s come before, and no real “understanding” of what it’s talking about, leading to disjointed conversations. Xiaolce solves this problem by using a context vector mechanism that keeps track of the broad topic of the conversation, alongside another set of attributes for the person it’s talking to. It also uses sentiment analysis to determine a person’s mood and adapts its responses accordingly, a form of robotic empathy.

Speaking of empathy, this was one of the goals Xiolce creators intended to reach. You see, Xiaolce researchers wanted to pass a particular type of Turing test known as the time-sharing test, where machines and humans coexist in a companion system with a time-sharing schedule. If a person enjoys its companionship (via conversation), we can call the machine “empathetic.”

This means that if a person is feeling lonely or depressed, they will be able to turn to Xiaolce for companionship and understanding. Xiaolce will remember their previous interaction and will try to cheer up the user. 

How Does Xiaolce Work?

Users can converse with Xiaolce, which is the Chinese equivalent of WhatsApp and other messaging services either via voice or text. Xiaolce communicates with users in two modes: full-duplex and taking turns. The full-duplex mode handles voice-stream-based conversations where a user and XiaoIce can talk to each other simultaneously. This mode is mainly used for the XiaoIce systems deployed on smart devices. The other mode deals with message-based conversations where a user and XiaoIce take turns talking.

When we delve deeper into the architecture there is a conversation engine layer, which is composed of a dialogue manager, an empathetic computing module, Core Chat, and dialogue skills. There is also a data layer that consists of a set of databases that store collected human conversational data (in text pairs or text-image pairs), non-conversational data, and knowledge graphs used by Core Chat and skills, and the profiles of XiaoIce and all its active users. This is where the data annotation aspect comes into play and Mindy Support can be of great assistance. 

What Types of Data Annotation are Necessary to Train Xiaolce? 

The training data used for Xiaolce came from popular messaging apps where humans communicated back and forth with one another. This means that entity annotation was necessary which is locating, extracting, and tagging entities in text. This can include things like: 

  • Named entity recognition – Annotating entities with proper names 
  • Keyphrase tagging – Certain keywords and phrases will need to be located and tagged accordingly
  • Part of speech tagging – This involves tagging functional elements of speech such as verbs, nouns, etc. 

We also mentioned earlier that Xialcoe is capable of sentiment analysis. This process involves annotating polarity i.e. positive, negative or neutral. It is possible to go a bit deeper than this and annotate evaluative sentences with respect to being on-topic or not and with respect to expressing a first-person opinion of the author or not. 

Mindy Support Provides Comprehensive Data Annotation Services 

As you can imagine, it takes a lot of data annotation to train an advanced system like Xiaolce. However, regardless of the volume of data you need to be annotated or the complexity of your project, Mindy Support will be able to assemble a team for you to actualize your project and meet deadlines. We are the largest data annotation company in Eastern Europe with more than 2,000 employees in six locations all over Ukraine and in other geographies globally. Our size and location allow us to source and recruit the needed number of candidates within a short time frame and we can scale your team without sacrificing the quality of the work provided. Contact us today to learn more about how we can help you.


    Stay connected with our latest updates by subscribing to our newsletter.

      ✔︎ Well done! You're on the list now