Building bots has been growing as a hobby in the tech world. Many companies offer bot building services, or integrate chatbots as smart assistants. However, things do not always run smoothly. Bots still have quite a lot to learn, and it is our mission at SAP Conversational AI to push the boundaries of what has been done to provide the best bot building experience.
One of the issues we always had when building bots was detecting keywords when they are not included in a sentence.
In this scenario, a user wants to find a good restaurant in Italy. Unfortunately, her name is the same as the city she’s in right now: Florence! When the bot asks for the location of her desired restaurant, she simply answers “Florence”. Unfortunately for her, the bot understands it as a name. In itself, it isn’t wrong, but that is still not what is expected of him for the current conversation. Florence is very disappointed, has lost all trust in this chatbot, and leaves the conversation.
This happens much more than we’d like to. In the past month, our team has been working on different ideas, and is today proud to announce that we’ve reached a solution! We are glad to announce our newest feature: context management.
What does context management mean?
Context management gives bots built with SAP Conversational AI a greater sense of conversational awareness. It helps you create human-like conversations by reducing the chances of misdetected entities. This release drastically improves the performance of your entity detection. During a conversation, your bot is learning as much as possible from the flow, the user answers, the vocabulary choice. By doing such an analysis, it is capable of predicting the answers at all times, therefore understanding them much better.
This feature is not something tangible: it is integrated into all bots using Bot Builder with Skills. And because of it, Florence can make her restaurant booking smoothly.
What’s next ?
We’re not stopping there, trust me! Today’s release is a big step in conversational technologies, but we want to go further than that. Tomorrow, we’ll be working on providing context on whole intents, languages, sentiment and emotions. These will allow the use of humor, sarcasm, foreign words, and overall improve the liveliness of chatbot conversations. And after mastering that in an intra-conversational level, we’ll just keep expanding to inter-conversational awareness and then global user awareness. Sweet, right? We’ll keep you updated.
If you want more technical information, please read our changelog. If you have any questions, please join our Slack Community to talk with the team!
Happy bot building.