Chatbots today are already so powerful that they are able to play an essential role in a modern customer service organisation. Nevertheless, sometimes their answers still seem rather awkward. Knowledge Graphs, however, can be used to provide significantly better answers.
Google and Facebook both use them, and numerous other big players in the technology world are paying increasing attention to them. But what exactly is behind this hot topic?
A Knowledge Graph is a knowledge database in which information is structured in such a way that knowledge can be generated from it. The term itself was first introduced by the IT company Google in 2012 and has now become a synonym for a special type of knowledge representation. In a Knowledge Graph, entities are placed in relation to each other, given attributes and arranged in a thematic context.
1. Knowledge graphs and Chatbots — An analytical approach.
2. 🤖 How to talk to Computers: A Framework for building Conversational Agents — Part 1
The basic structure consists of nodes and edges — the former represents the entities; the latter describes the type of relationship between the individual entities. In a Knowledge Graph, nodes are provided with attributes and classified according to entity types. In addition, the edges between the entities are labelled with the relationship type.
In the field of computer science, graph theory is usually used to represent and analyse relationships between objects. For example, as mentioned at the beginning, Facebook uses a Social Graph to analyse the relationships between user profiles. Netflix, on the other hand, uses Knowledge Graphs to recommend suitable films to users whilst Springer Verlag uses them to store its portfolio and make it searchable.
By contrast, Google has been using a Link Graph to analyse and evaluate relationships between documents and websites and a Knowledge Graph to map and analyse relationships between entities for quite some time. By way of summary, a Knowledge Graph can be used to identify the semantic meaning of terms, their semantic context and their similarity to other terms.
It goes without saying that this opens up numerous possibilities, since knowledge obtained from a Knowledge Graph can be easily mapped. Another factor that contributes to this is that this knowledge can be expanded very easily, simply by adding data. This offers decisive advantages, especially in the area of chatbots. The number of intents (predefined sample questions) can be enormously reduced by using a Knowledge Graph
Furthermore, in comparison to classic chatbots, considerably more complex questions and answers are possible. For example, by using chatbots to perform mathematical operations or comparisons.
To learn more about the symbiosis of Knowledge Graphs and Conversational AI, check out our whitepaper.
For numerous industries, such as the tourism or events sector, using a chatbot based on a Knowledge Graph opens up completely new possibilities for optimizing conversations. This way, questions such as
1) “What events are taking place in my neighbouring village at the weekend?”
2) “What are the best family hotels along the XYZ hiking trail?”
3) “I need a ski pass, where can I get one?”
can suddenly be answered directly or made more concrete thanks to corresponding questions from the bot.
Let’s stay with the example question about the ski pass: By comparing the information linked in the Knowledge Graph, the system would determine that further information is required — for which region and how many days the pass is needed — and would ask accordingly. The original question can then be answered based on how the user responds. Classic systems do not offer this possibility, or if they do, it requires a huge amount of programming effort.
In addition, Knowledge Graphs also simplify the ongoing management of the chatbot systems, as well as the data stored in them. Today, simple messenger chatbots can be set up within a few minutes. However, in order to arrive at more sophisticated solutions, classic systems have to be trained over a longer period of time by entering the intents. Instead of defining and creating new intents for each new question that a chatbot is supposed to answer, Knowledge Graphs allow you to compare which information is already available and which needs to be added. New products can also be integrated more easily.
Another important point is that several Knowledge Graphs can be seamlessly linked together. To do this, simply use existing nodes or add new edges if necessary. This way, it is possible to set up the modern management of corporate knowledge without any problems.
Using Knowledge Graphs for chatbots thus offers users tangible benefits. On the one hand, they enjoy . By using them, companies obtain an extremely powerful tool for automated dialogue management via chatbots. After all, an essential factor in the success of a chatbot is the enhanced data integration and on the other hand, there is a significant improvement in the conversations structure and quality of the data available for answering questions.
uses Knowledge Graphs for numerous customers and projects, among which are tourism companies, banks, energy suppliers, furniture store chains, logistics service providers, IT companies and many others. There are use cases for almost all scenarios.
Read more about chatbots and Conversational AI:
