Topical Summer: Improving context awareness in AI chat with topical mapping
The summer is upon us, and it's a great time to reflect on the achievements of the first half of the year and take some well-deserved rest. However, at TekstiAI, development and operations remain in full swing through the summer as we prepare to bring exciting new functionalities to our clients.
Tommi Kostilainen
• Product News • Reading time: 3 minutes
TekstiAI Summer Update
We all know well by now that the artificial intelligence field is moving forward at an accelerating pace. For companies like TekstiAI, which aim to enable AI ecosystems at scale for companies in all sizes with a one-size-fits-most approach, it means a constant cycle of strengthening the variety of functionalities we offer to our clients.
We exist so that we can cater to your AI needs
- Immediately
- Securely
- Without locking you into a single AI ecosystem
- Without an expensive implementation project
During the last few months, we have been cooking up our next killer feature to bring value to your company across all its functions. We are proud to introduce Topical maps.
Mapping your company
Possible problems with introducing your company data to the Artificial Intelligence system your company uses include data hygiene and proper contextual understanding. At TekstiAI, we have aimed to tackle these issues from the get-go.
Environments have been a feature in our application since the very beginning. This means that the material fed to our Chunking Engine has been introduced with context. The environmental context helps retrieval-augmented generative AI to be more specific about the context in which it works, as well as ensuring that hygiene between company functions remains intact. For example, R&D-specific material is often confidential, so only those with access to the correct environment should be able to ask questions regarding those subjects.
As useful as Environments are for data hygiene, mapping context with them might prove laborious, especially in larger companies. It might also, in some cases, contradict data hygiene rules, and all those of us who have worked with an access-rights nightmare know that this is to be avoided.
We are confident that the most useful way to bring immediate cross-functional value from Artificial Intelligence, as well as the best option for ease of use and understanding, is to map your company's material on a topical basis. Your company data will be chunked, analyzed, and placed into contextual topics with which you can then chat, all while having access to the documents used in said context, as well as receiving the document references you know and love in your answer. We call this a Topical map of your company.
Topical maps
In this example, you can chat with any context, from vague to more context-specific. Company overview will give you answers from all materials that you have access to. Sales will give answers based on all materials under Sales, Materials based on all materials related to Sales materials, and so forth.
Environments will still ensure that data hygiene remains intact, making sure that a user can only see a topic for which they have environmental access to at least one document. If the user chats about said topic, they will only get answers based on the documents that they actually have access to.
You can traverse topics on a high level to decide on the level of context you want to work with.
You can see the parent topic on top of the chat screen, and the child topics below the chat screen. The chat interface is familiar to those who have used our current material chat functionality. Notice that the reply is a mockup due to confidentiality reasons.
Don't hesitate to contact our sales team to see our application in action, and to hear more about how our new features will help your company grow.