Delivering Wiki knowledge to Slack
Memonia for Confluence
At Memonia, we’ve already provided the automatic knowledge extraction from Slack conversations and sharing it with the automatic semantic search as well. So, next step was quite obvious — expand it to other knowledge sources.
We decided to give a try with Atlassian Confluence Cloud because it is a very popular solution to save (mostly) important knowledge and share it inside your company.
Now, with the cloud version, you can easily start using Confluence without installing everything on your site. There are also Atlassian’s plans to provide free tier subscriptions which give you a start without any cost.
We’ve implemented a seamlessly integrated solution where you won’t need any additional accounts. You will need only your existing accounts from Slack and Atlassian.
Why we need one more Confluence integration application?
Atlassian Cloud API and many other integrations, which we’ve seen before, provide only keyword-based search, so we’ve decided to go much further and created Memonia for Confluence that made semantic search possible and the knowledge from Wiki pages might be delivered automatically to Slack channels when Memonia finds relevant questions.
The challenges to delivering knowledge from Confluence are:
- Wiki pages are unstructured and might be represented in many forms. In the case of Confluence, they actually look like the simplified and limited version of HTML. There is a big challenge to extract and clean important information in this form from Wiki pages and filter out additional content.
- Searching in those documents semantically is also a tough task. To avoid noisy results you should take into consideration of pages structure, titles and location.
How does this work in reality?
All of your wiki pages in the spaces, where Memonia is enabled, are analysing to clean up and filter out less important content. Memonia does it automatically:
In this simple example of the page, all important pieces are marked bold here and will be used to semantically answer any possible questions related to those sentences via automatic or manual search in Slack channels:
This implementation was made possible by using new and state-of-art ML and NLP models which we use to provide knowledge extraction, sharing and search capabilities.
Of course, there is always room for improvements and we’re working hard to improve those models with the whole industry.