Holistics NLP Questions Interface

Systems such as PowerBI and other startup companies offer and promote NLP based interfaces for business users to ask questions of the data they interact with, rather than having to either write SQL or even just drag and drop columns onto charts.

Is this something that Holistics is looking into for the product at all?

Hi @DataGeekDude

We are still considering the feasibility and potential benefits of incorporating NLP-based interfaces into our product. Therefore, I cannot provide you with a definite answer at this time. However, please be assured that our team is constantly exploring new ways to improve the user experience and make data exploration more accessible.

In the meantime, would you mind sharing with us why you believe a NLP Question Interface is superior to a Drag-and-drop Interface for business users?

Some business users I know find it hard to know which dataset to pick, how to choose the right column or graph and just want a question answered.

So if you had an NLP interface then you can set up synonyms of columns, and use other techniques to make the interface for everyone to be super simple to get a quick answer.

For instance, “what is our total sales revenue for the past 6 months” might be the question, but our columns are named “sales income” and we have multiple date fields. This might be confusing to a user, but for an NLP engine quite easy if we can configured it ahead of time.

5 Likes

Hi Craig, giving you a quick update here.

Some business users I know find it hard to know which dataset to pick, how to choose the right column or graph and just want a question answered.

I understand this problem 100%. Currently, given the current Holistics approach, business users face with the following problems:

  • Don’t know which dataset to pick for their particular question.
  • When they’re in a dataset, they don’t know which fields to choose from. They might feel overwhelmed because there are too many fields presented. Sometimes these fields are duplicates so they don’t know which are the right fields to choose.

We understand this hinders the ability to self-serve for business users. Ideally, they should not be concerned with the datasets and fields. They should just be able to ask questions: “I want to see metric A, broken down by B and C, for the time period of D”.

We’re starting a big project to rethink how we design our self-service experience for the end users here. I hope to share more details with you in the next couple of weeks.

3 Likes

Thanks @huy ,

This is great to hear. I think Holistics would be in a great position with the AML layer describing how all the data fits together, especially if the columns, models and datasets have descriptions, to help business users ask questions rather than have to figure things out. There is a wealth of information there for each business that can be used into an NLP style interface.

It would be great to hear more and even chat more if you’d like on this topic.

Hi @huy

Was wondering if you had an update or anything to share on this topic? With all the buzz around ChatGPT and some new platforms coming out that are very much focussed on NLP interfaces I can see this being an area of competition for Holistics and it is something that our users are asking about.

We know self-service is hard, it’s hard to get people to understand data let alone use a tool and drag and drop columns across unless they are data savvy so we are interested in getting more people using and analysing data, but so far that has been hard for us, so anything to make this easier would be awesome to learn about and or participate in.

1 Like

Hi @DataGeekDude

With these 2 recent posts written by our team LLM Is Not Enough (For Self-Service Analytics) and The Future of Self-service BI: AI-guided BI, I believe we all agree that NLP can significantly enhance the self-service experience. However, we want to make sure that we have a solid foundation in place before we incorporate NLP in order to assure its capabilities, reliability, accuracy, and security. This includes features like Dashboard as Code, Data Exploration as Code, and AMQL Expression.

As outlined in our roadmap, we are gradually building these foundational elements. Once they’re established, I think Holistics or even any users can integrate LLM interfaces to create a governed, AI-guided self-service experience.

In the meantime, we’re eager to hear more about your thoughts, the challenges you’re facing and how you’re trying to address them.

5 Likes