ThoughtSpot Natural Language Search on Databricks Data

0:02uh hi uh good morning so uh today I have a special guest and this special guest will show us how to get started quickly with to spot on data breaks thank you Yousef hello everyone good morning my name is Penelope I'm a Solutions engineer with hotsport and I

0:26will be very happily show you today how easy uh to start searching um based on your data breaks data using thoughtspot yes and uh and we will show you every single step like how to create the connection how to select a table how to edit a table add description and then generate automatically dashboards for

0:50you and the purpose is really to democratize the access to the data or leverage weld the data in your company and the floor is yours amazing thank you so much Yousef um all right so I'll start sharing my screen maybe I'll do it full um and this is thoughtspot I will go straight away to

1:12the data tab of thoughtspot where you can go and start creating your connections so I'm going to be using one I've already um created just to uh save us some time but the idea is that you um create the connection using your credentials um and then the next step would be to go and connect tables and

1:36columns um the different types of connections we support with Databricks you could use either a service account or all type of um connection as well if you want the authentication to happen within data breaks um the next step and as you can see um I've already selected uh three

1:59tables that I've connected but I basically will deselect them to show you the process so the first thing you see once the connection is there is essentially um an empty a blank canvas um so what I can do I can um uh find the data sets I'm looking for in this case it's like a retail um type of data set

2:21that I'm going to be using and you see the tables that I want to connect so let's go and connect them so this is my fact table so I'm going to go and select all of the columns fact table which includes my sales um I'm going to connect a dimension table that includes information about my stores and I'm also going to include a a product Dimension

2:44table that includes information about my products so as you can see now all three have gone into the uh selected let's say tab where I can go and review what thought spot is connected to and let's go ahead and update the connection great so um in case you've already

3:07established the joints within Databricks thoughtsp would inherit those joints so you see how for example I can see the number of joints per table so I can go in each of these tables and I can review those joints right so for example these are too many to ones H inner type of joints that my fact is connected to the dimension tables so this is really

3:31um what it takes to create uh the worksheet um on top of the T sorry to create the tables uh to connect Thoughts with to datab Bricks tables um and the next important step and essentially the step you need to do in thought spot so that you can take advantage of it is as we say h create a worksheet this is the example I've

3:53created but let's go and create it together so that you can see the process so I'm still within the data workspace um so the data tab within data workspace so let's go create new worksheet so first step is to choose my sources so um I've named the tables as my initials and then um data breaks so

4:16these are the three tables we just connected so let's go and select them so the worksheet is essentially the layer the data model if you like that you build on top of your tables um that you basically then expose to your business users so that they can start searching across all of these different tables together right because in a

4:40retail scenario we are interested about the sales but then we also want to enhance it with information about the product or the stores as we saw together um the only thing I would always recommend when you build a worksheet is to make sure you keep only what's important for your users what would make sense for the users because this is how you um don't tend to for example

5:04overwhelm them with everything that is available so let's go and maybe remove the IDS because in the IDS in this case I just use them for um uh join purposes and they're not really important for my users so let's say that this is the information I want to keep uh for this is the information I want to keep for my users so let's go and save it we can

5:29give a name test PC worksheet date something like that so this is the new worksheet we just created together um this is also the layer where you can go and configure and I guess find tune your model to make it as easy as possible for the business users to understand so as

5:51you can see here you can go and add descriptions what I could also do I could uh change easily the names all of that is editable so I make sure that my users understand that and it's a little bit more you know easy to use and understandable um you can add descriptions you can add synonyms um we

6:13also have a feature to automatically generate the synonyms if you want uh so for example I can go and I can um um I can click that so that the synonyms will be generated I need to enable that in the background um I can set up um aggregated um types aggregation types for my measures um importantly I guess

6:35things like you know you want to map your I don't know the zip code to be a zip code so I can come in here and I can map it to a zip code or for example the latitude is a Latitude the longitude is a longitude which basically means or my state I can also map my States uh so sub Nation um u based on you know the geographical type of information and the

6:58idea is that if my users are going um directly use these um columns automatically they will get the best feed visual which in this particular scenario would be a map right and similarly for example for this particular um uh geographical type of information I don't want them to be measures even though it's like a number

7:20I want them to be attributes so these are the sort of configurations that you are doing at the worksheet layer and now we are ready to start search um so I'm going back to the homepage and this is basically where you know the first thing I see uh is something we call Sage um and this is our natural language search capability

7:44so you select the worksheet you build uh so in this particular case what I can do I can come in here sorry Databricks is I named um and test pie see Databricks is the one we just built uh so so this is the uh worksheet we just built together and I'm ready to come start like to start asking natural language questions so where the magic will take

8:08place exactly and youf sorry for this monologue but everything makes sense so far do you have I mean to be honest is very very clear so that's why I didn't step in I mean everything is super clear it's very very easy to use so that's that's perfect awesome thank you so much so let's go and ask the first question something like um what are the best

8:34performing uh products let's say this year something like that so let's go see what happens so thoughtsp integrates with large language models and um GPT 3.5 and four in specific and essentially what we do we translate this question from the user into our keyword search

8:57for whoever is familiar with hotspot our key keyword search our token based search is traditionally the product we've been building for the past 11 years but now we took it a step further and we make it even easier for people to start searching so you see how I'm getting you know the top products so I'm getting the product name uh so it recognized it well enough to map it to

9:19the relevant um you know columns from my data model record date would be this year um sort by sales um uh in descending order of course because I asked the best performing so it would map it to the top keyword um but I don't have to stop there I can always continue drilling down drilling anywhere in the

9:41data so any of the data points from the visualization is fully interactive so if let's say I want to drill down um you see a list of columns that are available in a in my data model without me having to predefine that drill down so if for example I want to see the different let's say States we sold that particular

10:04uh product you see how then thoughts would will be give me the best fit visual so that I can understand the states and you can see that you know in Nevada and California we've got like uh stores performing very well so maybe I want to drill down to Nevada and I want to see the particular um let's say Stores um that we sold that particular

10:27product type so with thoughtspot you are not limited to a particular visual you can always go down to any level of granularity you want I could go for example all the way to the particular you know product name is actually what we're searching on but for for example I could go and see the specific I don't know um uh let's say dates we sold

10:52that particular product and so on so I'm free to continue exploring the data the way that I want hopefully that's a good example to show you the capabilities of what you can do with data breaks as well as thoughts for uh to quickly get insights out of the dates and your in your lake house that's really that's really that's really amazing I believe I have one more

11:16question for you uh if someone want to know more about tough spot uh to whom should this person reach out or is there any useful links yes absolutely so uh first of all um if you um anytime want to reach out um I'm very happy to you know to reach out

11:38directly to me however what is um something I would definitely recommend is um getting started um SP free trial so getting started with our free uh 30-day trial um and potentially you Yousef we can add that link to the description of the video or something something like that um so that you can

12:01uh start yourself uh follow the steps we showed you today and get started and if you have any questions or if you have any particular um you know uh things or U more things you can do always feel free to reach out um to us that's perfect then just for information tough spot is a part of partner connect so you can create your

12:25endpoint or SQL Warehouse very very easily and join the two platforms to get the most of your data that's really the important thing to know so thank you so much Penelope and hope to see you soon thank you so much user for having me