AI Insights - Using Text Analytics algorithms in Power BI
Updated: Dec 1, 2020
AI, Machine Learning… all these buzzwords mean anything to you? If you’re in the data world they probably do!
What about AI in Power BI? Applying a language detection algorithm to further analyse your data? A sentiment scoring algorithm?
If you’re thinking “naaa, that sounds incredible, but it must be way too complicated”, then you will love this post!
Power BI allows you to apply AI algorithms in a user-friendly way, and no, you don’t need to know how to program or anything like that, all you need is Power BI Premium!
Hint: you don’t need to spend huge amounts of money to buy Premium capacity, recently Microsoft announced it will introduce a Power BI Premium per user licensing option, which means you can have most of the capabilities you have when using Premium Capacity, but paying for only what you need/use. Awesome isn’t it???
Having all of this, today I’m putting the Detect Language algorithm to test. Let’s see how well it does!!
To test the algorithm, I downloaded a sample dataset from my favourite data science website: Kaggle!
This is the dataset I chose for this test:
Note: I noticed the dataset has some inconsistencies and wrong entries, so I just used 200 entries for this test. Also, for you to be able to fully use AI Insights you will need Power BI Premium!
1) To do this test yourself, just open Power BI Desktop, and load the data from the sample dataset. The dataset is a csv so you just need to use the Text/CSV connector:
2) After you loaded the data in Power BI, in the Home tab click on Text Analytics on the AI Insights section:
3) A popup window will show. You will see three algorithms you can choose to perform Text Analytics. Click on Detect Language and choose the column you want to detect the language on:
Remember that for the algorithm to work you need to have access to Premium Capacity.
4) Two more columns will show on your table, one with the name of the full name of the language detected and another one with the ISO abbreviation:
5) And it’s as simple as this! In my example, because I wanted to test the algorithm, I had a column already filled with the language for each text entry.
Next, I just loaded my data and built a very simple report to check how well Power BI detected the languages used in my data entries, and the result was…
From the 212 data entries, he only got 5 wrong! And if we analyse the incorrectly predicted entries… I give it to you Power BI! Some of those were tricky anyway.
My final verdict for the Detect Language algorithm: