Natural language processing for Kasaz

Artificial Intelligence – NLP – Data Improvement

Client: Kasaz

Challenge: Finding keywords in descriptions and improving the data quality through natural language processing.

What is Kasaz & their mission?

Kasaz is the real-estate portal of the future, the first truly buyer-focused property portal in mainland Europe. Where the status quo is duplicates, fake listings, and inaccurate information, Kasaz offers unparalleled transparency and listing quality combined with superior user experience.

How NLP improves data and UX

A lot of information is available in the text description of a listing. It contains important real estate related words which aren’t available in the metadata (data that can be processed by a computer such as a price, number of bedrooms, and more).

We improved the data of Kasaz by scanning their listing descriptions and getting tons of real estate related keywords such as terrace, open kitchen, bath, garage, fireplace, and more.

After a thorough look at the data of Kasaz and a detailed EDA report, we concluded that there was a big opportunity to enrich their data. Variables in the dataset such as ‘pool’, ‘garage’, ‘air_conditioning’, and more were available in the descriptions.

Structured data (metadata) such as pricing, address, and more were ok in general.

On average, Co-libry noticed that 80% of the unstructured data (features like pool, garage, balcony, and more) were not yet available in the metadata.

In this example (picture), you see that our tool detects the following keywords: ‘garaje’, ‘ascensor’, ‘trastero’, and ‘aire_acondicionado’. (In English: garage, elevator, storage room, and air conditioning)

The power of Natural Language Processing is not just detecting the exact keywords but also misspelled words and synonyms. This is an example of the word ‘ascensor’ (elevator) where our AI engine is searching for in descriptions.

31% more searchable features!

Based on this solution alone, we were able to improve their data with 31% more searchable features. What does this mean? That on average, X1.31 more listings will show up to visitors.

The result? Happier real estate agents because more or their listings show up and longer user session lengths.

Want to see more about NLP?
Check our quick demo here. (no email needed)

31% more searchable features!

Based on this solution alone, we were able to improve their data with 31% more searchable features. What does this mean? That on average, X1.31 more listings will show up to visitors.

The result? Happier real estate agents because more or their listings show up and longer user session lengths.

Want to see more about NLP?
Check our quick demo here. (no email needed)

REQUEST YOUR DEMO

AI and personalization made easy

Start today!

REQUEST YOUR DEMO

AI and personalization made easy

Start today!