You Can’t but Machine Learning Can Stop Fake Real Estate Listings
In most of the consumer studies in the Indian real estate, deficit of trust comes up as the big highlight and there have been various reasons for this in our country. Some of the regulations like RERA are definitely going to help reduce this trust deficit. On the digital side, platforms have mostly focused on solving the use case of search and discovery for home seekers. But even there lack of MLS (short form of multiple listing service, a unified source of listing content) has raised the issue of fake listings made up of incorrect information being uploaded to various platforms.
There are 2 ways to address this issue:
1. Go and verify each and every house detail yourself and only then allow the listing to be uploaded on platform - This approach is the tightest approach to verify accuracy of information but it poses the challenge of very high cost and low scalability. In a country as big as India, beyond a point it becomes almost impossible to scale this operation.
2. Technology intervention to figure out fraud content - This approach is definitely more scalable, but so far it has had huge limitations in terms of how much fake content one can verify using expert driven technology solutions and hence the adoption of this approach was highly questionable. But some serious advancements in the field of artificial intelligence have brought in big hope in solving this using machine learning.
Let me throw some more light on this:
Machine learning is a sub area of artificial intelligence, which gives computers the ability to learn without being explicitly programmed. The traditional approach of a rule based system driven by experts has its own limitations as it can only be as good as experts and also it is very difficult to deal with the complexity of the rules and their influences from each other beyond a complexity level.
Real estate platforms receive millions of listings and through various tech and manual ways they mark the quality of listings on various parameters like image, price and description. This data can be fed to a machine learning system for training and with more and more data they can keep on refining their internal model and after a point the system becomes more effective than any manual check. Overall the content inside a real estate listing can be divided into 3 major categories:
1. Images
2. Statistical content like price, size and amenities
3. Textual content like description of the house
The advancement in image classification technologies based on AI has created a very positive roadway for verifying images inside a listings. Not only one can flag images which are of non-real estate category or which are not safe for work (NSFW) but also perform auto tagging of a given image as a bathroom or a kitchen. The more you train your system the better the classification.
For statistical content there are multiple ways like finding out variance from the other verified listings in the same neighbourhood.
For textual content, NLP (natural language processing) and machine learning can be used very effectively to decipher whether the description written is in alignment with the other information set of the listings or not.
I am really excited by how these advanced technologies are able to address some of the deep rooted issues of different domains and I am sure that this will be the only scalable way to address these fundamental issues. There are very interesting developments going on the front of physical verification of the house using technology. As a technologist, I am excited to witness these experiences where technology is not only leading to better outcomes for our consumers but also helping companies remain leaner for better growth.
