Search engines such as Google or Bing are like Swiss army knives: not only they scout for new web pages and documents on Internet, they fetch them, index them and offer users a very efficient access to their information thanks to their full-text search capabilities. Full (or free) text search is a technique for locating words and strings of text in documents or databases. In technical jargon, the documents found matching a given query are called “hits”.
In contrast to the generalist engines, the great majority of property web sites pre-determine the user experience with field-restricted search of specific real estate terms. The search experience is organized by fields including location, prices and other such as:
* Type of property: house or flat
* Basic features: rooms, number of bathrooms
* Attributes like garden, parking, garage, swimming pool, etc.
We got to work at Lokku Labs to develop a full text search capabilities on our extensive database of property listings. We tested our prototypes of Gartoo with queries like
* easy access to Heathrow and Gatwick
* Canary Wharf flats to let with security
The results of the pilot project are very encouraging. We manage to get hits for many interesting queries involving properties and location names. The challenge is to “understand what the query really means” and to find and rank each property listing by its relevance to the query. For example, “Houses to rent in Sale” is one of those cases that defy the usual understanding of natural language. Our challenge will not end with the ranking of results; we will analyze searches that do not produce satisfactory results to try to improve our algorithms accordingly.
Besides learning a new search technology, we will experiment with an extremely simple design. We plan to run a live test of our work very soon indeed, please keep an eye on Lokku Labs.
