As you know, the conviction that drives us is the same since the creation of Qwant: there is a pressing need on the market for a search engine that respects the privacy of its users! But to develop this technology, we had to make choices…
The TECH ROCKS SUMMIT was an opportunity for me to put a little spotlight on what is at the heart of our adventure: the technology that runs our search engine and the choices we have made since my arrival last March!
As you know, the conviction that drives us is the same since the creation of Qwant: there is a pressing need on the market for a search engine that respects the privacy of its users! But to develop this technology, we had to make choices… and some compromises in order to offer the best possible service to our users while guaranteeing respect for our values.
During my speech, I explained how it was decided to delegate part of the indexing work to a partner: Microsoft Bing. Since the first production, Qwant has laid the founding bricks allowing it to only delegate to its partner. A substantial part of user requests is now entirely managed by Qwant.
Our SERPs, Search Engine Result Pages, are composed of several parts, and the list of documents is the one that concerns us in this post. Quality and relevance are key, and to improve them, we develop our Learning-to-Rank (LTR) algorithms. This is a crucial decision because the LTR is truly the heart of the reactor!
It is this technology that allows us to evolve Qwant towards a growing autonomy and ever more relevant results. Similarly, it was thus more interesting at first to rely on Elasticsearch for indexing.
This solution allowed us to take the first steps. But the latency increasing with the volume of the index, we have recently made the choice to change technology in order to lower our average response times from one second to sixty milliseconds, and thus free ourselves from a constraint prohibiting us from considering total autonomy.
In a few words I just talked about the response time of our index, compared to its size. Today, Qwant is able to respond in total autonomy to the requests of its users in 40% of cases in terms of the proposed list of documents. But so many other elements are to be presented, e.g. the crawl speed that allows the freshness of the results, the quality monitoring that validates the relevance, the artificial intelligence that makes it possible to work on this relevance. Other blog posts and technical interventions will allow us in the future to present Qwant’s efforts and advances.
By refocusing on our core business, we ensure that we have the fundamental skills and tools to be essential tomorrow.
