BERT makes Google more human

By now, it’s normal for Google to change something in the algorithm almost every day, making improvements to the search. Really big updates, as they are known from earlier, have become rare. Last week, however, it was time again and with BERT came the biggest update since RankBrain was introduced five years ago. BERT is the abbreviation for Bidirectional Encoder Representations from Transformers and is said to start with about ten percent of English-language queries for better results.

Google is quite proud of this update and sees it as an important milestone. With the procedure used here, which actually comes from the field of Natural Language Processing (NLP), they would like to be able to better answer search queries that are more in the form of keywords than in complete questions. So far, the problem often arose that these search queries could not be interpreted correctly because the previous Google algorithm could not always interpret the relationship of the words correctly. For example, prepositions have been completely ignored, although these are just important to answer the search query.

So far, this misinterpretation of search results has so far been the case if the request contained a “not”, “no” or similar. That was simply ignored and therefore the results were not helpful. For other requests, the keywords were recognized correctly, but without being related. Again, the results were then far off. This is exactly where BERT should start and make better results possible, which of course also helps the users and thus make them satisfied.

BERT should therefore be able to better interpret the human language and ensure that users receive the right results in their inquiries in the future – and without having to follow the “search engine language”. This should come in the field of voice search which has very positive effects. Here, the requests often come in complete sentences, which Google has often posed problems so far. BERT could remedy this situation and deliver the right results for these complex queries. As mentioned earlier, the update is only included in the English language search, but Google assures that BERT will be able to transfer everything he learns to other languages ​​if they get the update.

What does this mean for webmasters? It is not directly possible to optimize for this update. It is important, as always, that the content of the website is accessible and can be captured by the search engines. Furthermore: Texts should be written in natural words and not in nested sentences – therefore making the chances a lot better that they will do justice to the search queries and thus a good ranking is possible.