BERT – Latest in Google’s vault
- rwangchuk
- Nov 10, 2020
- 3 min read
Introduction with facts
One of the most impactful news came in the last quarter of 2019 when Google announced the introduction of BERT, which stands for Bidirectional Encoder Representation Transformers. It is considered to be most advanced update by Google since the very inception of Google itself! This specific update is powered by machine learning and the AI model and focuses on the context of the user’s query than just the string from one word to the next. With the help of BERT now, Google search can give the user not just results matched by the words one types but with the context of the searches, hence making the results more useful and relevant to the user.
Some of the contributions by BERT on the search landscape have been the following;
Auto-complete policy change, to be more conservative in its results, especially with the elections, and show less than more in an area
Google BERT used in full coverage in the news stories
Fact checking, more collaborative with some authoritative sites online
With partnership with Wikipedia, able to detect and remove vandalism online in the knowledge graphs
Google being able to break stories way faster within minutes Vs 40+ minutes earlier
Googles intention and search intent
BERT did stay an open-source for a year while testing its intention and making sure it did exactly what Google engineers wanted it to do, that is to refine the search queries, before officially launching it. The results followed with 1 in every 10 search queries being impacted, while keyword stuffing was anyway a thing of the past before 2019, great genuine content surely was the clear winner here!
Write naturally
As per Google, the cool part about the BERT is that regarding the content-creation side of things, there is nothing one needs to focus on apart from writing a great content for the user in mind (focus shifting on the search intent, hence resulting in creating the right search query )! All the traditional techniques, by the book rules so to speak just in order to be ranked on Google search engines including the clearly not so popular practice of keyword stuffing have been away with. Now, the focus for any content- creator should simply be about writing great content and writing naturally.
Search engine strategy
The traditional search engine strategy has always been about the process of search engine optimization on Google. How one matches keywords, tries to be ranked on page 1 of Google search engine. But with the introduction of BERT, the future of search engines and queries have certainly changed. Although machine learning and the AI model on which BERT works certainly is made by the experts for the better search queries and results consequently, I would still be curious to see how this works for the future. This is certainly good for the good content creators out there, where they are being rewarded, but I strongly feel there is still a science to the approach to search, search queries and the end-users one is and always will be trying to reach. Hence, a process to be followed which will never fade away.
Conclusively, though machine learning through AI are certainly doing away with the human involvement in the process of things, which is clearly visible through the introduction of BERT here, contributing and having a ripple effect too in the industry such as Microsoft’s Bing search engine and Youtube’s need for keywords for better search visibility on their respective platforms. Compared to the traditional ways of searching where one uses a string of words, keywords so to speak, what BERT has done is given the importance to natural ways of asking queries Vs the very abrupt ways of asking for queries with keywords etcetra. BERT simply emphasis importance on the nuances of things, the real intent behind the queries by the users, giving emphasis on the preposition too, if I can say so, words like ‘to’, ‘from’, in order to connect the nuance of a query. Though content is clearly the winner here for the long run, but one will still have to wait and see what becomes of the machine learning and Artificial Intelligence in the coming years, how granular or deep they would become.
References:
Comments