Understanding different languages can be quite a difficult task to perform. Plus whenever we construct a sentence, we can see that a sentence holding the same meaning can be written in many different ways. Thus leading to increase in complexity in the sentences.
These complex and difficult task is not only faced by humans but also, by googles algorithm.
Google itself uses more than 1000 languages. Hence it becomes a hurdle for the SEO itself to understand the different set of sentence in each line for more than 1000 languages.
Google always stood to give the best user experience to its users.
For this it has designed its algorithm in such a way that it should be able to provide the best answer possible for a particular question, under just a few millisecond.
Anybody who is a content writer will automatically be aware of the SEO Optimized content writing. Here it asks the content writer to make small sentence within the blog.
This itself shows the vulnerability of SEO which Google itself understands very vividly.
Although small sentences are not being used just because of the SEO but also for the users as well.
Since we humans are very impatient in nature it becomes difficult for us to read large sentences. Thus showing that Google does wants the best for its users.
Now, comes Bert. Bert is the algorithm used by google. It helps Google to understand a word naturally inside a whole sentence.
Bert considers approximately two or three words together i.e. it chooses one word and the word before and after the chosen word to extract its meaning from it.
Although Bert is powerful, it is not sufficient enough for best user experience. Indeed BETH is useful in searching results for the questions asked by the viewers but it lacks to understand the content as a whole.
Therefore to evolve from these situations SMITH algorithm can be used.
SMITH( Siamese Multi-depth Transformer-based Hierarchical Encoder)
This algorithm can easily extract meanings of big paragraphs, passages simply by understanding the relationship between the sentences.
SMITH can easily understand lengthy sentences and recommend the best articles as per the laid questions.
Smith always works in two towers
- Sentence Level- where it divides the sentences into different blocks.
- Document Level- here it tries to understand the meaning of the blocks.
We must remember SMITH is not here to replace BERT but it will be used as extension to enhance it.
It will not only help BERT in better understanding the quality of the content but also in understanding content with larger paragraphs and in better indexing.
Smith generally analyses lines and paragraphs by removing the styles and words which are irrelevant to the topic and are just placed to increase the length of the content.
Officially its hard to say that Google is applying it or not, since it never says anything openly.
But to conclude it we can say that sooner or later great content will be the king. Indeed even today content is the king but it will be given much more importance than it is today.