Selection of words for LSI-text

How to select words using LSI technology?

First you need to collect queries and synonyms relevant to the search phrase. According to LSI technology, you can use no more than 3 search queries in one text that are necessarily relevant to each other (toothbrushes and felt boots cannot be promoted with the same text). Moreover, key phrases do not have to be used directly. Here it is more important to correctly use the train, these same relevant and synonymous words. It is also important to keep the visitor on the page for 15 seconds, but this rule is general, not specific to LSI.

the essence of the principle:  search engines, analyzing the texts on the site, take into account not only keywords (the so-called semantic core), but also related words that reveal the topic (the so-called thematic core): synonyms, associations, hyponyms, hyperonyms, related concepts, etc. The larger and richer the thematic core, the more valuable the text for search engines looks and the higher it is ranked. That is why many SEO-specialists, when they give the authors a technical assignment (TOR), indicate not only the keys, but also the thematic dictionary – those same accompanying words for activating the LSI-factor.

Let’s look at a good example. Imagine you have a high-frequency (HF) golf request. And you are writing an article. Based on a single query, the search engine cannot understand what you are writing about: either about a game, or about a car, or about a long sock. Therefore, when indexing, the robot will begin to analyze the surrounding text (its thematic core). For example, if we have it as in the figure below, it will immediately become clear what we are talking about.

The same is true for more substantive keywords. When a user enters the query “golf rules”, one of the significant ranking factors is a well-developed thematic core to give a person an exhaustive answer to his question, reducing the time needed to find the necessary information to a minimum.

In other words, the direction of LSI copywriting is based on writing texts for requests, adding the maximum number of (related) thematic words (to work out an associative core so that it is better than competitors).

Creating an associative kernel (LSI core)

Creating an associative kernel (LSI core) of the first order

The thematic or, as it is also called, the associative core is not created immediately, but in stages. At the first stage, we take the local semantic core (list of keywords) for the text and write it out. Or we reduce to an intelligence card. It will be much easier and more visual in the future.

For example, for this article, the semantic core consists of four basic queries. There are also secondary ones, but I did not indicate them here to make it easier. I will then watch what positions they provide to the material in the search results.

Next, for each keyword (or a set of keywords, if there are many), we create a cloud of associations, synonyms, hyponyms, hyperonyms and other related words. These are the so-called first-order LSI requests. If you got enough words, and the topic is not particularly competitive, then this can be stopped.

If the queries are similar, as in this case, then the thematic kernels will largely coincide. Then it’s better to group your keywords in order to save time and save yourself from unnecessary work.

Creation of a thematic core of the second order

We have first order LSI queries. Next, we take each of these words and make our own associative mini-kernels for them. For example, for the SEO association, we get the following mini-kernel snippet. If there are words that intersect with the core of the first order, it’s not scary.

Important: the second-order kernel must be directly or indirectly related to the first-order kernel, i.e. if you include all the words that are related to SEO, but have no connection with copywriting, there will be little sense.