Lecture
Quorum – a certain minimum number of query words that must be present in a document for the system to start digging into its (word) coordinates.
Qualified quorum - the minimum total weight of the present terms.
The search engine results page (SERP) or search results — a web page generated by a search engine in response to a user's search query , using the processes of indexing, ranking, and the query's qualified quorum.
The quorum can be given as a function, or as a table:
Query computation:
1. Select documents that contain query terms whose combined weight is greater than
the threshold value.
2. Among the present terms in the document, select the most significant ones.
3. Find the cores of the tuples.
4. Decorate with less significant words.
As is known, for a document to appear in the results for each query, it must reach (pass) a certain quorum.
Quorum — the required share of the total weight (IDF) of the words from the search query that must be present in the text of the document and/or in the texts of the links pointing to it in order for it to appear in the search results (SERP).
As follows from the definition, for a document to appear in the search results, it must contain — within itself and/or in the texts of the links pointing to it — all or a specified minimum share of the weight of the words from the query. The share is calculated as a function of the length of the query (in words) and the weights of the words in it, according to the formula shown below (Fig.1), where:
Fig. 1. Formula for the quorum (weight share) from a presentation by one of Yandex's developers — Denis Raskovalov. The formula has been known for much longer; in particular, it also appears in a 2004 paper by the developers.
The numeric parameter 0.01 in the formula is called the softness and can change depending on the search engine's settings. There are certain grounds to believe that the softness value at Yandex may differ from 0.01 and take the value 0.06. For these two values, the minimum weight share (quorum) that must be found for a document to be included in the search results was calculated (Fig. 2).
Fig. 2. Calculated Quorum value for two softness values — 0.01 (blue bars) and 0.06 (red bars) — depending on the length of the query in words. On the Y axis — the minimum share of the total weight needed to pass the quorum, on the X axis — the number of words in the query, from 1 to 10.
As can be seen from the histogram, for a document to be included in the ranking for a five-word search query (with a softness coefficient of 0.06), it is enough for it to contain 4 words from the query (assuming equal weights for all words in the query). Moreover, the rules for passing the quorum can change depending on the query and the number of documents found for it.
Thus, we arrive at the first possible case, in which the text of the document and/or the texts of the links pointing to it contain not all the words from the query, but only part of them, and this part turns out to be sufficient to pass the quorum. An example is shown below (Fig. 3):
Fig. 3. Demonstration of a document passing the quorum in Yandex with a word that is deliberately absent from its text and from the link anchors.
[[b6535]]
Comments