Endeca Relevance Ranking
Relevance Ranking Module Detailed Explanation
Exact: (exact match)
The entire result is layered. The highest layer, that is, the layer that is first presented to the user, matches the term exactly, the second layer is the layer that splits the term and matches, and the third layer is the layer of other hits.
For example, if the property sku.displayName now has Nike Shoes, Nike XM Shoes, Nike-Shoes, "Nike Shoes", etc., then the highest layer will put Nike Shoes on the first layer, and then include Nike or Shoes on the first layer. Two layers, then others like Nike-Shoes, "Nike Shoes" on the third layer.
Filed: (field matching)
It is set according to the priority of the fields in the search interface in the entire member list list, and the first one has the highest priority.
1 will rate each result
2 Scoring is based on the static filed rank of the dimension/ property member in the search interface
3 The static field rank is determined according to the order of the member list in the search interface.
For cross-filed our default score is 0, but we can also change its indexator in Developer Studio.
for example:
MatchMode: All
SearchInterface: All
Search Term: Nike Plastic
Default fields in Search Interface:
material, color, displayName. (note that material is the first, the default rank value is the highest )
CROSS_FIELD_BOUNDARY="ALWAYS/ON-FAILURE"
If the current Filed Match matches the material, it will be placed on the top layer, followed by sku.color, clor, sku.displaName, etc. What is finally presented to the user is the order results in accordance with this level.
If it matches, then
FieldMatch: matter
If it doesn't match:
Field: cross-field (null)
- stratum rank - 0
- evaluation time - 0.0012 ms
Frequency(Freq)：
In the result text, a score is made based on the number of occurrences of the user's query term, the more the higher the score.
It is generally divided into 1-N levels:
It depends on how many of these terms your result text contains. If there is only 1, then there is only one layer:
strata:1, if there are N, then stata:N
Then the ones with the most will be presented to the user first.
Glom:
Matching on a single field takes precedence over matching on multiple fields. Generally used in conjunction with the MaxFiled module. In this mode, MatchAny is ineffective for cross-fileds because each individual term
Either can be used as a single-field.
Assuming the current
Search Terms: Nike Shoes
Search Interface: all
Match Mode: matchAll
When I search, first match Match All, that is, the field contains Nike and contains Shoes, and then the relevancy is sorted.
Then for a single field ranking, the ranking is finished, and then according to whether cross-filed is enabled, if it is enabled, then glom
There are only 2 layers, and only 1 layer if not enabled. In the case of cross-field, for example, if I have two fields that contain Nike Shoes, the record will be placed in this layer if so many fields match.
It is generally considered that there are enough terms to satisfy the search, so it is usually placed after NTerms.
Interpreted：
1 Non-partial matches are ranked ahead of partials
2 Single-field match ranking takes precedence over cross-field match
3 Non-spell check ranks ahead of spell check
4 Synonym matches are ranked after non-synonym matches
5 Stemming matches are ranked behind non-stemming ones
Maximum Field
In fact, it is similar to the Field ranking module, except that there is a little difference when matching across fields:
Unlike Field, which assigns a static score to cross-filed matches, MaxField will select the field with the highest rank value for the match
Numfields
What does that mean? According to the result set, in the current matching mode, the more fields that contain the user's query term, the larger the rank value, and the higher the ranking.
for example:
RecordA:
sku.displayName:Nike Shoes
sku.description: Alert Frame Nike Shoes
RecordB:sku.displayName:Nike
Because Record A has two fields that contain the query item that the user searches. And RecordB has only one field, so RecordA ranks in front of Record.
NTerms：
It means that the query terms provided by the user, the more words they contain, the higher the ranking, and the less they are at the back. For example, Nike Shoes, the rank value of all matches is higher than that of only matching one Nike or Shoes.
Phrase
Indicates that the user's term is completely matched or partially matched as a phrase. The higher the matching degree, the higher the rank value, and the no matches are ranked last.
The Phrase options are:
• Rank based on length of subphrases
• Use approximate subphrase/phrase matching
• Apply spell correction, thesaurus, and stemming
Static
Score each result based on a number or constant value, depending on the type of search and the parameters passed to the module, the first parameter being the property or dimension, and the second parameter being the sort parameter.
For example, using the module Static(Availability,descending) would sort result records in descending
order with respect to their assignments from the Availability property. Using the module
Static(Title,ascending) would sort result records in ascending order by their Title property assignments.
Relevance Ranking strategies are used in two main contexts in the MDEX Engine: There are two ranking strategies in the MDEX Engine:
1 Configure in the search interface
2 Set it at the query level, and then override the ranking strategy of the search interface. For example, through url query, we can set parameters when querying:
Nrk=search-interface
Nrt=relrank-terms
Nrr=relrank-strategy
Nrm=relrank-matchmode
For Example:<application>?N=0&Ntk=P_Description&Ntt=sonoma
&Nrk=All&Nrt=citrus&Nrr=maxfield&Nrm=matchall