Elasticsearch - How to decide the best mapping scheme?

Replies

• 

  Saandeep Sreerambatla

  what is the type of data that you are working on ? We need information on the business case?
  I didnt work on ES at all, but my team works on ES so much. So need more information on this so that i can go back and get the answers.
• 

  Kaustubh Katdare

  Thanks, #-Link-Snipped-# . Since posting the thread, I've learned more about ES. Right now, I'm in the 'fine-tuning' phase for @mention suggestions.
  
  Let's say a user starts typing a name. Ideally, we should return multiple matches that have the starting characters, but these matches have to be relevant. I'm therefor experimenting with 'edge-ngram' tokenizer.
  
  Here's the query I posted on a forum -
  
  Here's how my current settings look like -
  
  

  protected $settings = [
     'analysis'  => [
         'filter'    =>  [
             'autocomplete_filter'   =>  [
                 'type'  =>  'edge_ngram',
                 'min_gram'  =>  2,
                 'max_gram'  =>  12,
             ],
         ],
         'analyzer'  =>  [
             'autocomplete' =>  [
                 'type'  =>  'custom',
                 'tokenizer' =>  'standard',
                 'filter'    =>  ['lowercase', 'autocomplete_filter']
             ]
         ]
     ]
  ];
  

  and
  

  public function buildQueryPayload()
  {
     return [
         'must'  =>  [
             'multi_match'   =>  [
                 'query'     => $this->builder->query,
                 'fields'    => ['first_name', 'last_name', 'name^2'],
                 'type'      =>  'phrase_prefix'
             ]
         ]
     ];
  }
  

  This is my best try so far; but I'm not happy with the results. I've the following in my User model -
  
  

  public function toSearchableArray() {
     return [
         'first_name'    =>  $this->first_name,
         'last_name'     =>  $this->last_name,
         'name'          =>  $this->first_name . ' ' . $this->last_name
     ];
  }
  

  I'm thinking of adding an extra field from my database to the index called last_activity_at and then sort the suggested users according to those who've the most recent last_activity.
  
  Can someone suggest if I need to make changes to my filter, analyzer or the query? The end result I want to achieve is -
  
  User types: @Joh . The system should return
  

  1. 'John Doe' [most recently active]
  2. 'John Doe' [second most recently active]
  3. 'Johan Kent'
  4. 'Johar Woe'

  ... like that. If the last_activity_at field is null, ES should only match based upon the closest search.
  
  Would welcome suggestions.