Elasticsearch : multi_match best_fields or most_fields?

Question

I'm trying to search through this type of documents :

            {
               "id_stage": "4598",
               "n_stage": "1331400117S",
               "intitule_stage": "Comment travailler avec le délire dans la psychose ? - Niveau II -",
               "objectifs": ">Poursuivre et approfondir les enseignements théoriques du stage Niveau I sur le même thème.\r\n>Savoir mettre en pratique ces connaissances par l'analyse \"active\" du discours du patient.",
               "contenu": "*Rappel de la théorie découverte lors du Niveau I : travail à  partir de cas de schizophrénie, de paranoïa et de névrose afin d'affiner la notion de clinique différentielle.\r\n*Le psychotique et sa famille :\r\n-Notion de transmission intra-familiale.\r\n-Les relations parents/enfants dans la psychose.\r\n*Les structures maniaques et mélancoliques, spécificité du fonctionnement.\r\n*Absence de sentiment de culpabilité et confusion diagnostique entre psychose et perversion. Essai d'approche différentielle.\r\n*Les psychoses blanches (sans délire apparent).\r\n",
               "pre_requis": "",
               "mmp": "",
               "sessions": [
                  {
                     "id_session": "5332",
                     "debut": "2017-12-04",
                     "date_fin": "2017-12-08",
                     "ville": "LA ROCHELLE"
                  }
               ],
               "mots_cles": [
                  "DELIRE",
                  "PSYCHOSE"
               ]
            }

and here is my query :

            {
            "from": 0,
            "size": 100,
            "min_score": 4.5,
            "query": {
              "bool": {
                "must": 
                    [
                        {
                        "multi_match": {
                            "fields": 
                            [
                                "sessions^2",
                                "contenu^1",
                                "objectifs^1",
                                "intitule_stage^3",
                                "mmp^1",
                                "mots_cles^3"
                            ],

                            "query": "' . $request . '",
                            "fuzziness" : "AUTO",
                            "prefix_length" : 2,
                            "type" : "best_fields"
                            }
                        },
                        {
                        "query_string": {
                            "fields": 
                                [
                                    "n_stage"
                                ],
                            "query": "*' . $filter_year . 'S"
                            }
                        }
                    ]
                }
            }
          }

The problem is that the search : "handicap" returns 66 results and the result "handicap La Rochelle" returns 100 results(the limit) and it should return less besause "La Rochelle" can only be present in the session field of the document

Maybe should I use something else than multi_match ?

Thanks for your help.

I've tried to add "operator":"and" and now it returns 0 results :( — Eloise85
In my example the term "handicap" could be found in the fields "contenu", "objectifs, "intitule_stage", "mmp", "mots_cles" and the 2nd term "La Rochelle" could be found in the field "sessions so what I want is that we can find all the result where "handicap" or nearly with fuzziness and the highest score is when "La Rochelle" is found in addition in the sessions field(but not mandatory). Am I clear in my explanations ? — Eloise85

Jiulin Teng Jiulin Teng · Accepted Answer · 2020-07-30T13:00:25

I believe that this is normal behavior: the more keywords you supply, the more results will come with best_fields and most other multi_match results.

The key is whether handicap La Rochelle is ranked at the top.

Elasticsearch : multi_match best_fields or most_fields?

1 Answers