One way Elasticsearch synonyms

Question

I'm currently facing the same issue with the synonym analyzer.

Here is the behavior I'd like to achieve:

• Query for "cable" will return the results for "cable" and "cordon"
• Query for "cordon" will return only the results for "cordon".

For the moment, the queries below return all the results with "cable" and "cordon"

Query #1 with "cable":

{
    "query": {
        "bool": {
            "must": [
                {
                    "bool": {
                        "should": [
                            {
                                "simple_query_string": {
                                    "query": "cable",
                                    "analyzer": "french",
                                    "fields": [
                                        "label"
                                    ]
                                }
                            }
                        ]
                    }
                }
            ]
        }
    },
    "sort": {
        "_score": {
            "order": "desc"
        }
    }
}

Query #2 with "cordon"

{
    "query": {
        "bool": {
            "must": [
                {
                    "bool": {
                        "should": [
                            {
                                "simple_query_string": {
                                    "query": "cordon",
                                    "analyzer": "french",
                                    "fields": [
                                        "label"
                                    ]
                                }
                            }
                        ]
                    }
                }
            ]
        }
    },
    "sort": {
        "_score": {
            "order": "desc"
        }
    }
}

The index settings:

{
    "lifecycle": {
        "name": "pdb-policy"
    },
    "routing": {
        "allocation": {
            "include": {
                "_tier_preference": "data_content"
            }
        }
    },
    "mapping": {
        "total_fields": {
            "limit": "10000"
        }
    },
    "number_of_shards": "1",
    "provided_name": "products2",
    "creation_date": "1719059028223",
    "analysis": {
        "filter": {
            "alphanumeric": {
                "pattern": "[^a-zA-Z0-9]",
                "type": "pattern_replace",
                "replacement": ""
            },
            "french_stop": {
                "type": "stop",
                "stopwords": "_french_"
            },
            "limit1term": {
                "type": "limit",
                "max_token_count": "1"
            },
            "synonym_monodirectional": {
                "type": "synonym",
                "synonyms_path": "synonyms\/words.txt",
                "expand": "false"
            },
            "french_elision": {
                "type": "elision",
                "articles": [
                    "l",
                    "m",
                    "t",
                    "qu",
                    "n",
                    "s",
                    "j",
                    "d",
                    "c",
                    "jusqu",
                    "quoiqu",
                    "lorsqu",
                    "puisqu"
                ]
            },
            "limit3term": {
                "type": "limit",
                "max_token_count": "3"
            },
            "french_stemmer": {
                "name": "french",
                "type": "stemmer"
            }
        },
        "normalizer": {
            "transliterator": {
                "filter": [
                    "asciifolding",
                    "lowercase"
                ]
            }
        },
        "analyzer": {
            "space_addition": {
                "filter": [
                    "asciifolding",
                    "french_elision",
                    "lowercase",
                    "synonym_monodirectional",
                    "french_stop"
                ],
                "tokenizer": "standard"
            },
            "prefix1term": {
                "filter": [
                    "asciifolding",
                    "french_elision",
                    "lowercase",
                    "synonym_monodirectional",
                    "french_stop",
                    "limit1term"
                ],
                "tokenizer": "standard"
            },
            "space_removal": {
                "filter": [
                    "asciifolding",
                    "french_elision",
                    "lowercase",
                    "synonym_monodirectional",
                    "french_stop"
                ],
                "tokenizer": "standard"
            },
            "french": {
                "filter": [
                    "synonym_monodirectional"
                ],
                "tokenizer": "standard"
            },
            "prefix3term": {
                "filter": [
                    "asciifolding",
                    "french_elision",
                    "lowercase",
                    "synonym_monodirectional",
                    "french_stop",
                    "limit3term"
                ],
                "tokenizer": "standard"
            },
            "transliterator": {
                "filter": [
                    "asciifolding",
                    "lowercase"
                ],
                "tokenizer": "keyword"
            }
        }
    },
    "priority": "100",
    "number_of_replicas": "0",
    "uuid": "ePDgepPcTm-NLcaePn6f3Q",
    "version": {
        "created": "7170999"
    }
}

The synonyms file:

cable => cordon

The analyzers seem to be correct too:

{
    "tokens": [
        {
            "token": "cordon",
            "start_offset": 0,
            "end_offset": 6,
            "type": "<ALPHANUM>",
            "position": 0
        }
    ]
}

{
    "tokens": [
        {
            "token": "cordon",
            "start_offset": 0,
            "end_offset": 5,
            "type": "SYNONYM",
            "position": 0
        }
    ]
}

Do you have any idea? Thank you for your help.

Answer 1

There are two issues here. The first issues is that your "french" analyzer is basically just a standard tokenizer with synonyms. It will not even lowercase tokens. See https://stackoverflow.com/a/78634671/783043 for detailed discussion.

The second issue is that during search we need to expand cable into both cable and cardon instead of just cordon. There are several ways to do it depending on the types of synonyms, but the simplest way would probably just adding the original token to the mapping. If we combine both solutions for both issues together we will get something like this:

DELETE test

PUT test
{
  "mappings": {
    "properties": {
      "label": {
        "type": "text", 
        "analyzer": "french_index",
        "search_analyzer": "french_search"
      }
    }
  }, 
  "settings": {
    "analysis": {
      "filter": {
        "french_elision": {
          "type":         "elision",
          "articles_case": true,
          "articles": [
              "l", "m", "t", "qu", "n", "s",
              "j", "d", "c", "jusqu", "quoiqu",
              "lorsqu", "puisqu"
            ]
        },
        "french_stop": {
          "type":       "stop",
          "stopwords":  "_french_" 
        },
        "french_stemmer": {
          "type":       "stemmer",
          "language":   "light_french"
        },
        "synonym_monodirectional": {
          "type": "synonym",
          "synonyms": ["cable => cable,cordon"],
          "expand": "true"
        }
      },
      "analyzer": {
        "french_search": {
          "tokenizer":  "standard",
          "filter": [
            "french_elision",
            "lowercase",
            "french_stop",
            "synonym_monodirectional",
            "french_stemmer"
          ]
        },
        "french_index": {
          "tokenizer":  "standard",
          "filter": [
            "french_elision",
            "lowercase",
            "french_stop",
            "french_stemmer"
          ]
        }
      }
    }
  }
}

POST test/_bulk?refresh=true
{ "index": { "_id": "1" } }
{ "label": "This is a cable."}
{ "index": { "_id": "2" } }
{ "label": "C'est un cordon."}

POST test/_search
{
    "query": {
        "simple_query_string": {
            "query": "cable",
            "fields": [
                "label"
            ]
        }
    }
}

POST test/_search
{
    "query": {
        "simple_query_string": {
            "query": "cordon",
            "fields": [
                "label"
            ]
        }
    }
}