如何利用Elasticsearch Suggesters实现文本字段前缀词搜索建议？

Got it, let's tackle this problem step by step. You want to pull prefix-matching suggestions from individual words in your text field—like getting "af, aff, afff" when searching for the prefix "af". Elasticsearch has a perfect tool for this: the Completion Suggester, which is built specifically for fast, prefix-driven suggestion use cases.

Step 1: Set Up the Index Mapping #

First, we'll add a dedicated completion type field to your index. This field is optimized for suggestion queries and will store the individual words from your text field as separate suggestion candidates.

PUT /your_index_name
{
  "mappings": {
    "properties": {
      "text": {
        "type": "text" // Keep your original text field if you need to retain the full content
      },
      "word_suggest": {
        "type": "completion",
        "preserve_separators": false, // Disable separator handling since we're working with single words
        "preserve_position_increments": false
      }
    }
  }
}

Step 2: Auto-Split Text into Suggestion Candidates #

Instead of manually splitting your text field into words when indexing, use an Ingest Pipeline to automate this. This pipeline will split the text by whitespace and populate the word_suggest field with an array of individual words.

PUT /_ingest/pipeline/split_text_to_suggestions
{
  "processors": [
    {
      "split": {
        "field": "text",
        "separator": "\\s+", // Split on one or more whitespace characters
        "target_field": "word_suggest"
      }
    }
  ]
}

Step 3: Index Your Data #

When adding documents to your index, specify the pipeline to auto-populate the suggestion field:

POST /your_index_name/_doc/1?pipeline=split_text_to_suggestions
{
  "text": "af bf aff bff"
}

POST /your_index_name/_doc/2?pipeline=split_text_to_suggestions
{
  "text": "afff bfff "
}

Step 4: Run the Suggestion Query #

Now you can query for prefix-based suggestions. We'll use skip_duplicates to ensure we don't get repeated words in the results.

POST /your_index_name/_search
{
  "suggest": {
    "prefix_suggestions": {
      "prefix": "af",
      "completion": {
        "field": "word_suggest",
        "skip_duplicates": true,
        "size": 10 // Adjust this number based on how many suggestions you want to return
      }
    }
  }
}

Sample Response #

You'll get a response containing exactly the suggestions you want:

{
  ...
  "suggest": {
    "prefix_suggestions": [
      {
        "text": "af",
        "offset": 0,
        "length": 2,
        "options": [
          {"text": "af", "_score": 1.0},
          {"text": "aff", "_score": 1.0},
          {"text": "afff", "_score": 1.0}
        ]
      }
    ]
  }
}

Alternative: Term Suggester (For Small Datasets) #

If you're working with a small dataset and don't need the performance boost of the Completion Suggester, you can use the Term Suggester directly with your existing text field. It scans the inverted index for prefix-matching terms:

POST /your_index_name/_search
{
  "suggest": {
    "prefix_suggestions": {
      "prefix": "af",
      "term": {
        "field": "text",
        "sort": "frequency" // Sort suggestions by how often they appear in your data
      }
    }
  }
}

This works too, but it's slower for large datasets compared to the Completion Suggester's optimized data structure.

内容的提问来源于stack exchange，提问作者yaroslavTir