如何在文本字段的弹性搜索查询中执行基于余弦相似度的语义搜索？

Question

I am performing a match on a text field(skills). I don't want a exact match, instead i want cosine similarity based search on the field.

GET 2/_search
{
  "_source": ["Skills"], 
  "query": {
    "function_score": {
      "query": {
        "match": {
          "Job_Group": "sales"
        }
      },
      "functions": [
        
        {
          "filter": {
            "match":{
              "Skills":"Designation"
            }
          },
          "weight": 15
        }
      ]
    }
    }
}

The above query is for exact match. How do i include some sort of semantic search(Cosine similarity based in the query on skills field). The skills field is a free text field, so i want matching to happen based on their semantic meaning also. Example--- skills -Communication & talking should reflect some sort of similarity and boost the score.

Answer 1

Very simple- Elastic is wrapping Lucene , and Lucene has More-Like-This which implement TF-IDF (BM25 to be more precise) and some additional wisdom. Try it, it will give you good similarity results. Explanation can be found in this link and various others.