How to define specific field tokenization on Logstash - elasticsearch

Question

I am using logstash to index some mysql data on elasticsearch:
input {
jdbc {
// JDBC configurations
}
}
output {
elasticsearch {
index => ""
document_type => ""
document_id => ""
hosts => [ "" ]
}
}
When checking results I found that elasticsearch automatically tokenizes the text like this:
"Foo/Bar" -> "Foo", "Bar"
"The thing" -> "The", "thing"
"Fork, Knife" -> "Fork", "Knife"
Well, that is ok for most of my fields. But there is one specific field that I'd like to have a custom tokenizer. It is a comma separated field (or semi-colon separated). So it should be:
"Foo/Bar" -> "Foo/Bar"
"The thing" -> "The thing"
"Fork, Knife" -> "Fork", "Knife"
I wander if there is a way to configure this on my logstash configuration.
UPDATE:
This is one example of the index that I have. The specific field is kind:
{
"index-name": {
"aliases": {},
"mappings": {
"My-type": {
"properties": {
"#timestamp": {
"type": "date",
"format": "strict_date_optional_time||epoch_millis"
},
"#version": {
"type": "string"
},
"kind": {
"type": "string"
},
"id": {
"type": "long"
},
"text": {
"type": "string"
},
"version": {
"type": "string"
}
}
}
},
"settings": {
"index": {
"creation_date": "",
"number_of_shards": "",
"number_of_replicas": "",
"uuid": "",
"version": {
"created": ""
}
}
},
"warmers": {}
}
}

Answer 1

It's possible to do so by using an index template.
First delete your current index:
DELETE index_name
Then create the template for your index with the appropriate mapping for the kind field, like this:
PUT _template/index_name
{
"template": "index-name",
"mappings": {
"My-type": {
"properties": {
"#timestamp": {
"type": "date",
"format": "strict_date_optional_time||epoch_millis"
},
"#version": {
"type": "string"
},
"kind": {
"type": "string",
"index": "not_analyzed"
},
"id": {
"type": "long"
},
"text": {
"type": "string"
},
"version": {
"type": "string"
}
}
}
}
}
Then you can run Logstash again and the index will be re-created with the proper mapping.

Answer 2

Well, the right answer for this question is: you cannot do it by logstash. So I had to add an additional step as follow.
I finally got this done following the path showed by #Val. Thanks, pal. So, what I had to do was to create the index before the logstash ETL with a specific tokenizer:
{
"settings": {
"analysis": {
"analyzer": {
"simple_analyzer": {
"tokenizer": "simple_tokenizer"
}
},
"tokenizer": {
"simple_tokenizer": {
"type": "pattern",
"pattern": ","
}
}
}
},
"template": "my-index",
"mappings": {
"my-type": {
"properties": {
"kind": {
"type": "string",
"analyzer": "simple_analyzer"
}
}
}
}
}
This will create a tokenizer by comma to the kind field. After that, I can perform the logstash etl and it won't overwrite the kind properties.