Strange query result with the Polish analyzer

Elastic Stack Elasticsearch
1

I got a strange query result with the Polish analyzer (Stempel (Polish) Analysis plugin)

Here is the issue :

Use the Polish analyzer to index a document that contains in one of the fields the text "SIERPOWY Baton Energia #girl weganskie cudo #girl" :

{
	"_index" : "fts-polish",
	"_type" : "Document",
	"_id" : "bool3",
	"_source" : {
		"_analyzer" : "polish",
		"message" : "SIERPOWY 7:19 Baton Energia #girl weganskie cudo #girl"
	}
}

Run the following query :

{
	"from" : 0,
	"size" : 50,
	"query" : {
		"bool" : {
			"should" : [{
					"match" : {
						"_all" : {
							"query" : "value",
							"type" : "boolean"
						}
					}
				}
			]
		}
	},
	"highlight" : {
		"pre_tags" : [
			"<b>"
		],
		"post_tags" : [
			"</b>"
		],
		"fragment_size" : 0,
		"number_of_fragments" : 0,
		"fields" : {
			"*" : {}
		}
	}
}

You expect not to find anything, because the word "value" is not in the text, but you get the record that you indexed above, with the word "girl" highlighted :

{
	"took" : 11,
	"timed_out" : false,
	"_shards" : {
		"total" : 170,
		"successful" : 165,
		"failed" : 0
	},
	"hits" : {
		"total" : 1,
		"max_score" : 0.13561106,
		"hits" : [{
				"_index" : "fts-polish",
				"_type" : "Document",
				"_id" : "bool2",
				"_score" : 0.13561106,
				"_source" : {
					"_analyzer" : "polish",
					"streamId" : 1,
					"language" : "Polish",
					"message" : "SIERPOWY Baton Energia #girl weganskie cudo #girl"
				},
				"highlight" : {
					"message" : [
						"SIERPOWY Baton Energia #<b>girl</b> weganskie cudo #<b>girl</b>"
					]
				}
			}
		]
	}
}

Any ideas?

2

Both girl and value are translated into the same token ć:

curl "localhost:9200/_analyze?analyzer=polish&pretty" -d "girl" 
{
  "tokens" : [ {
    "token" : "ć",
    "start_offset" : 0,
    "end_offset" : 4,
    "type" : "<ALPHANUM>",
    "position" : 1
  } ]

curl "localhost:9200/_analyze?analyzer=polish&pretty" -d "value"    
{
  "tokens" : [ {
    "token" : "ć",
    "start_offset" : 0,
    "end_offset" : 5,
    "type" : "<ALPHANUM>",
    "position" : 1
  } ]
}

It looks like it's another variation of https://github.com/elastic/elasticsearch-analysis-stempel/issues/31

3

Thank you, Igor.
I received a similar answer on github: strange query result

4

@drippel Did you managed to find solution? I have the same problem, but both pasted links are dead now.

5

@jw404: There seems to be no solution. This looks like an inherent limitation of the plugin.