# Unexpected Shingle Behaviour

**URL:** <https://discuss.elastic.co/t/unexpected-shingle-behaviour/265461>\
**Category:** Elasticsearch\
**Created:** [February 25, 2021, 9:48am UTC](https://discuss.elastic.co/t/unexpected-shingle-behaviour/265461 "2021-02-25T09:48:17Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![rishabhm12](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/rishabhm12/32/84524_2.png) [@rishabhm12](https://discuss.elastic.co/u/rishabhm12)\
**Post date:** [February 25, 2021, 9:48am UTC](https://discuss.elastic.co/t/unexpected-shingle-behaviour/265461/1 "2021-02-25T09:48:17Z")

</div>

Hey,  
I have a field locality in my index having `'ashok vihar phase 2'` as one of the document. The corresponding setting is

```auto
{"settings": {
       "analysis": {
           "analyzer": {
               "custom_analyzer": {
                   "tokenizer": "standard",
                   "filter": ["shingle_filter", "remove_duplicates"],
               }
           },
           "filter": {
               "shingle_filter": {
                   "type": "shingle",
                   "min_shingle_size": 2,
                   "max_shingle_size": 4,
                   "output_unigrams": False,
                   "output_unigrams_if_no_shingles": True,
               }
           }
}
}

```

So it should ideally create 6 shingles i.e. `'ashok vihar, 'ashok vihar phase', 'ashok vihar phase 2', 'vihar phase'` and so on.

When my input search is: `'ashok vihar 2'`

and I use the explain-api to see how is it maching I get:

```auto
{
                        "value": 4.618802,
                        "description": "sum of:",
                        "details": [
                            {
                                "value": 4.618802,
                                "description": "weight(Synonym(locality_shingle:ashok vihar locality_shingle:ashok vihar 2) in 89) [PerFieldSimilarity], result of:",
                                "details": [
                                    {
                                        "value": 4.618802,
                                        "description": "score from ScriptedSimilarity(weightScript=[null], script=[Script{type=inline, lang='painless', idOrCode='double norm = 1.0/Math.sqrt(doc.length); return query.boost * norm;', options={}, params={}}]) computed from:",
                                        "details": [
                                            {
                                                "value": 1.0,
                                                "description": "weight",
                                                "details": []
                                            },
                                            {
                                                "value": 8.0,
                                                "description": "query.boost",
                                                "details": []
                                            },
                                            {
                                                "value": 17042,
                                                "description": "field.docCount",
                                                "details": []
                                            },
                                            {
                                                "value": 27606,
                                                "description": "field.sumDocFreq",
                                                "details": []
                                            },
                                            {
                                                "value": 27606,
                                                "description": "field.sumTotalTermFreq",
                                                "details": []
                                            },
                                            {
                                                "value": 111,
                                                "description": "term.docFreq",
                                                "details": []
                                            },
                                            {
                                                "value": 111,
                                                "description": "term.totalTermFreq",
                                                "details": []
                                            },
                                            {
                                                "value": 1.0,
                                                "description": "doc.freq",
                                                "details": []
                                            },
                                            {
                                                "value": 3,
                                                "description": "doc.length",
                                                "details": []
                                            }
                                        ]
                                    }

```

It creates weight(Synonym(locality\_shingle: ashok vihar)). I'm unable to understand how's the shingle matching working and also how the doc\_length turns out to be 3. Also despite of creating a separate field for shingles it seems to be using the lucene SynonymQuery .

---

<div class="post-metadata">

**Author:** ![system](https://us1.discourse-cdn.com/elastic/original/3X/1/a/1ac57faf039f6b580b3f104ef42a2a89e41014de.png) [@system](https://discuss.elastic.co/u/system)\
**Post date:** [March 25, 2021, 9:48am UTC](https://discuss.elastic.co/t/unexpected-shingle-behaviour/265461/2 "2021-03-25T09:48:46Z")

</div>

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.