Hi
Im using ES 2.3. I have a phrase suggester that's returning suggestions with a higher score than the searched text even though the searched text is in the index (suggest dictionary). Is this expected ? (although that's not what I need)
I have 1000's of documents with "Blue" in the name and couple of documents with "Blur". When I run the phrase suggester, I get "Blue" with a higher score than "Blur". I tried with "confidence:0.0" to actually see the exact score of blur since it was not returned with the default confidence.
This is my analyzer -
"simple_trigram": {
"type": "custom",
"tokenizer": "standard",
"filter": ["ascii_folding_preserve", "lowercase", "trim", "shingle_for_spellcheck"],
"char_filter" : ["punctuation_char_mapping"]
}
And this is shingle filter -
"shingle_for_spellcheck": {
"type": "shingle",
"min_shingle_size": 2,
"max_shingle_size": 3,
"output_unigrams": true
}
And this is how I query -
"ngram": {
"text": "blur",
"phrase": {
"field": "spell_check_field_en-us",
"confidence":0.0,
"direct_generator": [
{
"field": "spell_check_field_en-us",
"max_edits":1
}
]
}
}
Response : There are two hits for "blur" match query, however, I see these suggestions which I don't expect.
"ngram": [
{
"text": "blur",
"offset": 0,
"length": 4,
"options": [
{
"text": "blue",
"score": 0.033866636
},
{
"text": "blu",
"score": 0.010273874
},
{
"text": "blut",
"score": 0.0074840933
},
{
"text": "blur",
"score": 0.006959285
},
{
"text": "blum",
"score": 0.0062757228
}
]
}
]