I'm a bit surprised by the behavior that elasticsearch completion has sometimes. I've set up a mapping that has a suggest field. In the input of the suggest field I put 3 elements that are the name, the isin and the issuer of one security.
Here is the mapping that I use :
"suggest": {
"type" : "completion",
"analyzer" : "simple"
}
When I want to query my index with this query :
{
"suggest": {
"my_suggestion": {
"prefix": "FR0011597335",
"completion": {
"field": "suggest"
}
}
}
}
I get a list of results but not necessarilly with my exact prefix and most of the time with the exact match not at the top.
So I'd like to know if there is a way to boost exact matches in a suggestion and make such exact term matches to be in first position when possible.
I think my problem is solved by using a custom analyzer : the simple one was not convenient for the entries I had.
"settings": {
"analysis": {
"char_filter": {
"punctuation": {
"type": "mapping",
"mappings": [".=>"]
}
},
"filter": {},
"analyzer": {
"analyzer_text": {
"tokenizer": "standard",
"char_filter": ["punctuation"],
"filter": ["lowercase", "asciifolding"]
}
}
}
},
and
"suggest": {
"type" : "completion",
"analyzer" : "analyzer_text"
}
Related
I am using Elastic Search 6.This is query
PUT /semtesttest
{
"settings": {
"index" : {
"analysis" : {
"filter": {
"my_stop": {
"type": "stop",
"stopwords_path": "analysis1/stopwords.csv"
},
"synonym" : {
"type" : "synonym",
"synonyms_path" : "analysis1/synonym.txt"
}
},
"analyzer" : {
"my_analyzer" : {
"tokenizer" : "standard",
"filter" : ["synonym","my_stop"]
}
}
}
}
},
"mappings": {
"all_questions": {
"dynamic": "strict",
"properties": {
"kbaid":{
"type": "integer"
},
"answer":{
"type": "text"
},
"question": {
"type": "text",
"analyzer": "my_analyzer"
}
}
}
}
}
PUT /semtesttest/all_questions/1
{
"question":"this is hippie"
}
GET /semtesttest/all_questions/_search
{
"query":{
"fuzzy":{"question":{"value":"hippie","fuzziness":2}}
}
}
GET /semtesttest/all_questions/_search
{
"query":{
"fuzzy":{"question":{"value":"this is","fuzziness":2}}
}
}
in synonym.txt it is
this, that, money => sainai
in stopwords.csv it is
hello
how
are
you
The first get ('hippie') return empty
only the second get ('this is') return results
what is the problem? It looks like the stop word "this is" is filtered in the first query, but I have specified my stop words explicitly?
fuzzy is a term query. It is not going to analyze the input, so your query was looking for the exact term this is (applying some fuzzy fun).
So you either want to build a query off those two terms, or use a full text query instead. If fuzziness is important, I think the only full text query is match:
GET /semtesttest/all_questions/_search?pretty
{
"query":{
"match":{"question":{"query":"this is","fuzziness":2}}
}
}
If match phrases is important, you may want to look at this answer and work with span queries.
This might also help you so you can see how your analyzer is being used:
GET /semtesttest/_analyze?analyzer=my_analyzer&field=question&text=this is
I am indexing all the names on a web page with characters with accents like "José". I want to be able to search the this name with "Jose" and "José".
How should I set up my index mapping and analyzer(s) for a simple index with one field "name"?
I set up an analyzer for the name field like this:
"analyzer": {
"folding": {
"tokenizer": "standard",
"filter": ["lowercase", "asciifolding"]
}
}
But it folds all accents into ASCII equivalents and ignores the accent when indexing the "é". I want the "é" char to be in the index and I want to be able to search "José" with either "José" or "Jose".
You need to preserve the original token with the accent. To achieve that you need to redefine your own asciifolding token filter, like this:
PUT /my_index
{
"settings" : {
"analysis" : {
"analyzer" : {
"folding" : {
"tokenizer" : "standard",
"filter" : ["lowercase", "my_ascii_folding"]
}
},
"filter" : {
"my_ascii_folding" : {
"type" : "asciifolding",
"preserve_original" : true
}
}
}
},
"mappings": {
"my_type": {
"properties": {
"name": {
"type": "text",
"analyzer": "folding"
}
}
}
}
}
After that, both tokens jose and josé will be indexed and searchable
This is what I can think of to resolve the folding problem with diacritical marks:
Analyzer used:
{
"settings": {
"analysis": {
"analyzer": {
"folding": {
"tokenizer": "standard",
"filter": [ "lowercase", "asciifolding" ]
}
}
}
}
}
Below is the mapping to be used:
mappings used:
{
"properties": {
"title": {
"type": "string",
"analyzer": "standard",
"fields": {
"folded": {
"type": "string",
"analyzer": "folding"
}
}
}
}
}
The title field uses the standard analyzer and will contain the original word with diacritics in place.
The title.folded field uses the folding analyzer, which strips the diacritical marks.
Below is the search query I will use:
{
"query": {
"multi_match": {
"type": "most_fields",
"query": "esta loca",
"fields": [ "title", "title.folded" ]
}
}
}
I have the following elasticsearch settings:
"settings": {
"index":{
"analysis":{
"analyzer":{
"analyzer_keyword":{
"tokenizer":"keyword",
"filter":["lowercase", "asciifolding"]
}
}
}
}
}
The above works fine for the following keywords:
Beyoncé
Céline Dion
The above data is stored in elasticsearch as beyonce and celine dion respectively.
I can search for Celine or Celine Dion without the accent and I get the same results. However, the moment I search for Céline, I don't get any results. How can I configure elasticsearch to search for partial keywords with the accent?
The query body looks like:
{
"track_scores": true,
"query": {
"bool": {
"must": [
{
"multi_match": {
"fields": ["name"],
"type": "phrase",
"query": "Céline"
}
}
]
}
}
}
and the mapping is
"mappings" : {
"artist" : {
"properties" : {
"name" : {
"type" : "string",
"fields" : {
"orig" : {
"type" : "string",
"index" : "not_analyzed"
},
"simple" : {
"type" : "string",
"analyzer" : "analyzer_keyword"
}
},
}
I would suggest this mapping and then go from there:
{
"settings": {
"index": {
"analysis": {
"analyzer": {
"analyzer_keyword": {
"tokenizer": "whitespace",
"filter": [
"lowercase",
"asciifolding"
]
}
}
}
}
},
"mappings": {
"test": {
"properties": {
"name": {
"type": "string",
"analyzer": "analyzer_keyword"
}
}
}
}
}
Confirm that the same analyzer is getting used at query time. Here are some possible reasons why that might not be happening:
you specify a separate analyzer at query time on purpose that is not performing similar analysis
you are using a term or terms query for which no analyzer is applied (See Term Query and the section title "Why doesn’t the term query match my document?")
you are using a query_string query (E.g. see Simple Query String Query) - I have found that if you specify multiple fields with different analyzers and so I have needed to separate the fields into separate queries and specify the analyzer parameter (working with version 2.0)
I have very simple query:
POST /indexX/document/_search
{
"query": {
"match_phrase_prefix": {
"surname": "grab"
}
}
}
with mapping:
"surname": {
"type": "string",
"analyzer": "polish",
"copy_to": [
"full_name"
]
}
and definition for index (I use Stempel (Polish) Analysis for Elasticsearch plugin):
POST /indexX
{
"settings": {
"index": {
"analysis": {
"filter": {
"synonym" : {
"type": "synonym",
"synonyms_path": "analysis/synonyms.txt"
},
"polish_stop": {
"type": "stop",
"stopwords_path": "analysis/stopwords.txt"
},
"polish_my_stem": {
"type": "stemmer",
"rules_path": "analysis/stems.txt"
}
},
"analyzer": {
"polish_with_synonym": {
"tokenizer": "standard",
"filter": [
"synonym",
"lowercase",
"polish_stop",
"polish_stem",
"polish_my_stem"
]
}
}
}
}
}
}
For this query I get zero results. When I change phrase to GRA or GRABA it returns 1 result (GRABARZ is the surname). Why is this happening?
I tried max_expansions with values even as high as 1200 and that didn't help.
At the first glance, your analyzer stems the search term ("grab") and renders it unusable ("grabić").
Without going into details on how to resolve this, please consider getting rid of polish analyzer here. We are talking about people's names, not "ordinary" polish words.
I saw different techniques used in this case: multi-field searches, fuzzy searches, phonetic searches, dedicated plugins.
Some links:
https://www.elastic.co/blog/multi-field-search-just-got-better
http://www.basistech.com/fuzzy-search-names-in-elasticsearch/
https://www.found.no/play/gist/6c6434c9c638a8596efa
But I guess in case of polish names some kind of prefix query on non-analyzed field would suffice...
I want to use elasticsearch for multi-word searches, where all the fields are checked in a document with the assigned analyzers.
So if I have a mapping:
{
"settings": {
"analysis": {
"analyzer": {
"folding": {
"tokenizer": "standard",
"filter": [ "lowercase", "asciifolding" ]
}
}
}
},
"mappings" : {
"typeName" :{
"date_detection": false,
"properties" : {
"stringfield" : {
"type" : "string",
"index" : "folding"
},
"numberfield" : {
"type" : "multi_field",
"fields" : {
"numberfield" : {"type" : "double"},
"untouched" : {"type" : "string", "index" : "not_analyzed"}
}
},
"datefield" : {
"type" : "multi_field",
"fields" : {
"datefield" : {"type" : "date", "format": "dd/MM/yyyy||yyyy-MM-dd"},
"untouched" : {"type" : "string", "index" : "not_analyzed"}
}
}
}
}
}
}
As you see I have different types of fields, but I do know the structure.
What I want to do is starting a search with a string to check all fields using the analyzers too.
For example if the query string is:
John Smith 2014-10-02 300.00
I want to search for "John", "Smith", "2014-10-02" and "300.00" in all the fields, calculating the relevance score as well. The better solution is the one that have more field matches in a single document.
So far I was able to search in all the fields by using multi_field, but in that case I was not able to parse 300.00, since 300 was stored in the string part of multi_field.
If I was searching in "_all" field, then no analyzer was used.
How should I modify my mapping or my queries to be able to do a multi-word search, where dates and numbers are recognized in the multi-word query string?
Now when I do a search, error occurs, since the whole string cannot be parsed as a number or a date. And if I use the string representation of the multi_search then 300.00 will not be a result, since the string representation is 300.
(what I would like is similar to google search, where dates, numbers and strings are recognized in a multi-word query)
Any ideas?
Thanks!
Using whitespace as filter in analyzer and then applying this analyzer as search_analyzer to fields in mapping will split query in parts and each of them would be applied to index to find the best matching. And using ngram for index_analyzer would very improve results.
I am using following setup for query:
"query": {
"multi_match": {
"query": "sample query",
"fuzziness": "AUTO",
"fields": [
"title",
"subtitle",
]
}
}
And for mappings and settings:
{
"settings" : {
"analysis": {
"analyzer": {
"autocomplete": {
"type": "custom",
"tokenizer": "whitespace",
"filter": [
"standard",
"lowercase",
"ngram"
]
}
},
"filter": {
"ngram": {
"type": "ngram",
"min_gram": 2,
"max_gram": 15
}
}
},
"mappings": {
"title": {
"type": "string",
"search_analyzer": "whitespace",
"index_analyzer": "autocomplete"
},
"subtitle": {
"type": "string"
}
}
}
See following answer and article for more details.