Elasticsearch Soundex Match Query - NEST

1k views Asked by At

Can anyone think why this may not be working? I basically have two fields which I index using a soundex analyzer see configuration below but when I search using names similar to what is stored in the index it doesn't work.

anz.Custom("soundex_analyzer", dma => dma
                    .Tokenizer("keyword")
                    .Filters("lowercase", "icu_folding", "soundex_filter"));


tk.Phonetic("soundex_filter", ph => ph.Encoder(PhoneticEncoder.RefinedSoundex).Replace(false));

[String(Name = "surnameSoundex", Index = FieldIndexOption.Analyzed, Analyzer = "soundex_analyzer")]
public string SurnameSoundex { get; set; }

[String(Name = "forenameSoundex", Index = FieldIndexOption.Analyzed, Analyzer = "soundex_analyzer")]
public string ForenameSoundex { get; set; }


if (string.IsNullOrEmpty(oReq.person.ForenameSoundex) || oReq.person.ForenameSoundex.Length < 3)
  {
    _qc = _qd.Match(mt => mt.Field(fld => fld.SurnameSoundex).Query(oReq.person.SurnameSoundex));
    _AndQueries.Add(_qc);
    _qc = null;
  }
  else
     {
      //search on surname and combination of forename and surname
       _qc = _qd.Match(mt => mt.Field(fld => fld.SurnameSoundex).Query(oReq.person.SurnameSoundex))
                        || _qd.Match(mt => mt.Field(fld => fld.SurnameSoundex).Query(oReq.person.SurnameSoundex))
                        && _qd.Match(mt => mt.Field(fld => fld.ForenameSoundex).Query(oReq.person.ForenameSoundex));
       _AndQueries.Add(_qc);
       _qc = null;
       }

There isn't anything wrong with the construction of the query as I have checked this but basically those queries are passed to a Query Container list and turned into an array can passed into the bool query.

I wasn't sure whether I am not able to use the keyword tokenizer with this analyzer.

Thanks in advance!

EDIT:

So basically I have a person class with my POCO attributes defined:

[ElasticsearchType(Name = "person", IdProperty = "id")]
public class Person
{
    [String(Name = "id", Index = FieldIndexOption.NotAnalyzed)]
    public string id { get; set; }

    [String(Name = "forename", Index = FieldIndexOption.Analyzed, Analyzer = "low_whit_analyzer", SearchAnalyzer = "low_whit_analyzer")]
    public string forename { get; set; }

    [String(Name = "forenameSoundex", Index = FieldIndexOption.Analyzed, Analyzer = "soundex_analyzer", SearchAnalyzer = "soundex_analyzer")]
    public string forenameSoundex { get; set; }

    [String(Name = "surname", Index = FieldIndexOption.Analyzed, Analyzer = "low_whit_analyzer", SearchAnalyzer = "low_whit_analyzer")]
    public string surname { get; set; }

    [String(Name = "surnameSoundex", Index = FieldIndexOption.Analyzed, Analyzer = "soundex_analyzer", SearchAnalyzer = "soundex_analyzer")]
    public string surnameSoundex { get; set; }

    [Date(Name = "dob", Index = NonStringIndexOption.NotAnalyzed, Format = "date_optional_time")]
    public DateTime dob { get; set; }

    [String(Name = "postCode1", Index = FieldIndexOption.Analyzed, Analyzer = "keyword_analyzer", NullValue = null)]
    public string postCode1 { get; set; }

    [String(Name = "postCode2", Index = FieldIndexOption.Analyzed, Analyzer = "keyword_analyzer", NullValue = null)]
    public string postCode2 { get; set; }

    [String(Name = "identifier", Index = FieldIndexOption.Analyzed, Analyzer = "low_whit_analyzer", NullValue = null)]
    public string identifier { get; set; }

    [String(Name = "email", Index = FieldIndexOption.Analyzed, Analyzer = "keyword_analyzer", NullValue = null)]
    public string email { get; set; }

    [String(Name = "mobile", Index = FieldIndexOption.Analyzed, Analyzer = "low_whit_analyzer", NullValue = null)]
    public string mobile { get; set; }

    [String(Name = "gender", Index = FieldIndexOption.Analyzed, Analyzer = "keyword_analyzer")]
    public string gender { get; set; }

    [String(Name = "notes", Index = FieldIndexOption.NotAnalyzed)]
    public string notes { get; set; }

    [String(Name = "address1", Index = FieldIndexOption.NotAnalyzed, NullValue = null)]
    public string address1 { get; set; }

    [String(Name = "address2", Index = FieldIndexOption.NotAnalyzed, NullValue = null)]
    public string address2 { get; set; }

    [String(Name = "personalReferenceId", Index = FieldIndexOption.Analyzed, Analyzer = "low_whit_analyzer")]
    public string personalReferenceId { get; set; }
}

I then create the index using the following code:

 Uri eSAddress = new Uri(ConfigurationManager.AppSettings["ElasticSearchUrl"]);
        _clientSettings = new ConnectionSettings(eSAddress)
            .MapDefaultTypeIndices(i => i.Add(typeof(Person), "people"));
        _client = new ElasticClient(_clientSettings);

        var oRequest = new IndexExistsRequest("people");
        var bIndexExists = _client.IndexExists(oRequest);

        if (bIndexExists.Exists == false)
        {
            var oIndexResponse = _client.CreateIndex("people", c => c
             .Settings(st => st
                .RefreshInterval(-1)
                .Translog((ts) => SetupTranslogSettings(ts))
                .NumberOfShards(1)
                .NumberOfReplicas(0)
                    .Analysis(an => an
                        .TokenFilters((tf) => SetUpFilters(tf))
                        .Analyzers((anz) => SetUpAnalyzers(anz)
                 )))
                .Mappings(mp => mp.Map<Person>(m => m
                .AutoMap()
                .AllField(al => al.Enabled(false)))));

I then use logstash to import my records from the database using the following configuration:

statement => "SELECT IGF_UID AS id, IGF_FORENAME AS forename, IGF_SURNAME AS surname, IGF_FORENAME AS forenameSoundex, IGF_SURNAME AS surnameSoundex, 
              IGF_DATE_OF_BIRTH AS dob, IGF_POSTCODE1 AS postCode1, IGF_POSTCODE2 AS postCode2, IGF_NHS_NUMBER AS identifier, IGF_EMAIL AS email, 
              IGF_MOBILE AS mobile, (CASE IGF_SEX
                        WHEN 'male' THEN 'm'
                        WHEN 'female' THEN 'f'
                        WHEN 'transgender' THEN 't'
                        WHEN 'unknown' THEN 'u'
                        WHEN '' THEN NULL
                        ELSE IGF_SEX
                        END) AS gender, IGF_ADDRESS1 AS address1, IGF_ADDRESS2 AS address2 FROM dbo.IGT_PEOPLE"
   }
 }
  filter {
   mutate {
    remove_field => [ "@timestamp", "@version" ]
     }
    }

 output {
elasticsearch {
    hosts => "localhost"
    index => "people"
    document_type => "person"
    document_id => "%{id}"
    manage_template => false
    template_overwrite => false
    }
}

My analyzers are included below - note I have changed to the double metaphone token filter:

 public static void AddSoundexAnalyzer(ref AnalyzersDescriptor anz)
    {
        anz.Custom("soundex_analyzer", dma => dma
                    .Tokenizer("keyword")
                    .Filters("soundex_filter"));
    }

 public static void AddSoundexFilter(ref TokenFiltersDescriptor tk)
    {
        tk.Phonetic("soundex_filter", ph => ph.Encoder(PhoneticEncoder.DoubleMetaphone).Replace(true));
    }

Then I query using a bool query with the query within a must should so should match at least one query.

  public SearchDescriptor<Person> FuzzySearch(PersonSearchRequest oReq)
    {
        var oPerson = oReq.person;
        var oSearchParams = oReq.searchParams;
        _s = new SearchDescriptor<Person>();
        _b = new BoolQueryDescriptor<Person>();
        _AndQueries = new List<QueryContainer>();
        _OrQueries = new List<QueryContainer>();

        GetNameSearchClauses(oReq, ref _OrQueries, ref _AndQueries);

        if (_OrQueries.Count > 0 || _AndQueries.Count > 0)
        {
            _b.Should(_OrQueries.ToArray());
            _b.Must(_AndQueries.ToArray());
            return _s.Query(qu => qu.Bool((z) => _b)).Sort(srt => srt.Descending(SortSpecialField.Score));
        }
        else
        {
            return null;
        }
    }

Then my forenameSoundex and surnameSoundex queries are constructed in the following method:

  public void GetNameSearchClauses(PersonSearchRequest oReq, ref List<QueryContainer> _OrQueries, ref List<QueryContainer> _AndQueries)
    {
        if (oReq.searchParams.useSoundex == true && oReq.person.surnameSoundex.Length > 3)//use different analyzers
        {
            if (!string.IsNullOrEmpty(oReq.person.surnameSoundex))
            {//check if clause is null

                //if no first name then just search on surname
                if (string.IsNullOrEmpty(oReq.person.forenameSoundex) || oReq.person.forenameSoundex.Length < 3)
                {
                    _qc = _qd.Match(mt => mt.Field(fld => fld.surnameSoundex).Query(oReq.person.surnameSoundex));
                    _AndQueries.Add(_qc);
                    _qc = null;
                }
                else
                {
                    //search on surname and combination of forename and surname
                    _qc = _qd.Match(mt => mt.Field(fld => fld.surnameSoundex).Query(oReq.person.surnameSoundex))
                        || _qd.Match(mt => mt.Field(fld => fld.surnameSoundex).Query(oReq.person.surnameSoundex))
                        && _qd.Match(mt => mt.Field(fld => fld.forenameSoundex).Query(oReq.person.forenameSoundex));
                    _AndQueries.Add(_qc);
                    _qc = null;
                }
            }
        }

}

1

There are 1 answers

0
GWilkinson On BEST ANSWER

The problem was that with using the jdbc plugin for logstash it automatically lowercases column names. So within my sql statement when I was creating an alias to map directly to the elasticsearch field name, it was then being converted to lowercase when passing through logstash.

My jdbc config needed the following line: lowercase_column_names => false