Phrase Extraction Jobs

Table of Contents

Minimum configuration
Output documents
Output fields
Configuration properties

Identify multi-word phrases in signals.

Default job name

COLLECTION_NAME_phrase_extraction

Input

Raw signals (the COLLECTION_NAME_signals collection by default)

Output

Extracted phrases (the COLLECTION_NAME_query_rewrite_staging collection by default)

query

count_i

type

timstamp_tdt

user_id

doc_id

session_id

fusion_query_id

Required signals fields:

This job writes to the COLLECTION_NAME_query_rewrite_staging collection. It also uses reviewed documents from that collection to improve the accuracy of the job. You can review, edit, deploy, or delete output from this job using the Query Rewriting.

Fusion ships with the OpenNLP Maxent model already loaded in the blob store.

This job’s output, and output from the Token and Phrase Spell Correction job, can be used as input for the Synonym Detection job.

Minimum configuration

For most use cases, the minimum configuration for this job consists of these fields:

id/Spark Job ID

Give this job an arbitrary ID string.
trainingCollection/Training Collection

Specify the input collection.
fieldToVectorize/Field to Vectorize

Specify the field in the input collection where phrases can be found.
outputCollection/Output Collection

Specify the collection in which the output documents should be indexed.

When running this job over a content document collection, be sure to set attachPhrases/Extract Key Phrases from Input Text to "true". The default is "false", which works well when running the job over a signals collection.

Output documents

By default, the job only outputs the phrases found from the original document. In each row of the phrases output, these fields are most useful:

The phrase itself is in the phrases_s field, which can be used for faceting.
The likelihood_d field gives the likelihood that the phrase is legitimate, from 0 to infinity.

Low-probability phrases are automatically trimmed from the results.
When a phrase’s likelihood value is ambiguous, the review field is set to "true" to indicate that the phrase should be reviewed.
A phrase_count field indicates the number of instances of the phrase in the input collection.

The complete list of output fields is shown below.

Output fields

aggr_id_s

The name of the Phrase Extraction job that generated this document.

doc_type_s

This is always key_phrases for documents generated by a Phrase Extraction job.

id

A unique ID for this document.

input_collection

The collection used for this job’s input.

likelihood_d

The likelihood that this phrases_s is a phrase, from 0 to infinity.

phrase_count

The number of occurrences of this phrase in the input collection.

phrases_s

The phrase detected by the job.

review

"True" indicates that this may not be a valid phrase and should be reviewed.

score

This is always "1".

timestamp

The date and time when the document was generated.

word_num_i

The number of words in this phrase.

_version_

An internal Solr field used for partial updates.

If the attachPhrases/Extract Key Phrases from Input Text parameter is set to "true", then the job also outputs the original documents from the input collection with an appended field, phrases_extracted_tt, that lists the extracted phrases from this document.

The way to distinguish the phrases output from the original document output is by the field doc_type_s, with one of these values:

key_phrases denotes phrases output.
original_doc_with_phrases denotes the original documents.

Configuration properties

Use this job when you want to identify statistically significant phrases in your content.

analyzerConfig - stringrequired

The style of text analyzer you would like to use.

Default: { "analyzers": [{ "name": "StdTokLowerStop","charFilters": [ { "type": "htmlstrip" } ],"tokenizer": { "type": "standard" },"filters": [{ "type": "lowercase" }] }],"fields": [{ "regex": ".+", "analyzer": "StdTokLowerStop" } ]}

attachPhrases - boolean

Checking this will cause the job to associate extracted phrases from each source doc. and write them back to the output collection. If input data is signals, it is suggested to turn this option off. Also, currently it is not allowed to check this option while attempting to write to a _query_rewrite_staging collection.

Default: false

dataFormat - stringrequired

Spark-compatible format that contains training data (like 'solr', 'parquet', 'orc' etc)

>= 1 characters

Default: solr

dataOutputFormat - string

Spark-compatible output format (like 'solr', 'parquet', etc)

>= 1 characters

Default: solr

enableAutoPublish - boolean

If true, automatically publishes rewrites for rules. Default is false to allow for initial human-aided reviewing

Default: false

fieldToVectorize - stringrequired

Solr field containing text training data. Data from multiple fields with different weights can be combined by specifying them as field1:weight1,field2:weight2 etc.

>= 1 characters

id - stringrequired

The ID for this Spark job. Used in the API to reference this job. Allowed characters: a-z, A-Z, dash (-) and underscore (_). Maximum length: 63 characters.

<= 63 characters

Match pattern: [a-zA-Z][_\-a-zA-Z0-9]*[a-zA-Z0-9]?

minLikelihood - number

Phrases below this threshold will not be written in the output of this job.

minmatch - integer

The number of times a phrase must exist to be considered. NOTE: if input is non signal data, please reduce the number to e.g. 5.

>= 1

exclusiveMinimum: false

Default: 100

ngramSize - integer

The number of words in the ngram you want to consider for the sips.

>= 2

<= 5

exclusiveMinimum: false

exclusiveMaximum: false

Default: 3

outputCollection - string

Solr Collection to store extracted phrases; defaults to the query_rewrite_staging collection for the associated app.

partitionCols - string

If writing to non-Solr sources, this field will accept a comma-delimited list of column names for partitioning the dataframe before writing to the external output

randomSeed - integer

For any deterministic pseudorandom number generation

Default: 8180

readOptions - array[object]

Options used when reading input from Solr or other sources.

object attributes:{key required : {
display name: Parameter Name
type: string
}value : {
display name: Parameter Value
type: string
}}

sourceFields - string

Solr fields to load (comma-delimited). Leave empty to allow the job to select the required fields to load at runtime.

sparkConfig - array[object]

Spark configuration settings.

object attributes:{key required : {
display name: Parameter Name
type: string
}value : {
display name: Parameter Value
type: string
}}

sparkPartitions - integer

Spark will re-partition the input to have this number of partitions. Increase for greater parallelism

Default: 200

sparkSQL - string

Use this field to create a Spark SQL query for filtering your input data. The input data will be registered as spark_input

Default: SELECT * from spark_input

trainingCollection - stringrequired

Solr Collection containing labeled training data

>= 1 characters

trainingDataFilterQuery - string

Solr query to use when loading training data if using Solr

Default: *:*

trainingDataFrameConfigOptions - object

Additional spark dataframe loading configuration options

trainingDataSamplingFraction - number

Fraction of the training data to use

<= 1

exclusiveMaximum: false

Default: 1

type - stringrequired

Default: sip

Allowed values: sip

writeOptions - array[object]

Options used when writing output to Solr or other sources

object attributes:{key required : {
display name: Parameter Name
type: string
}value : {
display name: Parameter Value
type: string
}}