Description
The LDA td_lda_mle
function uses training data and parameters to build a
topic model, using an unsupervised method to estimate the correlation
between the topics and words according to the topic number and other parameters.
Optionally, the function generates the topic distributions for each training document.
Usage
td_lda_mle (
data = NULL,
topic.num = NULL,
docid.column = NULL,
word.column = NULL,
alpha = 0.1,
eta = 0.1,
count.column = NULL,
maxiter = 50,
convergence.delta = 1.0E-4,
seed = NULL,
out.topicnum = "all",
out.topicwordnum = "none",
initmodeltaskcount = NULL,
data.sequence.column = NULL
)
Arguments
data |
Required Argument.
Specifies the name of the tbl_teradata or view that contains the new
documents.
|
topic.num |
Required Argument.
Specifies the number of topics for all the documents in the input
table, an INTEGER value in the range [2, 1000], both lower and upper bounds
inclusive.
Types: numeric
|
docid.column |
Required Argument.
Specifies the name of the input column that contains the document
identifiers.
Types: character
|
word.column |
Required Argument.
Specifies the name of the input column that contains the words
(one word in each row).
Types: character
|
alpha |
Optional Argument.
Specifies a hyperparameter of the model, the prior smooth parameter
for the topic distribution over documents. As alpha decreases, fewer
topics are associated with each document.
Default Value: 0.1.
Types: numeric
|
eta |
Optional Argument.
Specifies a hyperparameter of the model, the prior smooth parameter
for the word distribution over topics. As eta decreases, fewer words
are associated with each topic.
Default Value: 0.1.
Types: numeric
|
count.column |
Optional Argument.
Specifies the name of the input column that contains the count of
the corresponding word in the row.
Types: character
|
maxiter |
Optional Argument.
Specifies the maximum number of iterations to perform if the model
does not converge, a positive INTEGER value.
Default Value: 50.
Types: numeric
|
convergence.delta |
Optional Argument.
Specifies the convergence delta of log perplexity, a NUMERIC value in
the range (0.0, 1.0), both lower and upper bounds exclusive.
Default Value: 1.0E-4.
Types: numeric
|
seed |
Optional Argument.
Specifies the seed with which to initialize the model, a numeric
value. Given the same seed, cluster configuration, and input table,
the function generates the same model. By default, the function
initializes the model randomly.
Types: numeric
|
out.topicnum |
Optional Argument.
Specifies the number of top-weighted topics and their weights to
include in the output tbl_teradata for each training document. The
value topics must be a positive numeric value and the value should
be enclosed in quotes. The default value "all" specifies all topics
and their weights.
Types: character
|
out.topicwordnum |
Optional Argument.
Specifies the number of top topic words and their topic identifiers
to include in the output tbl_teradata for each training document. The
value "out.topicwordnum" must be a positive numeric value and the value
should be enclosed in quotes.
The value "all" specifies all topic words and their topic identifiers.
The default value "none" specifies no topic words or topic identifiers.
Types: character
|
initmodeltaskcount |
Optional Argument.
Specifies the number of vWorkers that are adopted to generate
initalized model. It must be a positive INTEGER value.
By default, all the available vWorkers are used to initialize the model.
Types: numeric
|
data.sequence.column |
Optional Argument.
Specifies the vector of column(s) that uniquely identifies each row
of the input argument "data". The argument is used to ensure
deterministic results for functions which produce results that vary
from run to run.
Types: character OR vector of Strings (character)
|
Value
Function returns an object of class "td_lda_mle" which is a named list
containing Teradata tbl objects.
Named list members can be referenced directly with the "$" operator
using following names:
model.table
-
doc.distribution.data
output
.
Examples
# Get the current context/connection
con <- td_get_context()$connection
# Load example data.
loadExampleData("lda_example", "complaints_traintoken")
# Create remote tibble objects.
complaints_traintoken <- tbl(con, "complaints_traintoken")
# Example 1 - This function uses training data and parameters from complaints_traintoken to build a topic model.
td_lda_out <- td_lda_mle(data = complaints_traintoken,
topic.num = 5,
docid.column = "doc_id",
word.column = "token",
count.column = "frequency",
maxiter = 30,
convergence.delta = 1e-3,
seed = 2
)