textir (version 1.8-6)

summary.topics: topic summary

Description

Summary function for Topic Models

Usage

## S3 method for class 'topics':
summary( object, nwrd=5, tpk=NULL, verb=TRUE, ... )

Arguments

object

An output object from the topics function.

nwrd

The number of phrases to output for each topic.

tpk

Optional list of topics to summarize. Defaults to 1:x$K.

verb

Whether or not to print the summary.

...

Unused arguments from other functions, for S3 compatibility.

Value

• The function prints available model selection results (log Bayes factors, fitted dispersion, and p-value from a test for dispersion $>1$) along with usage percentages (i.e. colMeans(omega)) and the top nwrd phrases by term-lift for each topic in tpk. The matrix of top nwrd phrases and their associated lift is returned invisibly.

Details

This summary orders phrases for each topic according to the lift $\theta_{kj}/q_{j}$, where $q_j$ is the null-model probability estimate $\sum_i x_{ij} / \sum_{i} m_i$. This ordering of term relevance can be used to identify representative vocabulary for each topic.

References

Taddy (2012), On Estimation and Selection for Topic Models. http://arxiv.org/abs/1109.4518

See Also

topics, plot.topics, we8there, congress109, wsjibm