Biterm Topic Models for Short Text

Biterm Topic Models find topics in collections of short texts. It is a word co-occurrence based topic model that learns topics by modeling word-word co-occurrences patterns which are called biterms. This in contrast to traditional topic models like Latent Dirichlet Allocation and Probabilistic Latent Semantic Analysis which are word-document co-occurrence topic models. A biterm consists of two words co-occurring in the same short text window. This context window can for example be a twitter message, a short answer on a survey, a sentence of a text or a document identifier. The techniques are explained in detail in the paper 'A Biterm Topic Model For Short Text' by Xiaohui Yan, Jiafeng Guo, Yanyan Lan, Xueqi Cheng (2013) < https://github.com/xiaohuiyan/xiaohuiyan.github.io/blob/master/paper/BTM-WWW13.pdf>.


This is an R package wrapping the C++ code available at https://github.com/xiaohuiyan/BTM for constructing a Biterm Topic Model (BTM). This model models word-word co-occurrences patterns (e.g., biterms).

Installation

This R package is on CRAN, just install it with install.packages('BTM')

What

The Biterm Topic Model (BTM) is a word co-occurrence based topic model that learns topics by modeling word-word co-occurrences patterns (e.g., biterms)

  • A biterm consists of two words co-occurring in the same context, for example, in the same short text window.
  • BTM models the biterm occurrences in a corpus (unlike LDA models which model the word occurrences in a document).
  • It's a generative model. In the generation procedure, a biterm is generated by drawing two words independently from a same topic z. In other words, the distribution of a biterm b=(wi,wj) is defined as: P(b) = sum_k{P(wi|z)*P(wj|z)*P(z)} where k is the number of topics you want to extract.
  • Estimation of the topic model is done with the Gibbs sampling algorithm. Where estimates are provided for P(w|k)=phi and P(z)=theta.

More detail can be referred to the following paper:

Xiaohui Yan, Jiafeng Guo, Yanyan Lan, Xueqi Cheng. A Biterm Topic Model For Short Text. WWW2013. https://github.com/xiaohuiyan/xiaohuiyan.github.io/blob/master/paper/BTM-WWW13.pdf

Example

library(udpipe)
library(BTM)
data("brussels_reviews_anno", package = "udpipe")

## Taking only nouns of Dutch data
x <- subset(brussels_reviews_anno, language == "nl")
x <- subset(x, xpos %in% c("NN", "NNP", "NNS"))
x <- x[, c("doc_id", "lemma")]

## Building the model
set.seed(321)
model  <- BTM(x, k = 3, beta = 0.01, iter = 1000, trace = 100)

## Inspect the model - topic frequency + conditional term probabilities
model$theta
[1] 0.3406998 0.2413721 0.4179281

topicterms <- terms(model, top_n = 10)
topicterms
[[1]]
         token probability
1  appartement  0.06168297
2      brussel  0.04057012
3        kamer  0.02372442
4      centrum  0.01550855
5      locatie  0.01547671
6         stad  0.01229227
7        buurt  0.01181460
8     verblijf  0.01155985
9         huis  0.01111402
10         dag  0.01041345

[[2]]
         token probability
1  appartement  0.05687312
2      brussel  0.01888307
3        buurt  0.01883812
4        kamer  0.01465696
5     verblijf  0.01339812
6     badkamer  0.01285862
7   slaapkamer  0.01276870
8          dag  0.01213928
9          bed  0.01195945
10        raam  0.01164474

[[3]]
         token probability
1  appartement 0.061804812
2      brussel 0.035873377
3      centrum 0.022193831
4         huis 0.020091282
5        buurt 0.019935537
6     verblijf 0.018611710
7     aanrader 0.014614272
8        kamer 0.011447470
9      locatie 0.010902365
10      keuken 0.009448751
scores <- predict(model, newdata = x)

Make a specific topic called the background

# If you set background to TRUE
# The first topic is set to a background topic that equals to the empirical word distribution. 
# This can be used to filter out common words.
set.seed(321)
model  <- BTM(x, k = 5, beta = 0.01, background = TRUE, iter = 1000, trace = 100)
topicterms <- terms(model, top_n = 5)
topicterms

Support in text mining

Need support in text mining? Contact BNOSAC: http://www.bnosac.be

News

CHANGES IN BTM VERSION 0.2

  • Allow to get all biterms from the model using terms.BTM
  • Get the likelihood of a biterm alongside the model

CHANGES IN BTM VERSION 0.1

  • Initial release based on BTM commit 66cc9b475afec81f3e74bb393b874b3fe5d5a148
  • Contains functions BTM, predict.BTM and terms.BTM

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.

install.packages("BTM")

0.2 by Jan Wijffels, 5 months ago


Browse source code at https://github.com/cran/BTM


Authors: Jan Wijffels [aut, cre, cph] (R wrapper) , BNOSAC [cph] (R wrapper) , Xiaohui Yan [ctb, cph] (BTM C++ library)


Documentation:   PDF Manual  


Apache License 2.0 license


Imports Rcpp, utils

Suggests udpipe

Linking to Rcpp

System requirements: C++11


See at CRAN