https://github.com/till-tietz/gsdmm

GSDMM Short Text Clustering via Dirichlet Mixture Models
https://github.com/till-tietz/gsdmm

cpp r rcpp text-analytics text-clustering

Last synced: about 1 month ago
JSON representation

GSDMM Short Text Clustering via Dirichlet Mixture Models

• Host: GitHub
• URL: https://github.com/till-tietz/gsdmm
• Owner: till-tietz
• License: other
• Created: 2023-07-25T15:02:01.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-07-26T20:53:29.000Z (over 1 year ago)
• Last Synced: 2023-07-26T21:55:58.627Z (over 1 year ago)
• Topics: cpp, r, rcpp, text-analytics, text-clustering
• Language: C++
• Homepage:
• Size: 1.26 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.Rmd
  • License: LICENSE

Awesome Lists containing this project

README

        
---

output: github_document

---

```{r, include = FALSE}

knitr::opts_chunk$set(

  collapse = TRUE,

  comment = "#>",

  fig.path = "man/figures/README-",

  out.width = "100%"

)

```

# gsdmm

`gsdmm` implements short text classification via Dirichlet Mixture Models proposed by [Yin and Wang 2014](https://www.semanticscholar.org/paper/A-dirichlet-multinomial-mixture-model-based-for-Yin-Wang/d03ca28403da15e75bc3e90c21eab44031257e80?p2df). It provides a fast `c++` implementation and R interface for the Gibbs sampler described in the paper. Specifically, `gsdmm` implements the Likelihood function allowing for multiple occurrences of the same word in a given text (EQ4). 

**Benefits:** \

- very space and time efficient 

- unlike LDA it requires only an upper bound on the number of clusters 

**Development:** \

- I am planning to add a tuning function for the alpha and beta parameters of the gibbs sampler

## Installation

You can install the development version of gsdmm from [GitHub](https://github.com/) with:

``` r

# install.packages("devtools")

devtools::install_github("till-tietz/gsdmm")

```

## Usage 

Here is a minimal working example. 

```{r, message = FALSE, warning = FALSE}

# we lemmatize and tokenize creating a list of character vector representing each text

text <- c(

  "Rockets are amazing.",

  "Witnessing a rocket in flight is a marvel of engineering.",

  "We should take a rocket to Mars.",

  "Rocket",

  "Have you ever seen a cat?",

  "Cats are fun.",

  "Your cat seems sweet.",

  "Cat"

) |>

  tolower()  |>

  gsub(pattern = '[[:punct:] ]+', replacement = ' ') |>

  textstem::lemmatize_strings() |>

  text2vec::word_tokenizer() |>

  lapply(function(i) i[!i %in% stopwords::stopwords()])

gsdmm::gsdmm(texts = text, n_iter = 100, n_clust = 20, alpha = 0.1, beta = 0.2)

```