Library

embedprepro is a command-line tool designed for text analysis tasks, including embedding, clustering, dimensionality reduction, and visualization. This tool leverages various machine learning and data processing techniques to provide a comprehensive solution for text data analysis.

Table of Contents

• Installation
• Usage
  • Command-line Interface
    • Embedding
    • Dimensionality Reduction
    • Clustering
    • Visualization
  • Importing in Your Project

Installation

You can install the package directly using pip:

!pip install -U embedprepro

If you prefer to install the package from the source, clone the repository and install it using pip:

git clone https://github.com/Elma-dev/embedprepro-lib.git
cd text_analysis_cli
pip install .

Usage

Command-line Interface

The text_analysis_cli provides a command-line interface for performing various text analysis tasks.

embedprepro [OPTIONS] COMMAND [ARGS]...

The main commands available are:

• clustring
• embedding
• reduction
• visualization

To get help on any command, use the --help option:

embedprepro COMMAND --help

Embedding

Generate embeddings for text data using a specified model and embedder type.

embedprepro embedding <input_file> <output_file> [options]

• input file: the file contain your data (example.csv)
• output file: the file you want to saved the result (result.npy)

Options

Option	Description	Default
--et	Embedder type	sentence_transformer
--mn	Model name	all-MiniLM-L6-v2
--col	Column name in the input CSV file containing text	text
--bs	Batch size	32
--p	Number of parallel processes	1

• available emebdder type

Embedder	available
sentence_transformer	✅

embedprepro embedding input.csv output_embeddings.npy --et sentence_transformer --mn all-MiniLM-L6-v2 --col text --bs 32 --p 2

Dimensionality Reduction

Reduce the dimensionality of text embeddings.

embedprepro reduction <input_file> <output_file> [options]

• if input is text data then reduction firstly embed data before reduction
• you can also make embeding.npy as input_file bby adding --with_embedding 1 option

Option	Description	Default
--nc	Number of components to reduce to	2
--ng	Number of neighbors	15
--md	Minimum distance	0.5
--metric	Distance metric	euclidean
--et	Embedder type	sentence_transformer
--mn	Model name	all-MiniLM-L6-v2
--col	Column name in the input CSV file containing text	text
--algorithm	Dimensionality reduction algorithm	PCA
--with_embedding	Use precomputed embeddings	False (0)

• available Reduction Algorithm

  Algorithm	available
  PCA	✅
  UMAP	✅

embedprepro reduction input.csv dimreduction.npy --nc 2 --ng 15 --md 0.5 --metric euclidean --et sentence_transformer --mn all-MiniLM-L6-v2 --col text --algorithm PCA --with_embedding False

Clustering

Perform agglomerative clustering on text data or embeddings.

embedprepro clustering <input_file> <output_file> [options]

• if input is text data then clustering firstly embed data before clustering

Options

Option	Description	Default
--et	Embedder type	sentence_transformer
--mn	Model name	all-MiniLM-L6-v2
--col	Column name in the input CSV file containing text	text
--bs	Batch size	32
--p	Number of parallel processes	1
--threshold	Clustering threshold	0.5
--min_cluster_size	Minimum cluster size	1
--show_progress_bar	Show progress bar	True (1)
--with_embedding	Use precomputed embeddings	False (0)

example

 embedprepro clustering input.csv output_clusters.npy --et sentence_transformer --mn all-MiniLM-L6-v2 --col text --bs 32 --p 2 --threshold 0.5 --min_cluster_size 1 --show_progress_bar True --with_embedding False

Visualization

Visualize the results of dimensionality reduction and clustering.

embedprepro visualization <clusters_data> <reduced_data> [options]

• with visualization you can plot your clustered and reduced data with 2d or 3d plot.
• to use 3d plot you need just add —zi

Options

Option	Description	Default
--xi	Index of the first dimension	0
--yi	Index of the second dimension	1
--zi	Index of the third dimension	-1
--title	Title of the plot	Clusters
--xlabel	Label of the x-axis	X
--ylabel	Label of the y-axis	Y
--zlabel	Label of the z-axis	Z
--save	Save path for the plot	None

example

embedprepro visualization dimreduction.npy output_clusters.npy --xi 0 --yi 1 --zi 2 --title "Clusters" --xlabel "X" --ylabel "Y" --zlabel "Z"

Python Project

after installation you can use embedprepro inside your python project like this:

from preprocessing import *

from preprocessing package you can import:

• embedding_service
• agglomerative_clustering
• dimensionality_reduction
• visualization_service