Welcome to Textpresso Classifiers Module documentation!

textpresso_classifiers is a Python package to train and apply document classifiers to scientific papers, both in the form of pdf files and Textpresso Central documents (CAS format).

The package contains a Python library with classes to create and manage classifiers based on several supervised learning models. Specifically, the library provides a simplified interface to the models and gives a set of utilities to process the data needed to feed the models. For example, the library exposes functions to perform feature extraction and feature selection from data obtained from different file formats (pdf and CAS), which are automatically converted to plain text and used to create the feature sets.

The library currently supports the following classifiers:

• SVM (both linear and non-linear)
• Linear Discriminant Analysis
• Gaussian process
• Naive Bayes
• XGBoost
• Decision Tree
• Random Forest
• K-Nearest Neighbors
• Multi-Layer Perceptrons (Neural Network)
• Radial Basis Function Neural Network

Feature extraction in TCM allows the user to apply several text pre-processing phases (e.g., lemmatization, n-gram extraction). Feature selection is based on the chi-squared method.

TCM is based on scikit-learn Python library.

• Textpresso Documents Classifiers
• File Management Utilities
• Executable programs included in TCM
  • tp_doc_classifier.py
  • classifiers_comparison.py
  • convert_doc_to_txt.py