Skip to content

Latest commit

 

History

History
84 lines (39 loc) · 4.62 KB

NEWS.md

File metadata and controls

84 lines (39 loc) · 4.62 KB
Raw

flaiR 0.0.6 (2023-10-29)

  • The flair in {flaiR} was renamed from flair() to import_flair() to avoid overlapping with conventional practice import flair in Python.

  • install_python_package() and uninstall_python_package() are new functions to install and uninstall Python packages using pip in the environment used by your flaiR package.

  • Add new training data from grandstanding training data from Ju Yeon Park's paper.

  • zzz.R is a revised code that proceeds through three steps. First, when installing and loading the package, {flaiR} utilizes the system's environment tool and undergoes three evaluation stages. Initially, {flaiR} requires at least Python 3 to be installed on your device. If Python 3 is not available, you will be unable to install {flaiR} successfully. Once this requirement is met, the system then checks for the appropriate versions of PyTorch and Flair. The primary focus here is on Flair. If it is not already installed, you will see a message indicating that 'Flair is being installed from Python'. This process represents a new format for loading the Python environment used by your flaiR package.

  • Add example datasets (cc_muller and hatespeech_zh_tw) for tutorials and documentation.

flaiR 0.0.5 (2023-10-01)

  • Added more tests to monitor function operation.

  • Added wrapped functions integrating Python code.

  • Created a function for coloring entities.

  • Provided some tutorials for interacting with R and Python using Flair.

  • Notice that Python 3.x and flair may fail to install Python dependencies on windows-latest due to potential compatibility issues with the latest Python versions on Windows. To fix this, I modified the Python version in your actions/setup-python@v2 step to use Python 3.9 or a lower version.

  • Added two new example datasets for tutorials and documentation.

 

flaiR 0.0.3 (2023-09-10)

Modifications Overview

  • Added show.text_id and gc.active parameters to get_entities(), get_pos(), and get_sentiment().

  • Enhanced batch processing with the introduction of batch_size in functions get_entities_batch(), get_pos_batch(), and get_sentiment_batch(). Introduced device parameter to specify computation device.

Introduction of New Parameters:

  • show.text_id: When activated (TRUE), the actual text (labeled 'text_id') from which the entity was derived will be appended in the resulting dataset. Although enriching the output for validation and traceability, users should be cautious, as this might inflate the output size. By default, this option remains deactivated (FALSE). For context, previously, the 'text_id' was intrinsically generated, potentially elevating R's memory consumption.

  • gc.active: Activating this (TRUE) will trigger the garbage collector post-text processing. This action aids in memory optimization by relinquishing unallocated memory spaces, a crucial step, particularly when processing an extensive text dataset. While the default is set to FALSE, users managing larger texts should consider setting gc.active to TRUE. Though this action doesn't bolster computational efficiency, it does circumvent potential RStudio crashes.

Batch Processing Enhancement:

The inception of the batch_size parameter (defaulted at 5) in the get_entities_batch(), get_pos_batch(), and get_sentiment_batch() augments their batch processing capabilities. This addition led to the creation of an internal function named process_batch to proficiently manage each text batch and the linked doc_ids. The core functionality has been adapted to segregate the texts and doc_ids into specific batches, subsequently processed via the process_batch function, with the final results amalgamated seamlessly.

  • device: A descriptive character string pinpointing the computation device. Users can opt for "cpu" or a GPU device number in string format. For instance, representing the primary GPU as 0. When a GPU device number is furnished, the system will endeavor to harness that specific GPU, with "cpu" as the default setting.

  • batch_size: An integer specifying the size of each batch. Default is 5.

 

flaiR 0.0.1 (development version)

  • The features in flaiR currently include part-of-speech tagging, sentiment tagging, and named entity recognition tagging. flaiR requires Python version 3.7 or higher to operate concurrently.

  • create_flair_env(): A function to install the Flair Python library using the reticulate R package, which is automatically generated.