V0.46

New changes:

• Ability to add metadata and display them on a plot
• Fix bugs with Mistral Generation
• Add FlagEmbedding
• Remove Chroma
• Add tokens count
• Sample documents to improve the terms extraction process
• Save and load Bunka objects
• load embeddings into Bunka
• Ability to add your own projection model (TSNE, PCA, UMAP etc)
• Add outliers detection
• automatic language detection

In this major update, we have completely revamped our package to align it with the latest framework and to better facilitate data cleaning for the next generation of Large Language Models (LLMs).Bunkatopics version 0.45 continues to serve as a versatile package catering to a wide range of tasks, including Topic Modeling Visualization, Frame Analysis and cleaning capabilities.

Utilizing Large Language Models: Our package uses the power of Large Language Models (LLMs) to empower developers in extracting valuable insights from unstructured data.

Integration of well-known Libraries: Bunkatopics is now constructed using renowned libraries such as langchain, chroma, and transformers, ensuring seamless integration into diverse development environments.

In-depth insights: Bunkatopics excels at providing in-depth insights into specific topics within categories, such as exploring Technology topics on the Medium website.

Framing Analysis: Bunkatopics introduces a 2-dimensional unsupervised scale, the Bourdieu map, facilitating the visualization of textual data and topics. This feature offers valuable insights into data distribution patterns.

Manually Cleaning Topics: Users can now exercise greater flexibility and customization by manually changing topic names or labels to better suit their needs.

Data Filtering by Topics. This release enables users to construct tailored datasets by excluding topics that do not align with their specific interests, simplifying the process of fine-tuning models.

In this major update, we have completely revamped our package to align it with the latest framework and to better facilitate data cleaning for the next generation of Large Language Models (LLMs).Bunkatopics version 0.45 continues to serve as a versatile package catering to a wide range of tasks, including Topic Modeling Visualization, Frame Analysis and cleaning capabilities.

• Utilizing Large Language Models: Our package uses the power of Large Language Models (LLMs) to empower developers in extracting valuable insights from unstructured data.

• Integration of well-known Libraries: Bunkatopics is now constructed using renowned libraries such as langchain, chroma, and transformers, ensuring seamless integration into diverse development environments.

• In-depth insights: Bunkatopics excels at providing in-depth insights into specific topics within categories, such as exploring Technology topics on the Medium website.

• Framing Analysis: Bunkatopics introduces a 2-dimensional unsupervised scale, the Bourdieu map, facilitating the visualization of textual data and topics. This feature offers valuable insights into data distribution patterns.

• Manually Cleaning Topics: Users can now exercise greater flexibility and customization by manually changing topic names or labels to better suit their needs.

• Data Filtering by Topics. This release enables users to construct tailored datasets by excluding topics that do not align with their specific interests, simplifying the process of fine-tuning models.