Docs /Templates/ sinapsis-bertopic

BERTopicFitModelFromDataFrame

sinapsis-bertopicBERTopicBERTopicTopic_ModelingVisualizationNLP

Python template · sinapsis_bertopic.templates.bertopic_fit_model_from_dataframe

Overview

Fit and manage a BERTopic model using documents extracted from dataframe packets.

This class extends the BERTopicFitModel class, adding functionality to handle dataframe inputs.

This class provides methods for fitting a BERTopic model on input documents, generating visualizations, and saving both the model and visualizations to disk. It also supports exporting visualizations as images and saving training data used for fitting the model.

Attributes: columns_to_use (list[str]): A list of column names to extract text from within the dataframe packets. bertopic_model_params (BERTopicModelParams): Configuration parameters for the BERTopic model. Defaults to an empty dictionary which is converted to a BERTopicModelParams instance during validation. umap_model_params (UMAPModelParams): Configuration parameters for the UMAP dimensionality reduction model. Defaults to an empty dictionary which is converted to a UMAPModelParams instance during validation. hdbscan_model_params (HDBSCANModelParams): Configuration parameters for the HDBSCAN clustering model. Defaults to an empty dictionary which is converted to a HDBSCANModelParams instance during validation. bertopic_save_model_params (BERTopicSaveModelParams): Parameters for saving the BERTopic model. Defaults to BERTopicSaveModelParams(). export_visualization_to_image (bool): Whether to export visualizations as image files. Defaults to True. image_export_params (ToImageParams): Parameters for image export configuration. Defaults to ToImageParams(). root_dir (str): Root directory for storing model and visualization files. Defaults to SINAPSIS_CACHE_DIR. save_documents_visualization_path (str): Relative path for saving documents visualization HTML file. Defaults to "documents_visualization.html". save_model_path (str): Relative path for saving the BERTopic model. save_training_data (bool): Whether to save training data to disk. Defaults to False. save_training_data_path (str): Relative path for saving training data pickle file. Defaults to "training_data.pkl". sentence_model_name (str): Identifier for the sentence transformer model. Defaults to "sentence-transformers/all-MiniLM-L6-v2". visualize_documents_params (VisualizeDocumentsParams): Parameters for document visualization. Defaults to VisualizeDocumentsParams().

Capabilities

Capabilities: not set

Installation

Install the published package using pip or uv.

Terminalbash
pip install sinapsis-bertopic==1.0.0
uv pip install sinapsis-bertopic==1.0.0
uv add sinapsis-bertopic==1.0.0

Configuration

Effective fields from the inspector schema, including nested Pydantic model fields.

metadata
sinapsis_core.template_base.base_models.TemplateMetadata | None
Optional
Default: null
root_dir
str
Optional
Default: "/home/ariel/.cache/sinapsis"
columns_to_useRequired
list[str]
Required
Default: null
save_model_path
str
Optional
Default: "bertopic_model"
umap_model_params
UMAPModelParams
Optional
Default: null
a
float | None
Optional
Default: null
b
float | None
Optional
Default: null
init
typing.Literal['spectral', 'random', 'pca', 'tswspectral']
Optional
Default: "spectral"
metric
typing.Literal['euclidean', 'manhattan', 'chebyshev', 'minkowski', 'canberra', 'braycurtis', 'mahalanobis', 'wminkowski', 'seuclidean', 'cosine', 'correlation', 'haversine', 'hamming', 'jaccard', 'dice', 'russelrao', 'kulsinski', 'll_dirichlet', 'hellinger', 'rogerstanimoto', 'sokalmichener', 'sokalsneath', 'yule]']
Optional
Default: "euclidean"
n_jobs
int
Optional
Default: -1
spread
float
Optional
Default: 1
unique
bool
Optional
Default: false
densmap
bool
Optional
Default: false
verbose
bool
Optional
Default: false
min_dist
float
Optional
Default: 0.1
n_epochs
int | None
Optional
Default: null
dens_frac
float
Optional
Default: 0.3
tqdm_kwds
dict | None
Optional
Default: null
low_memory
bool
Optional
Default: true
dens_lambda
float
Optional
Default: 2
metric_kwds
dict | None
Optional
Default: null
n_neighbors
int
Optional
Default: 15
output_dens
bool
Optional
Default: false
n_components
int
Optional
Default: 2
random_state
int | None
Optional
Default: null
learning_rate
float
Optional
Default: 1
output_metric
str
Optional
Default: "euclidean"
target_metric
str
Optional
Default: "categorical"
target_weight
float
Optional
Default: 0.5
dens_var_shift
float
Optional
Default: 0.1
transform_mode
typing.Literal['embedding', 'graph']
Optional
Default: "embedding"
transform_seed
int
Optional
Default: 42
precomputed_knn
tuple
Optional
Default: [null,null,null]
set_op_mix_ratio
float
Optional
Default: 1
angular_rp_forest
bool
Optional
Default: false
local_connectivity
float
Optional
Default: 1
output_metric_kwds
dict | None
Optional
Default: null
repulsion_strength
float
Optional
Default: 1
target_metric_kwds
dict | None
Optional
Default: null
target_n_neighbors
int
Optional
Default: -1
negative_sample_rate
int
Optional
Default: 5
transform_queue_size
float
Optional
Default: 4
disconnection_distance
float | None
Optional
Default: null
force_approximation_algorithm
bool
Optional
Default: false
save_training_data
bool
Optional
Default: false
ctfidf_model_params
CTFIDFModelParams
Optional
Default: null
seed_words
list[str] | None
Optional
Default: null
bm25_weighting
bool
Optional
Default: true
seed_multiplier
int
Optional
Default: 2
reduce_frequent_words
bool
Optional
Default: true
image_export_params
ToImageParams
Optional
Default: null
scale
int | float | None
Optional
Default: null
width
int | None
Optional
Default: null
format
typing.Literal['png', 'jpg', 'jpeg', 'webp', 'svg', 'pdf']
Optional
Default: "png"
height
int | None
Optional
Default: null
validate_figure
bool
Optional
Default: true
sentence_model_name
str
Optional
Default: "sentence-transformers/all-MiniLM-L6-v2"
hdbscan_model_params
HDBSCANModelParams
Optional
Default: null
p
int | None
Optional
Default: null
alpha
float
Optional
Default: 1
kwargs
dict[str, typing.Any]
Optional
Default: null
metric
typing.Literal['euclidean', 'manhattan']
Optional
Default: "euclidean"
algorithm
typing.Literal['best', 'prims_kdtree', 'prims_balltree', 'boruvka_kdtree', 'boruvka_balltree']
Optional
Default: "best"
leaf_size
int
Optional
Default: 40
min_samples
int | None
Optional
Default: null
prediction_data
bool
Optional
Default: true
core_dist_n_jobs
int
Optional
Default: 4
max_cluster_size
int
Optional
Default: 0
min_cluster_size
int
Optional
Default: 5
gen_min_span_tree
bool
Optional
Default: false
allow_single_cluster
bool
Optional
Default: false
approx_min_span_tree
bool
Optional
Default: true
branch_detection_data
bool
Optional
Default: false
cluster_selection_method
typing.Literal['eom', 'leaf']
Optional
Default: "eom"
cluster_selection_epsilon
float
Optional
Default: 0
cluster_selection_epsilon_max
float
Optional
Default: null
cluster_selection_persistence
float
Optional
Default: 0
match_reference_implementation
bool
Optional
Default: false
bertopic_model_params
BERTopicModelParams
Optional
Default: null
language
str
Optional
Default: "english"
nr_topics
int | str | None
Optional
Default: null
low_memory
bool
Optional
Default: false
top_n_words
int
Optional
Default: 10
n_gram_range
tuple[int, int]
Optional
Default: [1,1]
min_topic_size
int
Optional
Default: 5
seed_topic_list
list[list[str]] | None
Optional
Default: null
zeroshot_topic_list
list[str] | None
Optional
Default: null
calculate_probabilities
bool
Optional
Default: false
zeroshot_min_similarity
float
Optional
Default: 0.7
save_training_data_path
str
Optional
Default: "training_data.pkl"
bertopic_save_model_params
BERTopicSaveModelParams
Optional
Default: null
save_ctfidf
bool
Optional
Default: true
serialization
typing.Literal['safetensors', 'pytorch', 'pickle']
Optional
Default: "safetensors"
visualize_documents_params
VisualizeDocumentsParams
Optional
Default: null
title
str
Optional
Default: "<b>Documents and Topics</b>"
width
int
Optional
Default: 1200
height
int
Optional
Default: 750
sample
float | None
Optional
Default: null
topics
list[int] | None
Optional
Default: null
custom_labels
bool
Optional
Default: false
hide_annotations
bool
Optional
Default: false
hide_document_hover
bool
Optional
Default: false
export_visualization_to_image
bool
Optional
Default: true
save_documents_visualization_path
str
Optional
Default: "documents_visualization.html"

Configuration example

Generated from documented defaults. Replace any placeholders and complete required nested settings before running.

yaml
templates:
  - template_name: "BERTopicFitModelFromDataFrame"
    class_name: "BERTopicFitModelFromDataFrame"
    attributes:
      root_dir: "/home/ariel/.cache/sinapsis"
      columns_to_use: "<required>"
      save_model_path: "bertopic_model"
      umap_model_params:
        a: null
        b: null
        init: "spectral"
        metric: "euclidean"
        n_jobs: -1
        spread: 1
        unique: false
        densmap: false
        verbose: false
        min_dist: 0.1
        n_epochs: null
        dens_frac: 0.3
        tqdm_kwds: null
        low_memory: true
        dens_lambda: 2
        metric_kwds: null
        n_neighbors: 15
        output_dens: false
        n_components: 2
        random_state: null
        learning_rate: 1
        output_metric: "euclidean"
        target_metric: "categorical"
        target_weight: 0.5
        dens_var_shift: 0.1
        transform_mode: "embedding"
        transform_seed: 42
        precomputed_knn: {}
        set_op_mix_ratio: 1
        angular_rp_forest: false
        local_connectivity: 1
        output_metric_kwds: null
        repulsion_strength: 1
        target_metric_kwds: null
        target_n_neighbors: -1
        negative_sample_rate: 5
        transform_queue_size: 4
        disconnection_distance: null
        force_approximation_algorithm: false
      save_training_data: false
      ctfidf_model_params:
        seed_words: null
        bm25_weighting: true
        seed_multiplier: 2
        reduce_frequent_words: true
      image_export_params:
        scale: null
        width: null
        format: "png"
        height: null
        validate_figure: true
      sentence_model_name: "sentence-transformers/all-MiniLM-L6-v2"
      hdbscan_model_params:
        p: null
        alpha: 1
        kwargs: null
        metric: "euclidean"
        algorithm: "best"
        leaf_size: 40
        min_samples: null
        prediction_data: true
        core_dist_n_jobs: 4
        max_cluster_size: 0
        min_cluster_size: 5
        gen_min_span_tree: false
        allow_single_cluster: false
        approx_min_span_tree: true
        branch_detection_data: false
        cluster_selection_method: "eom"
        cluster_selection_epsilon: 0
        cluster_selection_epsilon_max: null
        cluster_selection_persistence: 0
        match_reference_implementation: false
      bertopic_model_params:
        language: "english"
        nr_topics: null
        low_memory: false
        top_n_words: 10
        n_gram_range: [1, 1]
        min_topic_size: 5
        seed_topic_list: null
        zeroshot_topic_list: null
        calculate_probabilities: false
        zeroshot_min_similarity: 0.7
      save_training_data_path: "training_data.pkl"
      bertopic_save_model_params:
        save_ctfidf: true
        serialization: "safetensors"
      visualize_documents_params:
        title: "<b>Documents and Topics</b>"
        width: 1200
        height: 750
        sample: null
        topics: null
        custom_labels: false
        hide_annotations: false
        hide_document_hover: false
      export_visualization_to_image: true
      save_documents_visualization_path: "documents_visualization.html"

Inheritance

Immediate parent: sinapsis_bertopic.templates.bertopic_fit_model.BERTopicFitModel

Environment variables

Environment definitions belong to their canonical Python modules.

No directly related environment variables listed.