multimodal_fin.processing package

Subpackages

• multimodal_fin.processing.metadata package
  • Submodules
  • multimodal_fin.processing.metadata.coherence_analyzer module
    • CoherenceAnalysis
      • CoherenceAnalysis.consistent
      • CoherenceAnalysis.contradictions
      • CoherenceAnalysis.model_config
      • CoherenceAnalysis.summary
      • CoherenceAnalysis.topic_covered
    • CoherenceAnalyzer
      • CoherenceAnalyzer.analyze_coherence()
      • CoherenceAnalyzer.model_name
    • ContradictionDetail
      • ContradictionDetail.explanation
      • ContradictionDetail.model_config
      • ContradictionDetail.monologue_excerpt
      • ContradictionDetail.response_excerpt
  • multimodal_fin.processing.metadata.metadata_enricher module
    • MetadataEnricher
      • MetadataEnricher.device
      • MetadataEnricher.enrich()
      • MetadataEnricher.num_evaluations
      • MetadataEnricher.qa_analyzer_models
      • MetadataEnricher.sec10k_model_names
      • MetadataEnricher.verbose
  • multimodal_fin.processing.metadata.prompt_builder module
    • PromptBuilder
      • PromptBuilder.analize_qa()
      • PromptBuilder.check_coherence()
      • PromptBuilder.explain_why_other()
      • PromptBuilder.prompt_10k()
      • PromptBuilder.prompt_monologue()
      • PromptBuilder.prompt_qa()
  • multimodal_fin.processing.metadata.qa_analyzer module
    • EvaluatedQA
      • EvaluatedQA.answer_quote
      • EvaluatedQA.answer_summary
      • EvaluatedQA.answered
      • EvaluatedQA.model_config
      • EvaluatedQA.question
    • InterventionAnalysis
      • InterventionAnalysis.evaluations
      • InterventionAnalysis.intervention
      • InterventionAnalysis.model_config
      • InterventionAnalysis.response
    • QAAnalyzer
      • QAAnalyzer.NUM_EVALUATIONS
      • QAAnalyzer.analize_qa()
      • QAAnalyzer.evaluate_qa_model()
      • QAAnalyzer.get_pred()
      • QAAnalyzer.get_pred_question()
      • QAAnalyzer.model_name
  • multimodal_fin.processing.metadata.sec10k_analyzer module
    • Category10K
      • Category10K.category
      • Category10K.model_config
    • SEC10KAnalyzer
      • SEC10KAnalyzer.NUM_EVALUATIONS
      • SEC10KAnalyzer.classify_dataframe()
      • SEC10KAnalyzer.classify_text()
      • SEC10KAnalyzer.explain_other_category()
      • SEC10KAnalyzer.get_pred()
      • SEC10KAnalyzer.model
  • Module contents
• multimodal_fin.processing.multimodal package
  • Subpackages
    • multimodal_fin.processing.multimodal.audio package
      • Submodules
      • multimodal_fin.processing.multimodal.audio.audio_emotion_analyzer module
      • Module contents
    • multimodal_fin.processing.multimodal.text package
      • Submodules
      • multimodal_fin.processing.multimodal.text.text_emotion_analyzer module
      • Module contents
    • multimodal_fin.processing.multimodal.video package
      • Submodules
      • multimodal_fin.processing.multimodal.video.analyzer module
      • multimodal_fin.processing.multimodal.video.face_detector module
      • multimodal_fin.processing.multimodal.video.processor module
      • multimodal_fin.processing.multimodal.video.video_emotion_analyzer module
      • Module contents
  • Submodules
  • multimodal_fin.processing.multimodal.embeddings_extractor module
    • EmbeddingsExtractor
      • EmbeddingsExtractor.audio_model_name
      • EmbeddingsExtractor.device
      • EmbeddingsExtractor.extract()
      • EmbeddingsExtractor.text_model_name
      • EmbeddingsExtractor.verbose
      • EmbeddingsExtractor.video_model_name
  • multimodal_fin.processing.multimodal.multimodal_embeddings module
    • MultimodalEmbeddings
      • MultimodalEmbeddings.audio_emotion_analyzer
      • MultimodalEmbeddings.audio_file_path
      • MultimodalEmbeddings.cortar_audio_temporal()
      • MultimodalEmbeddings.cortar_video_temporal()
      • MultimodalEmbeddings.generar_embeddings()
      • MultimodalEmbeddings.path_csv
      • MultimodalEmbeddings.path_json
      • MultimodalEmbeddings.text_emotion_analyzer
      • MultimodalEmbeddings.video_emmotion_analyzer
    • dummy_npwarn_decorator_factory()
  • Module contents
• multimodal_fin.processing.preprocessing package
  • Submodules
  • multimodal_fin.processing.preprocessing.ensemble_classifier module
    • EnsembleInterventionClassifier
      • EnsembleInterventionClassifier.NUM_EVALUATIONS
      • EnsembleInterventionClassifier.annotate_question_answer_pairs()
      • EnsembleInterventionClassifier.classify_dataframe()
      • EnsembleInterventionClassifier.ensemble_predict()
      • EnsembleInterventionClassifier.monologue_model_names
      • EnsembleInterventionClassifier.qa_model_names
      • EnsembleInterventionClassifier.verbose
  • multimodal_fin.processing.preprocessing.monologue_classifier module
    • CategoryPresentation
      • CategoryPresentation.category
      • CategoryPresentation.model_config
    • MonologueClassifier
      • MonologueClassifier.NUM_EVALUATIONS
      • MonologueClassifier.classify_dataframe()
      • MonologueClassifier.classify_text()
      • MonologueClassifier.get_pred()
      • MonologueClassifier.model
      • MonologueClassifier.output_path
  • multimodal_fin.processing.preprocessing.preprocessor module
    • Preprocessor
      • Preprocessor.divide_conference()
      • Preprocessor.extract_qna_intro()
      • Preprocessor.monologue_model_names
      • Preprocessor.num_evaluations
      • Preprocessor.process()
      • Preprocessor.process_and_save()
      • Preprocessor.qa_model_names
      • Preprocessor.qna_key
      • Preprocessor.section_col
      • Preprocessor.text_col
      • Preprocessor.verbose
  • multimodal_fin.processing.preprocessing.qa_classifier module
    • CategoryQA
      • CategoryQA.category
      • CategoryQA.model_config
    • QAClassifier
      • QAClassifier.NUM_EVALUATIONS
      • QAClassifier.classify_dataframe()
      • QAClassifier.classify_text()
      • QAClassifier.get_pred()
      • QAClassifier.model
  • multimodal_fin.processing.preprocessing.transcript_preprocessor module
    • TranscriptPreprocessor
      • TranscriptPreprocessor.extract_qna_intro()
      • TranscriptPreprocessor.preprocess()
      • TranscriptPreprocessor.qna_key
      • TranscriptPreprocessor.section_col
      • TranscriptPreprocessor.text_col
  • Module contents

Submodules

multimodal_fin.processing.basics module

class multimodal_fin.processing.basics.LLMClient(model, host='http://127.0.0.1:11500')

Bases: object

Client wrapper for interacting with Ollama models via the chat API.

This class provides:

• Automatic model name normalization.

• Automatic model download if not available locally.

• Configurable Ollama server host.

chat(messages, schema=None)

Send a list of messages to the model and retrieve the response.

Parameters:

• messages (List[dict]) – List of message dictionaries in Ollama format.

• schema (Optional[str]) – JSON schema to enforce structured responses.

Returns:

The content string of the model’s response.

Return type:

str

host: str | None = 'http://127.0.0.1:11500'

model: str

class multimodal_fin.processing.basics.UncertaintyMixin

Bases: object

Provides uncertainty estimation via majority voting.

get_result_and_uncertainty(predict_fn, text, n=5)

Estimates category and confidence using majority voting.

Parameters:

• predict_fn (Callable[[str], str]) – Prediction function to apply repeatedly.

• text (str) – The input text to classify.

• n (int) – Number of evaluations to perform.

Returns:

• The most frequent predicted category.

• Confidence score as percentage.

Return type:

Tuple[str, float]

multimodal_fin.processing.pipeline module

class multimodal_fin.processing.pipeline.ConferencePipeline(settings)

Bases: object

Orchestrates the full processing pipeline for a financial conference folder:

Steps performed:

1. Preprocessing of the transcript and section segmentation.

2. Text classification and question-answer (Q&A) annotation.

3. Multimodal embedding extraction (text, audio, video).

4. Metadata enrichment using LLMs (topics, Q&A analysis, coherence).

5. Result persistence in CSV and enriched JSON format.

run()

Run the processing pipeline on each conference folder path defined in the input CSV.

Return type:

None

multimodal_fin.processing.processor module

class multimodal_fin.processing.processor.Processor(sec10k_model_names, qa_analyzer_models, audio_model_name=None, text_model_name=None, video_model_name=None, num_evaluations=5, device='cpu', verbose=1)

Bases: object

Orchestrates the multimodal analysis pipeline in two main steps:

1. Embedding extraction for audio, text, and video.

2. Metadata enrichment (QA analysis, coherence, topics).

3. JSON serialization of enriched output.

process_and_save(input_csv_path, original_dir, output_json_path)

Executes the full multimodal pipeline and writes enriched results to a JSON file.

Parameters:

• input_csv_path (str) – Path to classified interventions CSV.

• original_dir (Path) – Directory containing LEVEL_3.json and audio/video files.

• output_json_path (str) – Destination path for saving the final JSON.

Return type:

dict

Returns:

A dictionary containing the enriched multimodal results.

Module contents