Add TupleTimeTextProcessor for Temporal Text Data

[Rian354]

Add TupleTimeTextProcessor for Temporal Text Data

Summary/TLDR

Adds TupleTimeTextProcessor, a processor for handling clinical text paired with temporal information (time differences). Allows for modality routing in multimodal pipelines.

Key Features

• Temporal text handling: Processes (List[str], List[float]) tuples
• Automatic modality routing: type_tag enables automatic encoder selection
• Clean interface: Returns (texts, time_tensor, modality_tag)
• Registered processor: Available via string key "tuple_time_text"

I/O

Input:  Tuple[List[str], List[float]]
        - List[str]: Clinical text entries
        - List[float]: Time differences between entries

Output: Tuple[List[str], torch.Tensor, str]
        - List[str]: Same text entries (unmodified)
        - torch.Tensor: 1D float tensor of time differences
        - str: Type tag for modality routing (default: "note")

Files Added

• pyhealth/processors/tuple_time_text_processor.py (107 loc)
• tests/test_tuple_time_text_processor.py (1 test)
• docs/api/processors/pyhealth.processors.TupleTimeTextProcessor.rst (docs)

Files Modified

• pyhealth/processors/__init__.py (export added)
• docs/api/processors.rst (documentation index)
• examples/text_embedding_tutorial.ipynb (added example section)

Example Usage

from pyhealth.processors import TupleTimeTextProcessor

# Initialize processor with modality tag
processor = TupleTimeTextProcessor(type_tag="clinical_note")

# Process temporal text data
texts = ["Admission note", "Progress note", "Discharge summary"]
time_diffs = [0.0, 24.0, 72.0]  # hours since admission

texts_out, time_tensor, tag = processor.process((texts, time_diffs))
# texts_out: ["Admission note", "Progress note", "Discharge summary"]
# time_tensor: tensor([0., 24., 72.])
# tag: "clinical_note"

Testing

pytest tests/test_tuple_time_text_processor.py -v
# 1 passed

[jhnwu3]

Also pretty much lgtm, thanks!