PsychiatryNLPKit.analysis.sentence_level_perplexity

PsychiatryNLPKit.analysis.sentence_level_perplexity(sentence_generative_tokens, generative_model, sections=None, manage_lifecycle=True)[source]

Generative perplexity averaged across sentences per section.

Notes:

Theoretical basis - Perplexity can be computed both at the sentence and paragraph levels (Alqahtani et al., 2022). High perplexity predicts delusion and unusual thought content, as reflected by the total PANSS scores (He et al., 2024).

Args:

sentence_generative_tokens: Dict from: TextData.sentence_generative_tokens. Maps section names to batched token dicts (one per sentence).
generative_model: HFGenerativeLLM instance. Must be loaded unless: manage_lifecycle is True.

sections: Sections to process. None processes all sections in the dict. manage_lifecycle: If True (default), load and unload the model

internally. Set to False when the caller manages the lifecycle.

Returns:

Dict mapping section names to a dict with key "sentence_level_perplexity". Empty sections receive float("nan").

References:

Alqahtani, A., Kayi, E. S., Hamidian, S., Compton, M., & Diab, M. (2022, December). A quantitative and qualitative analysis of schizophrenia language. In A. Lavelli, E. Holderness, A. Jimeno Yepes, A. Minard, J. Pustejovsky, & F. Rinaldi (Eds.), Proceedings of the 13th International Workshop on Health Text Mining and Information Analysis (LOUHI) (pp. 173-183). Association for Computational Linguistics. https://doi.org/10.18653/v1/2022.louhi-1.20 He, R., Palominos, C., Zhang, H., Alonso-Sánchez, M. F., Palaniyappan, L., & Hinzen, W. (2024). Navigating the semantic space: Unraveling the structure of meaning in psychosis using different computational language models. Psychiatry Research, 333, 115752. https://doi.org/10.1016/j.psychres.2024.115752