Voices Across Scripts: A Protocol for a Multilingual Urdu–Punjabi Suicide Discourse Corpus and the Cultural Calibration of Psycho-Forensic Linguistic Detection

Psycho-Forensic Linguistic Surveillance Initiative (PFLSI) established, respectively, through series papers 1, 2 and 3 of the a small culturally rich corpus of Pakistani suicide notes (N=140) and a large-scale, statistically rigorous English-language Suicide Discourse Corpus (452,000 tokens). Series Paper 3 unified these findings theoretically and specified the seven-layer PFLS pipeline architecture, identifying its Layer L3 Cultural Calibration module as dependent on lexical resources that do not yet exist at corpus scale: a validated, quantitatively analyzed body of Urdu- and Punjabi-medium suicidal discourse. This paper addresses that dependency directly. It is a corpus-construction and protocol paper, not a results paper: it specifies the design of a Multilingual Suicide Discourse Corpus (M-SDC) comprising native-script Urdu and Punjabi (Shanmukhi) text together with Roman-Urdu code-switched digital discourse, sourced through ethically governed partnerships with Pakistani crisis helplines, moderated mental-health forums, and re-digitized original-language transcription of the Series Paper 1 note corpus. We specify target corpus scale (a proposed 500,000-token Urdu/Punjabi Suicide Discourse Corpus and a matched 500,000-token control corpus), a code-switch annotation scheme adapted from the CALCS shared-task tradition, an izzat-register lexicon derived inductively from Series Paper 1's qualitative findings, and an inter-annotator reliability protocol targeting Cohen's kappa above 0.75 for semantic-domain tags. Because data collection has not yet been completed at the time of writing, we present anticipated findings as a set of pre-registered, directional hypotheses rather than as reported results, and we provide placeholder result tables whose structure mirrors that of Series Paper 2 so that outcomes will be directly comparable across languages once collection concludes. We contend that the absence of such a corpus limits the empirical validation of the PFLS cultural calibration layer and constrains automated detection in linguistically underrepresented populations.