Mining the Pre-1948 Palestinian Press: Unsupervised Keyphrase Extraction and Temporal Discourse Analysis from Five Historical Arabic Newspapers

Abstract

The Palestinian Arabic-language press of the late Ottoman and British Mandate periods constitutes a rich but computationally under-explored archive for studying the evolution of political, cultural, and social discourse in pre-1948 Palestine. This paper presents an end-to-end pipeline for extracting and analyzing thematic content from five historically significant Palestinian newspapers: Lisān al-ʿArab (اﻟﻌﺮب ﻟﺴﺎن), Al-Bushrā (اﻟﺒﺸﺮى), Al-Karmil (اﻟﻜﺮﻣﻞ), Al-Difāʿ (اﻟﺪﻓﺎع), and Filasṭīn (ﻓﻠﺴﻄﻴﻦ). We describe (i) the construction of a five-source corpus from scanned newspaper images obtained from archival collections, processed using the Google Cloud Vision OCR (GCV-OCR) API, (ii) the adaptation of KeyBERT with an AraBERT backbone for unsupervised keyphrase extraction, and (iii) a purpose-built Python visualization toolkit that produces keyword-frequency heatmaps, longitudinal trend charts, and ranked bar charts with full Arabic script rendering. Experiments across the five subcorpora show that the pipeline yields topically diverse keyphrases reflecting each newspaper’s editorial orientation and its distinct representation of Palestinian native perspectives—from pan-Arab nationalism and anti-colonial resistance to religious and communal affairs. Temporal analysis reveals event-responsive patterns that align with major historical developments, including the 1936–1939 Arab Revolt and the intensification of sovereignty discourse toward 1948. The pipeline, data format specifications, and visualization code are provided as supplementary material.

Resources

Details