TL;DR
This paper introduces a novel Somali-Italian terminological resource built by combining historical explorer notebooks with a contemporary Somali corpus. By leveraging Semantic Web technologies and the Linked Open Data paradigm, the authors create a valuable linguistic asset that enriches our understanding of Somali language evolution and addresses its under-resourced status in the digital humanities.
What Problem Does This Paper Address?
Somali, despite being a widely spoken Cushitic language, is considered "under-resourced" in terms of available annotated datasets and language resources for NLP and AI research. This scarcity hinders comprehensive linguistic analysis and development. Furthermore, understanding the historical evolution of the language, particularly how it was documented by early explorers and how it has changed over time, is challenging due to the lack of accessible, structured historical data. This paper aims to bridge this gap by creating a structured terminological resource that links historical linguistic data with contemporary usage.
Methods
The core methodology involves constructing a termino-ontological resource using Semantic Web technologies. This means the resource has both linguistic (terminological) and conceptual (ontological) dimensions.
- Data Extraction: Terminological data was extracted from the notebooks of Italian explorer Ugo Ferrandi, written during his stay in Somalia in the late 19th century.
- Semantic Web Technologies:
- RDF (Resource Description Framework): Used to represent the data in a structured, graph-like format.
- OWL (Web Ontology Language): Employed to formally describe the prevailing conceptualization of the world in Somalia at that time, creating a foundational ontology.
- SPARQL: The query language used to interact with and retrieve data from the RDF store.
- OntoLex-Lemon Model: This standard RDF-based model was adopted for representing the lexical data, allowing for detailed descriptions of linguistic characteristics while keeping linguistic and conceptual dimensions separate.
- Contemporary Corpus Integration: Whenever feasible, terms from Ferrandi's lexicon were linked to a contemporary Somali corpus (created by Musse Jama) using
rdfs:sameAs. This enriched the historical data with modern usage and allowed for diachronic analysis. - Data Conversion Pipeline: A crucial technical contribution is a conversion pipeline from proprietary corpus formats to CONLL-U, and then to OntoLex-Lemon. This process makes the resource versatile and applicable to other similar data.
Datasets
- Ugo Ferrandi's Notebooks: The primary source of historical Somali-Italian terminological data. These notebooks were published in 1903 as "Lugh. Emporio Commerciale sul Giuba."
- Contemporary Somali Corpus: A large, syntactically annotated corpus created by Musse Jama, containing over seven million words. This corpus provided contemporary Somali lexicon for enrichment.
Key Insights and Ideas
- Bridging Historical and Contemporary Language: The most significant insight is the successful integration of historical linguistic data (Ferrandi's notebooks) with a contemporary corpus. This allows for tracing linguistic evolution, identifying archaic terms, and understanding semantic shifts over time.
- The Power of Linked Open Data (LOD): By adhering to the LOD paradigm and using Semantic Web technologies, the resource becomes FAIR (Findable, Accessible, Interoperable, Reusable), contributing to the broader Linguistic Linked Data cloud and making Somali data more discoverable.
- Historical Contextualization of Culture: Ferrandi's notebooks offer a unique window into the material and immaterial culture of Somali tribes at the dawn of the 20th century, providing rich context for the documented terminology.
- Dialectal Nuances: The paper highlights the importance of understanding dialectal variations, noting Ferrandi's use of terms that likely correspond to the Maay dialect and common Somali, and the need for further investigation into the "Lughiano" dialect.
- Challenges of Historical Data: The authors acknowledge the inherent challenges of working with historical data, including inconsistent spellings due to the lack of a standardized alphabet and the need to "purge" cultural biases and stereotypes present in early explorer narratives.
Key Results
- Construction of a Somali-Italian Terminological Resource: The primary result is the creation of this structured, bilingual resource.
- Extraction of over 400 Terms: Manually extracted terms from Ferrandi's notebooks, covering a wide range of semantic fields (e.g., cooking tools, weapons, agricultural tools).
- Demonstration of SPARQL Queries: The paper illustrates how SPARQL queries can be used to explore the resource, for example, identifying terms borrowed from Swahili related to artefacts.
- Development of a Data Conversion Pipeline: A functional tool for converting data from proprietary formats to CONLL-U and then to OntoLex-Lemon, which is presented as a valuable asset for similar projects.
Limitations and Future Work
- Incomplete Formalization: Not all terms and concepts from Ferrandi's lexicon have been fully formalized yet. The project is scheduled for completion in 2025.
- Validation of Historical Data: The linguistic data from Ferrandi requires continuous and meticulous validation due to potential inaccuracies in transcription and his limited understanding of local dialects.
- "Lughiano" Dialect Identification: Further research is needed to definitively identify the dialect Ferrandi referred to as "Lughiano."
- Future Accessibility: Once compiled, the data will be accessible on the CLARIN platform, offering advanced query and visualization tools. The authors envision this resource as a significant contribution to understanding Somali language and culture.
Summary updated