The Semitic family includes languages and dialects spoken by a large number of native speakers (around 300 million). Prominent members of this family are Arabic (and its varieties), Hebrew, Amharic, Tigrinya, Aramaic, Maltese and Syriac. Their shared ancestry is apparent through pervasive cognate sharing, a rich and productive pattern-based morphology, and similar syntactic constructions. In addition, there are several languages which are used in the same geographic area such as Amazigh or Coptic, which, while not Semitic, have common features with Semitic languages, such as borrowed vocabulary.
The recent surge in computational work for processing Semitic languages, particularly Modern Standard Arabic (MSA) and Modern Hebrew (MH), has brought modest improvements in terms of actual empirical results for various language processing components (e.g., morphological analyzers, parsers, named entity recognizers, audio transcriptions, etc.). Apparently, reusing existing approaches developed for English or French for processing Semitic language text/speech, e.g., Arabic parsing is not as straightforward as initially thought. Apart from the limited availability of suitable language resources, there is increasing evidence that Semitic languages demand modeling approaches and annotations that deviate from those found suitable for English/French. Issues such as the pattern-based morphology, the frequently head-initial syntactic structure, the importance of the interface between morphology and syntax, and the difference between spoken and written forms (especially in Colloquial Arabic(s)) exemplify the kind of challenges that may arise when processing Semitic languages. For language technologies, such as information retrieval and machine translation, these challenges are compounded by sparse data and often result in poorer performance than for other languages.
This Workshop intends to follow on topics of paramount importance for Semitic-language NLP that were discussed at previous events (LREC, MEDAR/NEMLAR Conferences, the workshops of the ACL Special Interest Group for Semitic languages, etc.) and which are worth revisiting.
The workshop will bring together people who are actively involved in Semitic language processing in a mono- or cross/multilingual context, and give them an opportunity to update the community through reports on completed or ongoing work as well as on the availability of LRs, evaluation protocols and campaigns, products and core technologies (in particular open source ones). We also invite authors to address other languages spoken in the Semitic language area (languages such as Amazigh, Coptic, etc.). This should enable participants to develop a common view on where we stand and to foster the discussion of the future of this research area. Particular attention will be paid to activities involving technologies such as Machine Translation and Cross-Lingual Information Retrieval/Extraction, Summarization, etc. Evaluation methodologies and resources for evaluation of HLT will be also a main focus.
We expect to elaborate on the HLT state of the art, identify problems of common interest, and debate on a potential roadmap for the Semitic languages. Issues related to sharing of resources, tools, standards, sharing and dissemination of information and expertise, adoption of current best practices, setting up joint projects and technology transfer mechanisms will be an important part of the workshop.