As part of the implementation of the National Artificial Intelligence Plan and the initiative to create research and development infrastructure and advanced capabilities in the field of Natural Language Processing (NLP) in spoken Hebrew and Arabic, the Innovation Authority today announced its approval of 17 different projects in industry and academia, with a total budget of approximately 30 million NIS (~$8 million USD), which will facilitate the leap in AI applications in Hebrew and Arabic, requiring natural language processing as part of the solution.
New solutions and applications in the field of artificial intelligence are emerging almost on a daily basis. However, the gap between existing capabilities for Hebrew and Arabic and more common languages (such as English) is significant.
To narrow this gap, a wide range of infrastructures are required for application developers dependent on language. This call for proposals has promoted infrastructure activity to implement natural language-based artificial intelligence technologies and called on researchers, developers, academia, and industry to rise to this challenge.
Funding for approved projects will provide infrastructures and data repositories in general and specific areas, enabling a wide range of tasks from academia and industry.
Here are the approved projects:
These projects will work to reduce existing gaps in the capabilities of Hebrew and Arabic language processing and some of them even challenge the global State of The Art in the field.
New solutions and applications in the field of artificial intelligence are emerging almost on a daily basis. However, the gap between existing capabilities for Hebrew and Arabic and more common languages (such as English) is significant.
To narrow this gap, a wide range of infrastructures are required for application developers dependent on language. This call for proposals has promoted infrastructure activity to implement natural language-based artificial intelligence technologies and called on researchers, developers, academia, and industry to rise to this challenge.
Funding for approved projects will provide infrastructures and data repositories in general and specific areas, enabling a wide range of tasks from academia and industry.
Here are the approved projects:
- Verbit will develop and provide a transcription and summarization model for clinical and business applications.
- Dr. Amir David Nissan HaCohen and Dr. Avi Katschulero will develop and provide a model that enables chatbot creation and semantic retrieval of documents based on queries.
- Binat will develop and provide a speech-to-text service with additional capabilities such as speaker separation, real-time alerts, and more.
- Clalit Health Services will create a very large and unique international-level database that combines visual data with written interpretation of the image while meticulously preserving data integrity.
- Briya will develop and provide a language model for analyzing medical texts in Hebrew combined with English. Project partners include Assuta Ashdod Hospital, Galilee Medical Center, Sourasky Medical Center Tel Aviv, and Shaare Zedek Medical Center.
- Sheba Impact will develop and provide Clinical Bert, a language model trained on clinical textual data from Sheba Hospital (over 100 million medical records).
- Prof. Ruth Apter will develop a model that allows the generation of structured information from unstructured data.
- Dr. Amos Azaria (Ariel University) will develop and provide a translation algorithm that allows the use of pre-trained models in English for queries in Hebrew.
- Dr. Haya Libskind (Lev Academic Center) will develop and provide a model for identifying harmful content in texts.
- Dr. Amos Azaria (Ariel University) will develop and provide an algorithm for detecting hallucinations (false positives) in Hebrew language models.
- Dr. Arnon Shetrom (Ben-Gurion University of the Negev) will develop and provide an entity database in the legal field and an automatic tagging model.
- Tel Aviv University and Ichilov Hospital will develop and provide a medical model in Hebrew based on medical summaries and time relationship classification between two events.
- Ben-Gurion University of the Negev and Sami Shamoon School of Engineering will develop and provide tools for certifying service-oriented language models.
- Prof. Ido Dagan (Bar Ilan University) will develop and provide infrastructure for dynamic execution of queries based on semantic relationships (similar to the use of structured data and knowledge).
- Prof. Lior Rokach (Bar Ilan University) will develop and provide bidirectional translation models for Hebrew and Arabic languages.
- Prof. Koby Gell (Ben-Gurion University of the Negev) will develop and provide a model for detecting distress.
- Dr. Shai Fine (Reichman University) will develop and provide a model that allows extraction of information from recorded audio.
These projects will work to reduce existing gaps in the capabilities of Hebrew and Arabic language processing and some of them even challenge the global State of The Art in the field.