A Bird’s-eye View of Language Processing Projects at the Romanian Academy

A Bird’s-eye View of Language Processing Projects at the Romanian Academy Dan Tufiș1, Dan Cristea2,3 1Research Institute for Artificial Intelligence “Mihai Drăgănescu” of the Romanian Academy, 2Institute for Computer Science of the Romanian Academy 3 “Alexandru Ioan Cuza” University of Iaşi 113Calea 13 Septembrie, Bucharest, Romania 22 T. Codrescu St, 700481, Iaşi, Romania 311 Carol I Blv, 700506, Iaşi, Romania 1tufis@racai.ro, 2dcristea@info.uaic.ro Abstract This article gives a general overview of five AI language-related projects that address contemporary Romanian language, in both textual and speech form, language related applications, as well as collections of old historic documents and medical archives. Namely, these projects deal with: the creation of a contemporary Romanian language text and speech corpus, resources and technologies for developing human-machine interfaces in spoken Romanian, digitization and transcription of old Romanian language documents drafted in Cyrillic into the modern Latin alphabet, digitization of the oldest archive of diabetes medical records and dialogue systems with personal robots and autonomous vehicles. The technologies involved for attaining the objectives range from image processing (intelligent character recognition for hand-writing and old Romanian documents) to natural language and speech processing techniques (corpus compiling and documentation, multi-level processing, transliteration of different old scripts into modern Romanian, command language processing, various levels of speech-text alignments, ASR, TTS, keyword spotting, etc.). Some of these projects are approaching the end, others have just started and others are about to start. All the reported projects are national ones, less documented than the international projects we are/were engaged in, and involve large teams of experts and master/PhD students from computer science, mathematics, linguistics, philology and library sciences. Keywords: contemporary language reference corpus, speech corpus, natural language dialogue systems, intelligent character recognition, medical records archive 1. Introduction Language technology is a discipline needed in more and more areas. The computational linguistics traditional domain has been extended towards dealing with vast and diversified quantities of linguistic data, both written and spoken. This extension requires putting together expertise from natural language processing (NLP), Internet-of- Things (IoT), robotics, image processing, cognitive computing, High Performance Computing (HPC), semantic web, social media and many others. In this article we will give a brief account on several projects at the Romanian Academy, their prospects and outline the challenges posed by each of them. The common characteristic of these projects, apart from being all conducted by teams affiliated to different institutes of the Romanian Academy, is the use of a solid background of language resources, tools and models, acquired along the last two decades and permanently enhanced. As such, we argue that a constant accumulation of human expertise, language resources and software utilities can bring about a level that allows for the development of extremely complex projects. Without this know-how, data and processing platforms, these enterprises would have not been feasible. 2. CoRoLa The CoRoLa project is a priority program of the Romanian Academy and is developed by two institutes of the Romanian Academy: “Mihai Drăgănescu” Research Institute on Artificial Intelligence in Bucharest (RACAI) and the Institute of Computer …

Content truncated to 500 words