Mihai Dascalu - ReadME – Improving Writing Skills through Personalized Feedback

Organized by: 
Dominique Vaufreydaz
Mihai Dascalu

Detailed information: 


Mihai Dascalu (1400+ citations, h-index 19) was head of class in 2009 (i.e., GPA 10/10; ranked 1st across specialization and university) at University Politehnica of Bucharest and holds a double PhD with the highest distinctions in Computer Science (Excellent, UPB) and Educational Sciences (Très Honorable avec Felicitations, University Grenoble Alpes, France), with his thesis published as book in Springer, Studies in Computational Intelligence. He has extensive experience in national and international research projects (POC D HUB-TECH, POC G NETIO, H2020 RAGE, ERASMUS+ ENeA-SEA, FP7 LTfLL, FP7 ERRIC and CNCSIS K-TEAMS) with more than 150 published papers, including 23 articles at top-tier conferences (AAAI, CogSci, AIED, ITS, CSCL), 49 paper indexed ISI at renowned international conferences (ICALT, EC-TEL, ICWL, ISPDC, AIMSA), and 7 Q1 journals (ijCSCL, Elsevier Computers in Human Behavior). Currently Mihai is an associate professor at UPB, responsible for the courses of Object Oriented Programming, Semantic Web Applications, and Data Mining and Data Warehousing. Complementary to his competencies in NLP, technology-enhanced learning (TEL) and discourse analysis, Mihai holds a multitude of professional certifications (e.g. PMP, PMI-RMP, PMI-ACP, CBAP, CISA, C|EH and CISSP) and extensive experience on strategic projects on non-refundable funds (EU, WB, USTDA). Moreover, Mihai has received the distinction “IN TEMPORE OPPORTUNO” in 2013 as the most promising young researcher in UPB, has obtained a Senior Fulbright scholarship in 2015, and holds the US patent # 9734144 B2 " Three-Dimensional Latent Semantic Analysis".



Writing is a central learning activity that requires both practice and experience. However, providing comprehensive feedback to students about their writing is a cumbersome and time-consuming task that can dramatically impact the learning outcomes and learners' performance. The aim of this talk is to introduce our Automated Writing Evaluation system called ReadMe which supports in its current release Romanian and English languages. Based on an extensible rule engine system, ReadMe provides personalized feedback at four granularity levels: word, sentence, paragraph, and document. The open-source ReaderBench framework (http://www.readerbench.com) generates hundreds of textual complexity indices which are afterwards grouped using a PCA and are considered as input for our custom rule-based engine. Personalized feedback is provided using a user-friendly web interface that facilitates evaluation based on a combination of text color highlights, coupled with comprehensive text suggestions for improvement.