Yazar "Nasar, Zara" seçeneğine göre listele

Listeleniyor 1 - 1 / 1
  • Küçük Resim
    Yayın
    Information extraction from full-length scientific articles / Tam metin bilimsel makalelerden bilgi çıkarımı
    (Maltepe Üniversitesi, 2019) Nasar, Zara
    In last few decades, advent of computers and later internet has changed human civilization dramatically. Now we live in the world which is being overloaded with the data and the information. This information overload is posing new challenges to human intellect and hence creating opportunities for innovation. Scientific research is one of key beneficiaries under these new trends. Recently a tremendous increase in scientific publications is observed due to increase in researchers across the globe. This growth in scientific content consequently results in various challenges during determination of underlying key-insights from bulk of scientific articles. Hence, this process of reading and extracting key information nuggets out of multiple research articles has become a quite laborious, time consuming and cumbersome job for researchers. Therefore, a dire need is felt to automatically extract potential information out of immense set of research papers. Hence, in this study, aim is to develop a system that would be able to assist research community during literature review. Existing work in this regard, deals with key-insights extraction from single passage. This study, on the other hand, is focused on extracting key-insights from the full-length scientific articles. Therefore, in order to carry out this study, three major tasks are to be fulfilled. First task deals with the development of annotation guidelines, that describes the precise definitions of key-insights, to be extracted. Second task deals with annotation of complete scientific articles using devised annotation guidelines. Last task is focused on training of various machine learning modes to perform key-insights extraction. Current work applies three widelyused sequence labeling algorithms including Conditional Random Fields, Hidden Markov Models and Maximum Entropy Markov Models. These techniques are applied on self-annotated dataset of Computer Science scientific articles. Results show that a great deal of improvement is required in order to effectively carry out the task of automatic key-insights extraction from full-length scientific articles.