Makale gönderimi için sinir ağ tabanlı yayın tavsiyesi

This thesis is about recommending the best journal to authors for article submission. This task is a challenge because of the need to ensure that the journal is relevant. The significance of the relevancy of a journal is that an article published in a less relevant journal will have less exposure to the intended target audience, and consequently have less of an impact. The conceptual content contained in the abstracts of articles accepted for publication in a journal can be used to characterize that journal, which can be thought of as a "finger print" or "signature" of the journal. In turn, the content of these abstracts can be imprinted in the weights of neural network models. The abstracts are first converted to vector representations obtained through the methods of natural language processing, in order to be used with neural networks. The main purpose of this work will be exploring neural network architectures for discovering and recommending appropriate journals to those seeking to publish their research. In this thesis, the current state of the art will be extended and a robust and generic article-to-journal matching tool for all publishers, using Web of Science data, will be proposed.