I investigate whether the domain shift incurred when transferring the language models on to the target task can be actively bridged using domain adversarial training. The experimental results suggest that it can be beneficial in low resource scenarios, but is redundant on tasks with ample training data.
We provide an overview of deep learning approaches for question answering over knowledge graphs where we discuss numerous family of approaches, and the challenges they tackle best. We hope this paper provides an excellent resource to get started with research in the field.
In this tutorial, we provided the participants with an overview of the field of Question Answering over Knowledge Graphs, insights into commonly faced problems, its recent trends and developments. Alongwith, a hands-on session aimed at introducing the participants with a barebones implementation of a neural ranking based QA system.
We propose a new neural ranking architecture for predicting the correct logical form given a question. We experiment with this schema using RNN, and Transformers (BERT) based encoders. We also demonstrate that domain adaption is an effective technique to offset the lack of training data in some scenarios.
We demonstrate that reading comprehension over large legal documents can be reasonably accomplished using a paragraph selection mechanism, and a neural question answering model.
Gold standard Question Answering dataset over DBpedia, with 5,000 datapoints. And a framework to substantially reduce human efforts involved in making one such.
Over 400 downloads as of Sept 2018.
Computing document similarity by computing local alignment in the documents' topics, thereby capturing and comparing its thematic flow.
We create a rule based automatic Description Logics (DL) ontology learner over constrained, unstructured text. We also propose mechanisms to check the validity of generated ontology.
We propose an algebraic similarity measure for assigning semantic similarity score to concept definitions in ALCH+ (a fragment of Description Logics).