Hypothesis and objectives

Hypothesis

The initial hypothesis of this project is as follows: The effective integration of world and external knowledge in NLG architectures improves the commonsense reasoning capabilities of NLG systems. We consider that enhancing commonsense reasoning capabilities of NLG systems is needed to automatically produce accurate, correct, and reliable texts that will be in line with real facts. 

We propose a new generation of knowledge-enhanced NLG systems that overcome the “hallucination” phenomenon and avoid natural language generation that is “economical with the truth.” This is possible given that Transformers, a type of “end-to-end” architecture that predominates in text generation, can be fine-tuned, trained and adapted, so that: 1) they can learn to generalise and identify implicit information; and, 2) they can take into account certain characteristics that the generated text must reflect, such as structure, length, style, formality, etc., thus allowing a more controlled and accurate generation. These types of adjustments would help, on the one hand, to generate texts in a more natural, diverse, and semantic way, and on the other, to detect strange patterns or biases that should be avoided in the generated text. In addition, the control of different aspects of the text is key for successfully applying and transferring NLG systems to real scenarios that are relevant to industry and society (Len et al., 2020).

Objectives

The following specific objectives have been set to develop the project and advance the state of the art in NLG:

  • OB1. Gather and analyse the available generic existing language models and knowledge sources (structured and unstructured), identifying the type of information they contain and discovering new knowledge that can be inferred from them.
  • OB2. Determine what kind of knowledge is most appropriate to improve the NLG process, as well as
    compile, and adapt models that will enable the semantic enrichment of NLG approaches.
  • OB3. Research, propose and develop novel NLG approaches that can integrate, be guided by, orsimply use the obtained knowledge, thus leading to more accurate, flexible, and dynamic common sense and conscious generation approaches.
  • OB4. Propose and develop various scenarios and use cases that demonstrate the validity andapplication of the NLG task, together with the positive impact derived from the integration of commonsense knowledge in the proposed approaches.
  • OB5. Evaluate intrinsically and extrinsically each of the proposed techniques and approaches andscenarios with the most suitable standard metrics, or create novel metrics, if necessary.
  • OB6. Promote and disseminate the research results obtained from the project through differentnational and international media –including well indexed journals, conferences, seminars, etc., as well as exploit the potential for transferring this technology to society.