Loading...

Natural Language Inference in Persian Texts based on Natural Logic

Naji, Kiavash | 2020

400 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 53329 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Izadi, Mohammad
  7. Abstract:
  8. The progress and development in the field of Natural Language Processing (NLP) have led to the emergence of Natural Language Understanding (NLU). One of the main objectives in this field is inference, deduction, or entailment from the phrases (sentences) of a natural language. The crucial question here is whether it is possible to semantically infer a natural language hypothesis from a set of natural language premises. Since in many applications of Natural Language Inference (NLI) in the real world, it has a decisive role, this problem is of the utmost importance. Furthermore, the previously proposed approaches to tackle this problem incorporate a wide range of classic logic-based and new learning-based methods, which ultimately are fundamentally disparate. In this study, we will first present a thorough literature review of the different proposed methods to solve this problem including their datasets, then we will focus on two of these methods that are based on logic and share the same theoretical basis of natural logic, namely the MacCartney’s semantic model and the Abzianidze’s proof system. We will go through each of these methods and explain them one by one. Having reviewed these methods, we will compare them and afterwards, investigate the feasibility of implementing these two methods in the Persian language. Next, we will present a semantic method which systematically is similar to natural logic considering its logical system. Our proposed method is highly scalable and can be implemented more independent of external tools. Finally, we will present an experimental analysis estimation of this method to estimate its implementation cost and feasibility. This study is the first step of NLI in the Persian language. It gives a brief introduction to this problem, offers an evaluation of different approaches, and hopefully, provides the necessary foundation for future research in this field
  9. Keywords:
  10. Natural Language Understanding (NLU) ; Recognizing Textual Entailment (RTE) ; Semantic Model ; Natural Proof System ; Natural Language Inference (NLI) ; Natural Logic

 Digital Object List

 Bookmark

No TOC