I am Jeremy, a first-year MPhil student supervised by Dr Matthew Shardlow, and I am working on natural language processing, which is a cross-subject of computer science and linguistics.
I believe anyone interested in sci-fiction must be familiar with the word “artificial intelligence (AI)”. Although you might be still not very clear about what AI really is, it is shaping our lives in the next decade silently. Voice assistance such as Siri, Alexa and Google Assistance is just one example of how AI is changing our lives, and how these AI can understand your instructions is one focus of natural language processing (NLP). The natural language or human language, which is the one you are reading, is different from the language used by AI and hard for them to understand, so we need to build a bridge of communication between humans and AI and this topic is called natural language understanding. In addition to this, NLP covers nearly all topics related to natural language in verbal or text, including natural language understanding, question and answering, machine translation, etc.
The specific topic I am currently working on is text simplification. The idea of text simplification is rather simple — to increase the readability of text. Text simplification aims to reduce the obstacle for humans or even AI to understand natural language. It can help children or people who have difficulties in reading to comprehend the text easily and even assist untrained people in learning and understanding expert content.
Actually, there is research showing that many elder people are suffering from dyslexia in different extent. The reason why I am interested in this topic is that it might be a good way to smooth the learning difficulty curve and help people to read smoothly in many different fields. For now, text simplification at the sentence level mainly refers to the operations to reduce the complexity at both lexical and syntactical levels. Apart from this, it may also cover the explanation of technical terms and simplification at the text level.
Although the idea for this task is simple, the potential for this technique is quite promising. Imagining there is a teacher who explains all new concepts or ideas to you with plain and simple sentences, which might prompt the learners to master a new skill faster. It requires combined techniques in NLP to achieve this goal, including natural language understanding, question answering and text simplification.
To give you a quick example, here is a link to an existing project about explaining academic papers — https://www.explainpaper.com/. Although the output of the current model is not perfect, it still shows the potential of this field.
If you found yourself interested in the related topics, please join the community on Hugging Face.
