Mamoru Komachi Laboratory
Associate Professor Mamoru Komachi
Natural Language Processing (machine translation) / Computer Assisted Language Learning / Information Extraction / Web Mining / Machine Learning
Due to the advancement of computers and the rise of big data, research and development in the field of artificial intelligence are attracting global attention. Natural language processing, or computational linguistics, is a research field that uses computers to understand and generate human language. Words are represented mathematically and processed by computer programs. Natural language processing provides fundamental theories and technologies for various applications, including search engines, machine translation, dialogue systems, automatic summarization, and predictive text. Let's study cutting-edge artificial intelligence in our university.
Prospective Undergraduate Students
Natural language processing uses computers to analyze human language. It is an essential part of the information system you use every day, including predictive text, web search engines, and recommendation systems. It is one of the everyday technologies that we do not even realize the importance because it is too natural, but we will be in trouble if it is not there.
Many of the students in our laboratory are science and engineering majors because they use mathematics to understand language, but there are also people from various humanities and social sciences majors such as linguistics, education, law, and economics. Natural language processing is an interdisciplinary field. Figure 1 shows a large-scale incorrect example search engine developed in our laboratory for Japanese language learners. It is implemented by a student from the department of foreign languages. Also, 30% of the graduate students have work experience and 30% of the PhD students are international students in our laboratory.
Prospective PhD Students
In the field of natural language processing, the concept of deep learning has emerged since 2012, surpassing the performance of conventional methods in a wide range of applications, including machine translation. One of the biggest benefits of deep learning in natural language processing is a task called language generation, in which the output is text. Besides machine translation, many language generation tasks witness dramatic improvements including document summarization, dialogue systems, and image caption generation. In the past, it was difficult to obtain fluent output with statistical methods, but by using deep learning, fluent text can be obtained, and language processing is attracting attention.
In addition to mathematical formulation and programming frameworks, the success of deep learning is supported by large-scale high-quality data and appropriate evaluation metrics. Our research group focuses on deep learning techniques that can be applied even in the absence of large-scale data, on the construction of high-quality language resources that can be used widely in addition to deep learning, and on quantitative evaluation measures for language generation. Figure 2 shows a neural machine translation system for web texts developed in our laboratory. As one of the few universities in Japan that has a machine translation research group, we regularly participate in machine translation shared tasks at international conferences and workshops.
