In this study, we proposed a method for generating suggested keywords for search behavior analysis using large-scale language models. While search engine suggestion functions provide important clues for understanding user search intent, the proposed method aimed to improve search behavior analysis accuracy by generating more diverse and comprehensive suggested keywords utilizing the knowledge of large-scale language models.

In this study, we proposed a new method for effective keyword extraction from text. While conventional methods extract keywords based on word frequency and statistical features, they do not fully utilize surrounding context information. The proposed method achieved more context-appropriate keyword extraction by acquiring and utilizing phrase-level context information surrounding words.

In this study, we proposed corpus construction and paraphrase generation methods for Japanese grammatical simplification (Easy Japanese). We developed a method to simplify grammatical structures while preserving the meaning of sentences for users who have difficulty understanding complex grammatical structures, such as Japanese language learners and the elderly. We constructed a corpus consisting of pairs of sentences before and after simplification, and evaluated a paraphrase generation model using this corpus.

In this study, we constructed a Japanese grammatical simplification corpus and worked on automatic grammatical simplification using this corpus. In addition to the machine translation approach traditionally used for automatic simplification, we proposed a method that applies grammatical transformation rules acquired from the corpus, as well as a hybrid approach combining both methods. These approaches were compared through both automatic and human evaluation. Although the proposed rule-based method scored lower than the machine translation approach, the hybrid method achieved the highest score, demonstrating its effectiveness in grammatical simplification.

This study reported on the construction of a corpus for simplifying Japanese grammatical structures. Focusing on grammatical structures rather than vocabulary, we created a parallel corpus consisting of pairs of original and simplified sentences. This corpus can be used to develop systems that automatically simplify Japanese sentences with complex grammatical structures.

This study reported on the construction of a parallel corpus of general Japanese and Easy Japanese. Easy Japanese is a simplified form of Japanese that is easier for non-native Japanese speakers, children, and the elderly to understand. This corpus is useful for developing systems that automatically convert to Easy Japanese.