Hirotoshi Taira

Question Classification using HDAG Kernel

This paper proposes a machine learning based question classification method using a kernel function, Hierarchical Directed Acyclic Graph (HDAG) Kernel. The HDAG Kernel directly accepts structured natural language data, such as several levels of chunks and their relations, and computes the value of the kernel function at a practical cost and time while reflecting all of these structures. We examine the proposed method in a question classification experiment using 5011 Japanese questions that are labeled by 150 question types. The results demonstrate that our proposed method improves the performance of question classification over that by conventional methods such as bag-of-words and their combinations.

A Japanese Predicate Argument Structure Analysis using Decision Lists

This paper describes a new automatic method for Japanese predicate argument structure analysis. The method learns relevant features to assign case roles to the argument of the target predicate using the features of the words located closest to the target predicate under various constraints such as dependency types, words, semantic categories, parts of speech, functional words and predicate voices. We constructed decision lists in which these features were sorted by their learned weights. Using our method, we integrated the tasks of semantic role labeling and zero-pronoun identification, and achieved a 17% improvement compared with a baseline method in a sentence level performance analysis.

Assigning gene ontology categories (GO) to yeast genes using text-based supervised learning methods

We propose a method for assigning upper level gene ontology terms (GO categories) to genes using relevant documents. This method represents each gene as a vector using relevant documents to the gene. Then, binary classifiers are made for the GO categories using such supervised learning methods as support vector machines and maximum entropy method. We applied this method for assigning GO categories to yeast genes and achieved an average F-measure of 0.67, which is > 0.3 higher than the existing method developed by Raychaudhun et al. We also applied this method to genome-wide annotation for yeast by all GO Slim categories provided by SGD and achieved average F-measures of 0.58, 0.72, and 0.60, respectively, for the three GO parts: cellular component, molecular function, and biological process.

Text Categorization Using Transductive Boosting

In natural language tasks like text categorization, we usually have an enormous amount of unlabeled data in addition to a small amount of labeled data. We present here a transductive boosting method for text categorization in order to make use of the large amount of unlabeled data efficiently. Our experiments show that the transductive method outperforms conventional boosting techniques that employ only labeled data.