arXiv subjects

Hiroshi Matsuda

Publications and source records attributed to Hiroshi Matsuda.

10 recordsLinked to original sources

arXiv2025

Step-by-step Instructions and a Simple Tabular Output Format Improve the Dependency Parsing Accuracy of LLMs

Recent advances in large language models (LLMs) have enabled impressive performance in various tasks. However, standard prompting often struggles to produce structurally valid and accurate outputs, especially in dependency parsing. We propose a novel step-by-step instruction strategy, where universal part-of-speech tagging precedes the prediction of syntactic heads and dependency labels, and a simplified CoNLL-U like output format, our method achieves state-of-the-art accuracy on Universal Dependencies datasets across 17 languages without hallucination or contamination. We further show that multilingual fine-tuning simultaneously improves cross-language generalization performance. Our results highlight the effectiveness of explicit reasoning steps in LLM-based parsing and offer a scalable, format-consistent alternative to bracket-based approaches.

cs.CL↗

arXiv2024

LLM-jp: A Cross-organizational Project for the Research and Development of Fully Open Japanese LLMs

This paper introduces LLM-jp, a cross-organizational project for the research and development of Japanese large language models (LLMs). LLM-jp aims to develop open-source and strong Japanese LLMs, and as of this writing, more than 1,500 participants from academia and industry are working together for this purpose. This paper presents the background of the establishment of LLM-jp, summaries of its activities, and technical reports on the LLMs developed by LLM-jp. For the latest activities, visit https://llm-jp.nii.ac.jp/en/.

cs.CL↗

arXiv2020

Towards Productionizing Subjective Search Systems

Existing e-commerce search engines typically support search only over objective attributes, such as price and locations, leaving the more desirable subjective attributes, such as romantic vibe and worklife balance unsearchable. We found that this is also the case for Recruit Group, which operates a wide range of online booking and search services, including jobs, travel, housing, bridal, dining, beauty, and where each service is among the biggest in Japan, if not internationally. We present our progress towards productionizing a recent subjective search prototype (OpineDB) developed by Megagon Labs for Recruit Group. Several components within OpineDB are enhanced to satisfy production demands, including adding a BERT language model pre-trained on massive hospitality domain review corpora. We also found that the challenges of productionizing the system are beyond enhancing the components. In particular, an important requirement in production-quality systems is to instrument a proper way of measuring the search quality, which is extremely tricky when the search results are subjective. This led to the creation of a high-quality benchmark dataset from scratch, involving over 600 queries by user interviews and a collection of more than 120,000 query-entity relevancy labels. Also, we found that the existing search algorithms do not meet the search quality standard required by production systems. Consequently, we enhanced the ranking model by fine-tuning several search algorithms and combining them under a learning-to-rank framework. The model achieves 5%-10% overall precision improvement and 90+% precision on more than half of the benchmark testing queries making these queries ready for AB-testing. While some enhancements can be immediately applied to other verticals, our experience reveals that benchmarking and fine-tuning ranking algorithms are specific to each domain and cannot be avoided.

cs.DB↗

arXiv2019

2-knot homology and Yoshikawa move

Ng constructed an invariant of knots in ${\mathbb{R}}^3$, a combinatorial knot contact homology. Extending his study, we construct an invariant of surface-knots in ${\mathbb{R}}^4$ using marked graph diagrams.

math.GT↗

arXiv2019

2-knot homology and Roseman move

math.GT↗

arXiv2009

Morse-Novikov theory, Heegaard splittings and closed orbits of gradient flows

The works of Donaldson and Mark make the structure of the Seiberg-Witten invariant of 3-manifolds clear. It corresponds to certain torsion type invariants counting flow lines and closed orbits of a gradient flow of a circle-valued Morse map on a 3-manifold. We study these invariants using the Morse-Novikov theory and Heegaard splitting for sutured manifolds, and make detailed computations for knot complements.

math.GT↗

arXiv2007

On rectangular diagrams, Legendrian knots and transverse knots

A correspondence is studied by H. Matsuda between front projections of Legendrian links in the standard contact structure for 3-space and rectangular diagrams. In this paper, we introduce braided rectangular diagrams, and study a relationship with Legendrian links in the standard contact structure for 3-space. We show Alexander and Markov Theorems for Legendrian links in 3-space.

math.GT↗

arXiv2003

Knot Floer homology of (1,1)-knots

We present a combinatorial method for a calculation of knot Floer homology with Z-coefficient of (1,1)-knots, and then demonstrate it for non-alternating (1,1)-knots with ten crossings and the pretzel knots of type (-2,m,n). Our calculations determine the unknotting numbers and 4-genera of the pretzel knots of this type.

math.GT↗

arXiv2003

Average size of random polygons with fixed knot topology

We have evaluated by numerical simulation the average size $R_K$ of random polygons of fixed knot topology $K = \emptyset, 3_1, 3_1\sharp4_1$, and we have confirmed the scaling law $R^2_K \sim N^{2ν_K}$ for the number $N$ of polygonal nodes in a wide range; $N = 100$ -- 2200. The best fit gives $2 ν_K \simeq 1.11$ -- 1.16 with good fitting curves in the whole range of $N$. The estimate of $2 ν_K$ is consistent with the exponent of self-avoiding polygons. In a limited range of $N$ ($N \gtrsim 600$), however, we have another fit with $2 ν_K \simeq 1.01$ -- 1.07, which is close to the exponent of random polygons.

cond-mat.stat-mech↗

arXiv2001

On the Dominance of Trivial Knots among SAPs on a Cubic Lattice

The knotting probability is defined by the probability with which an $N$-step self-avoiding polygon (SAP) with a fixed type of knot appears in the configuration space. We evaluate these probabilities for some knot types on a simple cubic lattice. For the trivial knot, we find that the knotting probability decays much slower for the SAP on the cubic lattice than for continuum models of the SAP as a function of $N$. In particular the characteristic length of the trivial knot that corresponds to a `half-life' of the knotting probability is estimated to be $2.5 \times 10^5$ on the cubic lattice.

cond-mat.stat-mech↗