Hiroki Sakaji

Hiroki Sakaji, Jason Bennett, Risa Murono et al.

What would happen if temperatures were subdued and result in a cool summer? One can easily imagine that air conditioner, ice cream or beer sales would be suppressed as a result of this. Less obvious is that agricultural shipments might be delayed, or that sound proofing material sales might decrease. The ability to extract such causal knowledge is important, but it is also important to distinguish between cause-effect pairs that are known and those that are likely to be unknown, or rare. Therefore, in this paper, we propose a method for extracting rare causal knowledge from Japanese financial statement summaries produced by companies. Our method consists of three steps. First, it extracts sentences that include causal knowledge from the summaries using a machine learning method based on an extended language ontology. Second, it obtains causal knowledge from the extracted sentences using syntactic patterns. Finally, it extracts the rarest causal knowledge from the knowledge it has obtained.

Masahiro Suzuki, Hiroki Sakaji

We introduce JFinTEB, the first comprehensive benchmark specifically designed for evaluating Japanese financial text embeddings. Existing embedding benchmarks provide limited coverage of language-specific and domain-specific aspects found in Japanese financial texts. Our benchmark encompasses diverse task categories including retrieval and classification tasks that reflect realistic and well-defined financial text processing scenarios. The retrieval tasks leverage instruction-following datasets and financial text generation queries, while classification tasks cover sentiment analysis, document categorization, and domain-specific classification challenges derived from economic survey data. We conduct extensive evaluations across a wide range of embedding models, including Japanese-specific models of various sizes, multilingual models, and commercial embedding services. We publicly release JFinTEB datasets and evaluation framework at https://github.com/retarfi/JFinTEB to facilitate future research and provide a standardized evaluation protocol for the Japanese financial text mining community. This work addresses a critical gap in Japanese financial text processing resources and establishes a foundation for advancing domain-specific embedding research.

Issey Sukeda, Masahiro Suzuki, Hiroki Sakaji et al.

In the ongoing wave of impact driven by large language models (LLMs) like ChatGPT, the adaptation of LLMs to medical domain has emerged as a crucial research frontier. Since mainstream LLMs tend to be designed for general-purpose applications, constructing a medical LLM through domain adaptation is a huge challenge. While instruction-tuning is used to fine-tune some LLMs, its precise roles in domain adaptation remain unknown. Here we show the contribution of LoRA-based instruction-tuning to performance in Japanese medical question-answering tasks. In doing so, we employ a multifaceted evaluation for multiple-choice questions, including scoring based on "Exact match" and "Gestalt distance" in addition to the conventional accuracy. Our findings suggest that LoRA-based instruction-tuning can partially incorporate domain-specific knowledge into LLMs, with larger models demonstrating more pronounced effects. Furthermore, our results underscore the potential of adapting English-centric models for Japanese applications in domain adaptation, while also highlighting the persisting limitations of Japanese-centric models. This initiative represents a pioneering effort in enabling medical institutions to fine-tune and operate models without relying on external services.

Masanori Hirano, Hiroki Sakaji, Kiyoshi Izumi

This study proposes a new generative adversarial network (GAN) for generating realistic orders in financial markets. In some previous works, GANs for financial markets generated fake orders in continuous spaces because of GAN architectures' learning limitations. However, in reality, the orders are discrete, such as order prices, which has minimum order price unit, or order types. Thus, we change the generation method to place the generated fake orders into discrete spaces in this study. Because this change disabled the ordinary GAN learning algorithm, this study employed a policy gradient, frequently used in reinforcement learning, for the learning algorithm. Through our experiments, we show that our proposed model outperforms previous models in generated order distribution. As an additional benefit of introducing the policy gradient, the entropy of the generated policy can be used to check GAN's learning status. In the future, higher performance GANs, better evaluation methods, or the applications of our GANs can be addressed.

Masahiro Suzuki, Masanori Hirano, Hiroki Sakaji

Instruction tuning is essential for large language models (LLMs) to become interactive. While many instruction tuning datasets exist in English, there is a noticeable lack in other languages. Also, their effectiveness has not been well verified in non-English languages. We construct a Japanese instruction dataset by expanding and filtering existing datasets and apply the dataset to a Japanese pre-trained base model. We performed Low-Rank Adaptation (LoRA) tuning on both Japanese and English existing models using our instruction dataset. We evaluated these models from both quantitative and qualitative perspectives. As a result, the effectiveness of Japanese instruction datasets is confirmed. The results also indicate that even with relatively small LLMs, performances in downstream tasks would be improved through instruction tuning. Our instruction dataset, tuned models, and implementation are publicly available online.

Takehiro Takayanagi, Hiroki Sakaji, Kiyoshi Izumi

Stock embedding is a method for vector representation of stocks. There is a growing demand for vector representations of stock, i.e., stock embedding, in wealth management sectors, and the method has been applied to various tasks such as stock price prediction, portfolio optimization, and similar fund identifications. Stock embeddings have the advantage of enabling the quantification of relative relationships between stocks, and they can extract useful information from unstructured data such as text and network data. In this study, we propose stock embedding enhanced with textual and network information (SETN) using a domain-adaptive pre-trained transformer-based model to embed textual information and a graph neural network model to grasp network information. We evaluate the performance of our proposed model on related company information extraction tasks. We also demonstrate that stock embeddings obtained from the proposed model perform better in creating thematic funds than those obtained from baseline methods, providing a promising pathway for various applications in the wealth management industry.

Meiyun Wang, Masahiro Suzuki, Hiroki Sakaji et al.

Large Language Models (LLMs) have demonstrated exceptional capabilities across various machine learning (ML) tasks. Given the high costs of creating annotated datasets for supervised learning, LLMs offer a valuable alternative by enabling effective few-shot in-context learning. However, these models can produce hallucinations, particularly in domains with incomplete knowledge. Additionally, current methods for knowledge distillation using LLMs often struggle to enhance the effectiveness of both teacher and student models. To address these challenges, we introduce DualChecker, an innovative framework designed to mitigate hallucinations and improve the performance of both teacher and student models during knowledge distillation. DualChecker employs ContextAligner to ensure that the context provided by teacher models aligns with human labeling standards. It also features a dynamic checker system that enhances model interaction: one component re-prompts teacher models with more detailed content when they show low confidence, and another identifies borderline cases from student models to refine the teaching templates. This interactive process promotes continuous improvement and effective knowledge transfer between the models. We evaluate DualChecker using a green innovation textual dataset that includes binary, multiclass, and token classification tasks. The experimental results show that DualChecker significantly outperforms existing state-of-the-art methods, achieving up to a 17% improvement in F1 score for teacher models and 10% for student models. Notably, student models fine-tuned with LLM predictions perform comparably to those fine-tuned with actual data, even in a challenging domain. We make all datasets, models, and code from this research publicly available.

Masahiro Suzuki, Hiroki Sakaji

Natural language processing (NLP) tasks in English and general domains are widely available and are often used to evaluate pre-trained language models. In contrast, fewer tasks are available for languages other than English and in the financial domain. Particularly, tasks in the Japanese and financial domains are limited. We develop two large datasets using data published by a Japanese central government agency. The datasets provide three Japanese financial NLP tasks, including 3- and 12-class classifications for categorizing sentences, along with a 5-class classification task for sentiment analysis. Our datasets are designed to be comprehensive and updated by leveraging an automatic update framework that ensures that the latest task datasets are publicly always available.

Hiroki Sakaji, Ryotaro Kobayashi, Kiyoshi Izumi et al.

In this paper, we attempt to summarize monthly reports as investment reports. Fund managers have a wide range of tasks, one of which is the preparation of investment reports. In addition to preparing monthly reports on fund management, fund managers prepare management reports that summarize these monthly reports every six months or once a year. The preparation of fund reports is a labor-intensive and time-consuming task. Therefore, in this paper, we tackle investment summarization from monthly reports using transformer-based models. There are two main types of summarization methods: extractive summarization and abstractive summarization, and this study constructs both methods and examines which is more useful in summarizing investment reports.

Takehiro Takayanagi, Masahiro Suzuki, Ryotaro Kobayashi et al.

Causality is fundamental in human cognition and has drawn attention in diverse research fields. With growing volumes of textual data, discerning causalities within text data is crucial, and causal text mining plays a pivotal role in extracting meaningful patterns. This study conducts comprehensive evaluations of ChatGPT's causal text mining capabilities. Firstly, we introduce a benchmark that extends beyond general English datasets, including domain-specific and non-English datasets. We also provide an evaluation framework to ensure fair comparisons between ChatGPT and previous approaches. Finally, our analysis outlines the limitations and future challenges in employing ChatGPT for causal text mining. Specifically, our analysis reveals that ChatGPT serves as a good starting point for various datasets. However, when equipped with a sufficient amount of training data, previous models still surpass ChatGPT's performance. Additionally, ChatGPT suffers from the tendency to falsely recognize non-causal sequences as causal sequences. These issues become even more pronounced with advanced versions of the model, such as GPT-4. In addition, we highlight the constraints of ChatGPT in handling complex causality types, including both intra/inter-sentential and implicit causality. The model also faces challenges with effectively leveraging in-context learning and domain adaptation. We release our code to support further research and development in this field.

Yuna Haseyama, Tomoki Ito, Hiroki Sakaji et al.

In high-stakes domains such as healthcare, the reliability of Large Language Models (LLMs) is critical, particularly when generating clinical insights from incident reports. This study proposes a tag-based few-shot example selection method for prompting LLMs to generate background/causal factors and preventive measures from details of the medical incidents. For our experiments, we use the Japanese Medical Incident Dataset (JMID), a structured dataset of 3,884 real-world medical accident and near-miss reports. These reports are variably annotated with a wide range of tags--some include descriptive information (e.g., "medications," "blood transfusion therapy"). We compare three few-shot example selection strategies--random sampling, cosine similarity-based selection, and our proposed tag-based method--using GPT-4o and LLaMA 3.3. Results show that the tag-based approach achieves the highest precision and most stable generation behavior, while similarity-based selection often leads to unintended outputs and safety filter activation. These findings suggest that selecting examples based on human-interpretable dataset tags can improve generation precision and stability in clinical LLM applications.

Hiroki Sakaji, Noriyasu Kaneda

In this study, we propose a methodology to extract, index, and visualize ``climate change narratives'' (stories about the connection between causal and consequential events related to climate change). We use two natural language processing methods, BERT (Bidirectional Encoder Representations from Transformers) and causal extraction, to textually analyze newspaper articles on climate change to extract ``climate change narratives.'' The novelty of the methodology could extract and quantify the causal relationships assumed by the newspaper's writers. Looking at the extracted climate change narratives over time, we find that since 2018, an increasing number of narratives suggest the impact of the development of climate change policy discussion and the implementation of climate change-related policies on corporate behaviors, macroeconomics, and price dynamics. We also observed the recent emergence of narratives focusing on the linkages between climate change-related policies and monetary policy. Furthermore, there is a growing awareness of the negative impacts of natural disasters (e.g., abnormal weather and severe floods) related to climate change on economic activities, and this issue might be perceived as a new challenge for companies and governments. The methodology of this study is expected to be applied to a wide range of fields, as it can analyze causal relationships among various economic topics, including analysis of inflation expectation or monetary policy communication strategy.

Kota Tanabe, Masahiro Suzuki, Hiroki Sakaji et al.

We construct an instruction dataset for the large language model (LLM) in the Japanese finance domain. Domain adaptation of language models, including LLMs, is receiving more attention as language models become more popular. This study demonstrates the effectiveness of domain adaptation through instruction tuning. To achieve this, we propose an instruction tuning data in Japanese called JaFIn, the Japanese Financial Instruction Dataset. JaFIn is manually constructed based on multiple data sources, including Japanese government websites, which provide extensive financial knowledge. We then utilize JaFIn to apply instruction tuning for several LLMs, demonstrating that our models specialized in finance have better domain adaptability than the original models. The financial-specialized LLMs created were evaluated using a quantitative Japanese financial benchmark and qualitative response comparisons, showing improved performance over the originals.

Masahiro Suzuki, Hiroki Sakaji

We aim to enhance a price sentiment index and to more precisely understand price trends from the perspective of not only consumers but also businesses. We extract comments related to prices from the Economy Watchers Survey conducted by the Cabinet Office of Japan and classify price trends using a large language model (LLM). We classify whether the survey sample reflects the perspective of consumers or businesses, and whether the comments pertain to goods or services by utilizing information on the fields of comments and the industries of respondents included in the Economy Watchers Survey. From these classified price-related comments, we construct price sentiment indices not only for a general purpose but also for more specific objectives by combining perspectives on consumers and prices, as well as goods and services. It becomes possible to achieve a more accurate classification of price directions by employing a LLM for classification. Furthermore, integrating the outputs of multiple LLMs suggests the potential for the better performance of the classification. The use of more accurately classified comments allows for the construction of an index with a higher correlation to existing indices than previous studies. We demonstrate that the correlation of the price index for consumers, which has a larger sample size, is further enhanced by selecting comments for aggregation based on the industry of the survey respondents.

Eriko Shigetsugu, Hiroki Sakaji, Itsuki Noda

In this paper, we design indices of economic fluctuation narratives derived from economic surveys. Companies, governments, and investors rely on key metrics like GDP and industrial production indices to predict economic trends. However, they have yet to effectively leverage the wealth of information contained in economic text, such as causal relationships, in their economic forecasting. Therefore, we design indices of economic fluctuation from economic surveys by using our previously proposed narrative framework. From the evaluation results, it is observed that the proposed indices had a stronger correlation with cumulative lagging diffusion index than other types of diffusion indices.

Yuta Ojima, Hiroki Sakaji, Tadashi Nakamura et al.

This paper presents a knowledge management system for automobile failure analysis using retrieval-augmented generation (RAG) with large language models (LLMs) and knowledge graphs (KGs). In the automotive industry, there is a growing demand for knowledge transfer of failure analysis from experienced engineers to young engineers. However, failure events are phenomena that occur in a chain reaction, making them difficult for beginners to analyze them. While knowledge graphs, which can describe semantic relationships and structure information is effective in representing failure events, due to their capability of representing the relationships between components, there is much information in KGs, so it is challenging for young engineers to extract and understand sub-graphs from the KG. On the other hand, there is increasing interest in the use of Graph RAG, a type of RAG that combines LLMs and KGs for knowledge management. However, when using the current Graph RAG framework with an existing knowledge graph for automobile failures, several issues arise because it is difficult to generate executable queries for a knowledge graph database which is not constructed by LLMs. To address this, we focused on optimizing the Graph RAG pipeline for existing knowledge graphs. Using an original Q&A dataset, the ROUGE F1 score of the sentences generated by the proposed method showed an average improvement of 157.6% compared to the current method. This highlights the effectiveness of the proposed method for automobile failure analysis.

Kota Tanabe, Masanori Hirano, Kazuki Matoya et al.

The domain adaptation of language models, including large language models (LLMs), has become increasingly important as the use of such models continues to expand. This study demonstrates the effectiveness of Composition to Augment Language Models (CALM) in adapting to the financial domain. CALM is a model to extend the capabilities of existing models by introducing cross-attention between two LLMs with different functions. In our experiments, we developed a CALM to enhance the financial performance of an LLM with strong response capabilities by leveraging a financial-specialized LLM. Notably, the CALM was trained using a financial dataset different from the one used to train the financial-specialized LLM, confirming CALM's ability to adapt to various datasets. The models were evaluated through quantitative Japanese financial benchmarks and qualitative response comparisons, demonstrating that CALM enables superior responses with higher scores than the original models and baselines. Additionally, comparative experiments on connection points revealed that connecting the middle layers of the models is most effective in facilitating adaptation to the financial domain. These findings confirm that CALM is a practical approach for adapting LLMs to the financial domain.

Meiyun Wang, Kiyoshi Izumi, Hiroki Sakaji

Recently, Large Language Models (LLMs) have attracted significant attention for their exceptional performance across a broad range of tasks, particularly in text analysis. However, the finance sector presents a distinct challenge due to its dependence on time-series data for complex forecasting tasks. In this study, we introduce a novel framework called LLMFactor, which employs Sequential Knowledge-Guided Prompting (SKGP) to identify factors that influence stock movements using LLMs. Unlike previous methods that relied on keyphrases or sentiment analysis, this approach focuses on extracting factors more directly related to stock market dynamics, providing clear explanations for complex temporal changes. Our framework directs the LLMs to create background knowledge through a fill-in-the-blank strategy and then discerns potential factors affecting stock prices from related news. Guided by background knowledge and identified factors, we leverage historical stock prices in textual format to predict stock movement. An extensive evaluation of the LLMFactor framework across four benchmark datasets from both the U.S. and Chinese stock markets demonstrates its superiority over existing state-of-the-art methods and its effectiveness in financial time-series forecasting.

Masanori Hirano, Masahiro Suzuki, Hiroki Sakaji

This study constructed a Japanese chat dataset for tuning large language models (LLMs), which consist of about 8.4 million records. Recently, LLMs have been developed and gaining popularity. However, high-performing LLMs are usually mainly for English. There are two ways to support languages other than English by those LLMs: constructing LLMs from scratch or tuning existing models. However, in both ways, datasets are necessary parts. In this study, we focused on supporting Japanese in those LLMs and making a dataset for training or tuning LLMs in Japanese. The dataset we constructed consisted of various tasks, such as translation and knowledge tasks. In our experiment, we tuned an existing LLM using our dataset and evaluated the performance qualitatively. The results suggest that our dataset is possibly beneficial for LLMs. However, we also revealed some difficulties in constructing LLMs in languages other than English.