Masafumi Hagiwara

Inductive reasoning involves abstracting general principles from specific instances. It primarily relies on rules derived from relationships and is minimally dependent on specific data about entities, such as people, places, or organizations.Pre-trained language models (PLM) tend to focus on learning statistical features within a corpus. Thus, when inductive reasoning is expressed in a natural language, PLMs face challenges in learning the logical relations behind the text. Recently, researchers have explored graph neural network (GNN) architectures that excel in inductive inference on knowledge graphs (KGs) for inductive link prediction tasks; however, their application to natural language remains limited. To address the natural language inductive reasoning tasks, we propose a framework that utilizes a specific GNN module specialized in inductive link prediction as a reasoning mechanism. To construct inputs inferable to GNN from natural language, we first apply insights from a study of relation extraction tasks and use PLM to obtain embeddings for edge-related inferences. Subsequently, the newly designed module performs edge scoring and initializes the relation embeddings. The scores are used to prune the edges and are learned through edge weighting within the decoder. Experimental results on text datasets requiring logical inductive reasoning demonstrate that the proposed method notably improves PLM performance, outperforming the baselines. Furthermore, robustness evaluation on subsets provided by CLUTRR shows that our model surpasses other relational reasoning-based models in its ability to learn from and generalize noisy data.

In this paper, we propose a novel support system for generating comic videos from text. The comic video in this paper refers to the animation in the form of comic appearing on the screen one frame at a time in sequence. In the proposed system, comic is first generated from text, then each frame is extracted. Next, video animation is generated by adding motion to each frame. Polarity estimation of emotions is performed on the input text, and the results are used to control the motion of each frame of the comic and the background color of the video. The addition of movement and background color contributes greater clarity and impact, and makes it possible to generate impressive videos. In addition, the display time of the frame is appropriately controlled according to the number of characters in the speech balloon of each frame. Therefore, viewers can comfortably watch the comic video. Evaluation experiments have shown the validity of the proposed system and the usefulness of the generated comic videos.

The authors propose analogical reasoning systems based on first-order predicate logic using deep learning. The proposed systems consist of a model combining recursive neural networks and Word2Vec. When unknown data is input in this trained model, similar outputs to those of the trained data are obtained. Previous studies on symbolic inference using deep learning have focused on deductive and inductive inferences, and they require rule templates that are manually created and provided to the models in advance. However, it is not enough to infer unknown rules from a large amount of data on the Internet. Analogical reasoning is a mode of inference that attempts to solve a new case based on similarity. Thus, it enables inference of rules that were challenging to comprehend in previous studies. As a result of the experiments, unknown rules can be efficiently inferred from known rules contained in the knowledge bases. Using the proposed systems, the authors can extract unknown rules from five of the eight datasets consisting of three types of knowledge bases described in Prolog. The proposed method is the first case of analogical reasoning based on the first-order predicate logic using deep learning. Furthermore, the models used in the proposed systems have shown high robustness when using Word2Vec. For this reason, it is suggested that the practical applications of the proposed systems enable efficient inferences from a large amount of data that include noisy data on the Internet.

Most characters in the Chinese and Japanese languages are ideographic compound characters composedof subcharacter elements arranged in a planar manner. Token-based and image-based subcharacter-level modelshave been proposed to leverage the subcharacter elements. However, on the one hand, the conventional tokenbasedsubcharacter-level models are blind to the planar structural information; on the other hand, the image-basedmodels are weak with respect to ideographic characters that share similar shapes but have different meanings.These characteristics motivate us to explore non-image-based methods to encode the planar structural informationof characters. In this paper, we propose and discuss a method to encode the planar structural information by learningembeddings of the categories of the structure types. Our proposed model adds the structure embeddings to theconventional subcharacter embeddings and position embeddings before they are input into the encoder. In this way,the model learns the planar structural and positional information and retains the uniqueness of each character. Weevaluated the method in a text classification task. In the experiment, the embeddings were encoded by a CNN encoder,and then the encoded vectors were input into an LSTM classifier to classify product reviews as positive or negative.We compared the proposed model with models that use only the subcharacter embeddings, the structure embeddingor the position embeddings as well as with the conventional models in previous works. The results show that addingboth structure embeddings and position embeddings leads to more rich and representative features and better fittingon the dataset. The proposed method results in at best 1:8% better recall, 0:63% better F-score, and 0:55% better accuracy on the testing datasets compared to previous methods.

Current image quality assessment (IQA) methods require the original images for evaluation. However, recently, IQA methods that use machine learning have been proposed. These methods learn the relationship between the distorted image and the image quality automatically. In this paper, we propose an IQA method based on deep learning that does not require a reference image. We show that a convolutional neural network with distortion prediction and fixed filters improves the IQA accuracy.

Moral Judgment is the task of judging whether the action invoked by the input sentence is good or bad from the viewpoint of the morality. In this paper, we propose a new moral judgment system using attention-based distributed representation and co-occurrence information. Compared to the previous methods, the proposed attention-based system can utilize better representation of the input sentence. In addition, the proposed system can take into account not only distributed representation but also co-occurrence information which is effective in this task. The proposed system can be divided into two parts. First one is a distributed representation learning network using attention-based Long Short-Term Memory. Second one is a moral judgment network using distributed representation and co-occurrence information. In the experiments, the proposed system significantly outperforms the other models. Furthermore, we confirmed that attention-based distributed representation and co-occurrence information contribute to the accuracy improvement.

The character vocabulary can be very large in non-alphabetic languages such as Chinese and Japanese, which makes neural network models huge to process such languages. We explored a model for sentiment classification that takes the embeddings of the radicals of the Chinese characters, i.e, hanzi of Chinese and kanji of Japanese. Our model is composed of a CNN word feature encoder and a bi-directional RNN document feature encoder. The results achieved are on par with the character embedding-based models, and close to the state-of-the-art word embedding-based models, with 90% smaller vocabulary, and at least 13% and 80% fewer parameters than the character embedding-based models and word embedding-based models respectively. The results suggest that the radical embedding-based approach is cost-effective for machine learning on Chinese and Japanese.