One of the challenges in text generation is to control text generation as intended by the user.Previous studies proposed specifying the keywords that should be included in the generated text.However, this approach is insufficient to generate text that reflect the user's intent.For example, placing an important keyword at the beginning of the text would help attract the reader's attention; however, existing methods do not enable such flexible control.In this paper, we tackle a novel task of controlling not only keywords but also the position of each keyword in the text generation.To this end, we propose a task-independent method that uses special tokens to control the relative position of keywords.Experimental results on summarization and story generation tasks show that the proposed method can control keywords and their positions.The experimental results also demonstrate that controlling the keyword positions can generate summary texts that are closer to the user's intent than baseline.

We propose a fundamental theory on ensemble learning that evaluates a given ensemble system by a well-grounded set of metrics. Previous studies used a variant of Fano’s inequality of information theory and derived a lower bound of the classification error rate on the basis of the accuracy and diversity of models. We revisit the original Fano’s inequality and argue that the studies did not take into account the information lost when multiple model predictions are combined into a final prediction. To address this issue, we generalize the previous theory to incorporate the information loss. Further, we empirically validate and demonstrate the proposed theory through extensive experiments on actual systems. The theory reveals the strengths and weaknesses of systems on each metric, which will push the theoretical understanding of ensemble learning and give us insights into designing systems.

Abstract Mining an argument structure from text is an important step for tasks such as argument search and summarization. While studies on argument(ation) mining have proposed promising neural network models, they usually suffer from a shortage of training data. To address this issue, we expand the training data with various auxiliary argument mining corpora and propose an end-to-end cross-corpus training method called Multi-Task Argument Mining (MT-AM). To evaluate our approach, we conducted experiments for the main argument mining tasks on several well-established argument mining corpora. The results demonstrate that MT-AM generally outperformed the models trained on a single corpus. Also, the smaller the target corpus was, the better the MT-AM performed. Our extensive analyses suggest that the improvement of MT-AM depends on several factors of transferability among auxiliary and target corpora.

This demonstration paper presents i-Parser, a novel development kit that produces high-performance semantic parsers. i-Parser converts training graphs into sequences written in a context-free language, then our proposed model learns to generate the sequences. With interactive configuration and visualization, users can easily build their own parsers. Benchmark results of i-Parser showed high performances of various parsing tasks in natural language processing.

This paper describes the first report on crosslingual transfer for semantic dependency parsing.We present the insight that there are two different kinds of cross-linguality, namely surface level and semantic level, and try to capture both kinds of cross-linguality by combining annotation projection and model transfer of pre-trained language models.Our experiments showed that the performance of our graph-based semantic dependency parser almost achieved the approximated upper bound.

State-of-the-art argument mining studies have advanced the techniques for predicting argument structures. However, the technology for capturing non-tree-structured arguments is still in its infancy. In this paper, we focus on non-tree argument mining with a neural model. We jointly predict proposition types and edges between propositions. Our proposed model incorporates (i) task-specific parameterization (TSP) that effectively encodes a sequence of propositions and (ii) a proposition-level biaffine attention (PLBA) that can predict a non-tree argument consisting of edges. Experimental results show that both TSP and PLBA boost edge prediction performance compared to baselines.

For the motion planning problem, an iterative algorithm has been proposed as a general method. However, when we apply the algorithm to the problem of under-actuated mechanical systems, there are many cases when the convergence performance becomes too poor and we can not obtain the solution. We investigate the reason why the performance becomes poor and propose a method to improve the convergence performance. The effectiveness of the proposed method is shown by using an index that expresses the convergence performance.

A feedback gain tuning method is proposed to control mechatronic systems which have the characteristic applicable to iterative trial movements. The tuning algorithm is composed of an optimization procedure based on the complex method and has the following features: (1) it does not require any dynamic model of controlled systems; (2) it tunes the feedback gain under an arbitrary performance index and is applicable to both linear and nonlinear control laws; (3) the algorithm is concise and has a few adjusting parameters; and (4) the tuning process proceeds well and the on-site tuning for actual systems is possible. It is also confirmed by the experiments using both simulations and actual systems that the proposed tuning method is effective to the stabilization control of an inverted pendulum.

A method is proposed which describes the joint trajectories of a manipulator with B-spline curves, and which optimizes the values of these control points using the complex method under collision-free, full dynamic and kinematic constraints. This method utilizes the B-spline curve's uniqueness, continuity, and local controllability to the control points, and incorporates the complex method which does not require evaluation of the gradient in the optimization. This method has the following advantages: (1) It is applicable to manipulators with arbitrary degrees of freedom; (2) It achieves the generation of the trajectory which takes the priority of specified constraints into consideration; (3) Weighting factors of the performance indices corresponding to the constraints are automatically computed. The method has been applied to the collision-free trajectory generation problems of a manipulator with three degrees of freedom and the results show its effectiveness.

A pattern recognition method using an artificial hand to identify the pattern of a grasped object and to locate its position relative to the hand coordinates is discussed. In the proposed recognition method, an object is recognized as a pile of two-dimensional slices, namely, a pile of plane closed curves. Each curve (i.e., contour) is described as the distribution pattern of unit-vectors, called the unit-vector distribution (UVD). The proposed recognition method using UVD functions is as follows. 1) Given the a priori information about the contours of objects to be grasped, the UVD's of their contours are calculated and memorized. These are called the original patterns (OP's). 2) Then a sampled UVD is formed from real tactile data and called the data pattern (DP). 3) The DP is compared with the memorized patterns (MP's), which are obtained by blurring the corresponding OP, so as to make it easier to single out the best fitted OP. 4) Finally, the contour is regenerated in the hand coordinate system by using the UVD of the singled out pattern. An experiment using a trial sensor and a model hand has shown the effectiveness of the proposed method in the recognition of several kinds of objects having cylindrical and prismatic contours.