| 山根 壮一 | 1451118: M, 2回目発表 | 中村 哲，杉本 謙二，戸田 智基，Sakriani Sakti，Graham Neubig |
|
title: An estimation method of voice timbre evaluation values for voice data retrieval system of singing voice synthesis system
abstract: Recently, singing voice synthesis system is used in creating music. Singing voice synthesis system can change voice timbre of synthesized singing voice freely to replace the voice data. But there are a lot of usable voice data, so it is difficult to retrieve these data which the users need. In our research, we consider a voice data retrieval system by intuitive index. And we suggest an automatically estimation method of voice timbre evaluation values. language of the presentation: Japanese 発表題目 : 歌声合成システム用音源データ検索のための声質評価値推定法 発表概要 : 近年，主に楽曲製作において，歌声合成システムが広く利用されている． 歌声合成システムは，歌手の歌声である音源データを入れ替えることによって，合成される歌声の声質を自由に変更する事ができるが， 使用可能な音源データは大量に存在するため，所望の声質を持った音源データを検索することが困難である． 本研究では，膨大な音源データを直感的な指標によって検索するシステムについて検討し， 検索指標となる声質に関する評価値の自動推定手法を提案する． | ||
| 三浦 明波 | 1451101: M, 2回目発表 | 中村 哲，松本 裕治，戸田 智基，Graham Neubig，Sakriani Sakti |
|
title: Improving Pivot Translation by Remembering the Pivot
abstract: In statistical machine translation, pivot translation allows for translation of language pairs with little or no parallel data by introducing a third language for which data exists. In particular, the triangulation method, which translates by combining source-pivot and pivot-target translation models into a source-target model is known for its high translation accuracy. However, after the conventional triangulation method, information of pivot phrases is forgotten, and not used in the translation process. In this research, we propose a novel approach to remember the pivot phrases in the triangulation stage, and use a pivot language model as an additional information source at translation time. Experimental results on the Europarl corpus showed gains of 0.4-1.2 BLEU points in all tested combinations of languages. language of the presentation: Japanese 発表題目: 中間言語モデルを用いたピボット翻訳の精度向上 発表概要: 統計的機械翻訳において，特定の言語対で十分な文量の対訳コーパスが得られない場合，中間言語を用いたピボット翻訳が有効な手法の一つである．複数のピボット翻訳手法が考案されている中でも，特に中間言語を介する2つの翻訳モデルを合成するテーブル合成手法で，高い翻訳精度を得られることが知られている．ところが，従来のテーブル合成手法では，フレーズ対応推定時に用いた中間言語の情報が 「忘却」 され，翻訳時には利用できない問題が発生する．本研究では，合成時に用いた中間言語の情報も 「記憶」 し，ピボットの言語モデルを追加の情報源として翻訳に利用する新たなテーブル合成手法を提案する．また，欧州議会議事録による多言語コーパスを用いた実験により，本手法で評価を行った全ての言語の組合せで従来手法よりも有意に高い翻訳精度が得られた． | ||
| 水上 雅博 | 1461010: D, 中間発表 | 中村 哲，松本 裕治，戸田 智基，Graham Neubig，Sakriani Sakti，吉野 幸一郎 |
|
title: Example-based dialogue modeling considering satisfaction and
individuality
abstract: We describe an example-based dialogue model(EBDM) with considering user satisfaction based on prediction models and collaborative filtering. In dialogue systems, dialogue modeling is one of the most important factors contributing to user satisfaction. Current EBDM relies for a example database which constructed by heuristic rules, it does not consider the effect of example for users expressly. Thus, user satisfaction is significantly decreased if relied example database was low quality or user feels bad for example database by user preference based on individuality. To avoid this decrease, we propose 2 example selection methods as a EBDM framework considering satisfaction and individuality for user. First one is the direct satisfaction prediction model, user satisfaction were directly predicted from only example information and independent on user. Second one is the adaptive satisfaction prediction model, user satisfaction were predicted based on collaborative filtering technique using user preference which estimated from user feedback. Experimental results showed that the proposed methods can estimate user satisfaction and adapt to user preference, improving user satisfaction score. language of the presentation: Japanese | ||
| 山田 祐実 | 1451117: M, 2回目発表 | 松本 裕治，中村 哲，新保 仁，進藤 裕之 |
|
title: Translation of Classical Japanese into Contemporary Japanese Using MT
abstract:We are trying to translate classical Japanese to contemporary Japanese using statistical machine translation. There are limited corpora in Classical Japanese although a plenty of corpora are needed in translation tasks in general. In this study, we examine the accuracy of translation by trying some corpora in different types to train language models, which have influence on literary style of output sentences. Furthermore, we tried translation using corpora in different period in order to examine additional errors. language of the presentation: Japanese | ||
| 早田 昂生 | 1451084: M, 2回目発表 | 加藤 博一，小笠原 司，Christian Sandor，山本 豪志朗 |
|
title: Sprint Training Support System connecting the ground reaction force arrows and the recorded training video
abstract:In the track and field, people who belong to sprint running field are taking "Ground Reaction Forces" into account , not only their form and running to improve their speed. "Ground Reaction Forces" is considered as one of the most important factors in sprint running. The forces are calculated from the sensors that are embedded to ground with each step. Some factors are related to the speed of the sprinter directly, such as force direction, magnitude and braking power. In order to train efficiently, these components are evaluated frequently. Therefore sprinters and coaches wanted a system that visualises the force data of sprinters on their training videos. In our research, we estimate the camera rotation from the recorded video and synthesize CG force animation with the data we have to realise this system. language of the presentation: Japanese 発表題目: 録画したトレーニング映像上への床反力可視化によるスプリントトレーニング支援 発表概要: トラック競技においてフォームや走行法だけでなく，地面反力という視点から走行を改善しようとする動きがある．地面反力は足が接地時に地面から受ける力で，その向きや時間，力の大きさが直接速度に関係する，トラック競技において最も重要な要素のひとつである．この地面反力を分析することでトレーニングをより効率的に行うことができる．そのため走行を撮影した動画上に地面反力のデータを可視化することが，トラック競技選手やそのコーチなどから求められている．本研究ではこれを実現するため，撮影された動画からカメラの回転を推定し，地面反力を可視化したCGアニメーションの実動画上へのスムーズな重畳を行う． | ||
| KAPLAN ORAL | 1451124: M, 2回目発表 | 加藤 博一，小笠原 司，Christian Sandor，山本 豪志朗 |
|
Title: Turbine : Post-Visualisation of Pedalling Force Effectiveness for Long Distance Cycling Abstract: The goal of this research is to develop a system which supports training in long distance cycling. Having an effective pedalling action through the whole session such as a stage is important for professional cyclists. Graphs of force information have been used extensively to analyse this effectiveness independently from video information. Yet, these graphs do not encourage intuitive understanding when it comes to technology-novice trainees. Therefore, we apply AR technology onto video images for visualising the force effectiveness information which uses a new area-based representation that we proposed. We implement the system and confirm the responses of our users. Language of the Presentation: English | ||
| YAP CHRISTOPHER MICHAEL | 1461022: D, 中間発表 | 山口 英，藤川 和利，加藤 博一，砂原 秀樹，門林 雄基 |
|
Title: Using the Player-Side Emergence Concept for Reconciling Ludo-Narrative Dissonance in Interactive Video Games Abstract: The field of Game Studies (which is akin to similarly-styled media studies fields such as Film/Cinematic Studies or Literature Studies) is an extremely new and developing field dedicated to the multi-disciplinary study of the emerging media of the interactive video game. In this study, we focus on the area of narrative in an interactive game in order to understand how stories function when a considerable portion authorial control is allocated to the game player via the interactive nature of games. In particular, we address the problem of Ludo-Narrative Dissonance (LND), in which there is a trade-off between the overall narrative cohesion of the Directed Narrative (the narrative written by the developers of the game) and the Personal Narrative (the narrative which results from the player’s actions and decisions). With LND, the more interactivity there is in a game, the less that the resulting Personal Narrative will resemble the original developer’s intended narrative. Conversely, less interactivity has a tendency to result in a more focused, consistent Personal Narrative at the expense of turning the game into a static experience. In order to help modern game developers navigate and reconcile this LND tradeoff, this research proposes a concept of Player-Side Emergence in Narrative Games which utilizes recent findings in Narrative Game Studies and observations/analyses of experimental Indie Games in order to propose a concept which seeks to preserve both interactivity and authorial intent from the game developer, with respect to the resulting Personal (and Collective) Narratives. Language of the Presentation: English | ||