著者

吉田 舜 北園 淳 小澤 誠一 菅原 貴弘 芳賀 達也

出版者

一般社団法人 電気学会

雑誌

電気学会論文誌. C (ISSN:03854221)

巻号頁・発行日

vol.136, no.3, pp.340-347, 2016

被引用文献数

3

---

In recent years, along with the popularization of SNS, the incidents, which are called flaming, that the number of negative comments surges are on the increase. This becomes a problem for companies because flamings hurt companies reputation. In order to minimalize the damage of reputation, we propose the method that detects flamings by estimating the sentiment polarities of SNS comments. Because of the unique SNS characteristics such as repetition of same comments, the polarities of words are sometimes wrongly estimated. To alleviate this problem, transfer learning is introduced. In this research, the sentiment polarities of words are trained in every domain. This will enable to extract the words that are domain-specific and dictate the polarity of comments. These words are occurred in retweets. Transfer learning is implemented to non-extracted words by averaging the occurrence probabilities in other domains. These processes keep the polarities of important words that dictate the polarity of comments and modify the wrongly estimated polarities of words. The experimental results show that the proposed method improves the performance of estimating the sentiment polarity of comments. Moreover, flamings can be detected without missing by monitoring time course of the number of negative comments.

著者

朴 正泌 小澤 誠一 阿部 重夫

出版者

一般社団法人 電気学会

雑誌

電気学会論文誌. C (ISSN:03854221)

巻号頁・発行日

vol.121, no.5, pp.899-905, 2001

---

A main problem with dynamical associative memories (DAMs) is that when memory patterns are stored, pseudo-memories (false fixed points and limit cycles) are also generated and they hinder proper association of input patterns. To overcome this problem, Hassoun proposed a heuristic method of reducing pseudo-memories. In this method, DAMs are constructed such that a zero vector called “ground state” as well as stored patterns is stabilized and sparsely activated states (sparse patterns) converge to the ground state. Such dynamical properties of neural networks can be described with linear inequalities, and connection weights of networks are obtained by solving these inequalities using the Ho-Kashyap algorithm. In this paper, we propose an extended Hassoun model in which network dynamics are modified such that dense patterns, mix-ture patterns and inhibition patterns are also converged to the ground state. In simulations, we compare association performance of this extended Hassoun model with conventional associative memory models, and demonstrate the usefulness of our proposed model as a dynamical associative memory.

著者

小澤 誠一 竹内 洋平 阿部 重夫

出版者

一般社団法人電子情報通信学会

雑誌

電子情報通信学会論文誌. D, 情報・システム (ISSN:18804535)

巻号頁・発行日

vol.93, no.6, pp.826-836, 2010-06-01

被引用文献数

2

---

本論文では,初期データにのみ教師情報が与えられる「準教師付き学習タスク」において,ストリーミングデータからオンラインで非線形な特徴を抽出できる追加学習型カーネル主成分分析(IKPCA)を提案する.提案するIKPCAでは,学習データが入力されるたびにカーネル主成分分析の固有値問題を更新し,それを解くことで固有ベクトルの更新を行う.特徴固有空間で一次独立なデータを選択して固有ベクトルを表現するため,データの一次独立性を判定する必要がなく,追加学習時に保持するデータ数が少なくなって学習が高速化される.ベンチマークデータを用いた評価実験において,主成分分析(PCA)と追加学習型主成分分析(IPCA),更にカーネル主成分分析(KPCA)と比較し,IKPCAで得られる特徴量の評価を行った.その結果,IKPCAによってバッチ学習のKPCAと同等の認識性能が得られ,安定した追加学習が行われることを示した.このことは,IKPCAとKPCAにおいて,固有ベクトルや固有値の一致度を調べた実験からも確認された.また,多くの評価データでPCAやIPCAよりも,認識性能の優れた特徴が得られることを示した.