129,730 Hits in 4.2 sec

Sequence-Level Self-Learning with Multiple Hypotheses

Kenichi Kumatani, Dimitrios Dimitriadis, Yashesh Gaur, Robert Gmyr, Sefik Emre Eskimez, Jinyu Li, Michael Zeng
2020 Interspeech 2020  
In this work, we develop new self-learning techniques with an attention-based sequence-to-sequence (seq2seq) model for automatic speech recognition (ASR).  ...  The seq2seq network is updated through the MTL framework so as to find the common representation that can cover multiple hypotheses.  ...  Proposed self-learning method Loss function approximation with multiple hypotheses In many cases, the best ASR output contains errors.  ... 
doi:10.21437/interspeech.2020-2020 dblp:conf/interspeech/KumataniDGGELZ20 fatcat:ok6cdndf2jbubh4jhpqfjqesei

Learning from Observing: Vision and POIROT - Using Metareasoning for Self Adaptation

Mark Burstein, Robert Bobrow, William Ferguson, Robert Laddaga, Paul Robertson
2010 2010 Fourth IEEE International Conference on Self-Adaptive and Self-Organizing Systems Workshop  
We present a cognitive architecture that heavily utilizes metareasoning for self adaptation,.  ...  We also discuss how this architecture is applied in the POIROT system, which learns web services workflow from "observing" a small number of expert examples.  ...  with respect to additional examples, and to self-evaluate the quality of its component's outputs.  ... 
doi:10.1109/sasow.2010.61 dblp:conf/saso/BursteinBFLR10 fatcat:s2hgkt7wxfgzteqx274al5nvsq

Poetry to Prose Conversion in Sanskrit as a Linearisation Task: A Case for Low-Resource Languages

Amrith Krishna, Vishnu Sharma, Bishal Santra, Aishik Chakraborty, Pavankumar Satuluri, Pawan Goyal
2019 Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics  
The first pretraining step learns task specific token embeddings from pretrained embeddings. In the next step, we generate multiple hypotheses for possible word arrangements of the input .  ...  The word ordering in a Sanskrit verse is often not aligned with its corresponding prose order. Conversion of the verse to its corresponding prose helps in better comprehension of the construction.  ...  Pretraining Step 2 -Self-Attention Based Word-Ordering (SAWO): SAWO allows us to generate multiple permutations of words as hypotheses, which can be used as input to a seq2seq model.  ... 
doi:10.18653/v1/p19-1111 dblp:conf/acl/KrishnaSSCSG19 fatcat:vfyr3thuzndipjjgtbe5mjqp5e

MHFormer: Multi-Hypothesis Transformer for 3D Human Pose Estimation [article]

Wenhao Li, Hong Liu, Hao Tang, Pichao Wang, Luc Van Gool
2022 arXiv   pre-print
representations; (ii) Model self-hypothesis communication, merge multiple hypotheses into a single converged representation and then partition it into several diverged hypotheses; (iii) Learn cross-hypothesis  ...  To relieve this limitation, we propose a Multi-Hypothesis Transformer (MHFormer) that learns spatio-temporal representations of multiple plausible pose hypotheses.  ...  Suppose there are M different hypotheses and L 1 layers in the MHG, it takes a sequence of 2D poses X∈R N ×J×2 with N video frames and J body joints as input and outputs multiple hypotheses [X 1 L1 , X  ... 
arXiv:2111.12707v4 fatcat:cvhsefnvj5clpaanwkdzcnsxj4

ASAPP-ASR: Multistream CNN and Self-Attentive SRU for SOTA Speech Recognition [article]

Jing Pan, Joshua Shapiro, Jeremy Wohlwend, Kyu J. Han, Tao Lei, Tao Ma
2020 arXiv   pre-print
In this paper we present state-of-the-art (SOTA) performance on the LibriSpeech corpus with two novel neural network architectures, a multistream CNN for acoustic modeling and a self-attentive simple recurrent  ...  In the hybrid ASR framework, the multistream CNN acoustic model processes an input of speech frames in multiple parallel pipelines where each stream has a unique dilation rate for diversity.  ...  All of our self-attentivce SRU language models are trained at the utterance level (i.e., the model does not leverage any context past sentence boundaries), with a maximum sequence length of 275 tokens.  ... 
arXiv:2005.10469v1 fatcat:2w2bphgbjfhzzinnnshf4knnba

ASAPP-ASR: Multistream CNN and Self-Attentive SRU for SOTA Speech Recognition

Jing Pan, Joshua Shapiro, Jeremy Wohlwend, Kyu J. Han, Tao Lei, Tao Ma
2020 Interspeech 2020  
We further improve the performance via N -best rescoring using a 24-layer self-attentive SRU language model, achieving WERs of 1.75% on test-clean and 4.46% on test-other.  ...  In the hybrid ASR framework, the multistream CNN acoustic model processes an input of speech frames in multiple parallel pipelines where each stream has a unique dilation rate for diversity.  ...  All of our self-attentivce SRU language models are trained at the utterance level (i.e., the model does not leverage any context past sentence boundaries), with a maximum sequence length of 275 tokens.  ... 
doi:10.21437/interspeech.2020-2947 dblp:conf/interspeech/PanSWH0M20 fatcat:pfvnsjbilfdzhmadi24doeql5e

Examining self-efficacy during learning: variability and relations to behavior, performance, and learning

Matthew L. Bernacki, Timothy J. Nokes-Malach, Vincent Aleven
2014 Metacognition and Learning  
Findings suggest that self-efficacy varies during learning, that students consider multiple aspects of performance to inform their efficacy judgments, and that changes in efficacy influence self-regulated  ...  Self-regulated learning (SRL) theorists propose that learners' motivations and cognitive and metacognitive processes interact dynamically during learning, yet researchers typically measure motivational  ...  Self efficacy as a dynamic component of SRL Because our research hypotheses are predicated on the assumption that learners' level of selfefficacy changes over a learning task, we first must confirm that  ... 
doi:10.1007/s11409-014-9127-x fatcat:hzhox4lpizevlmtq5n3rwedvly

The Influence of WebCT Information Technology and Structure of Instruction on Students Academic Performance

Konstantin Taskov
2007 Americas Conference on Information Systems  
The purpose of this research is to investigate the influence of WebCT Information Technology, students' perceived computer self-efficacy, students' motivation to learn and the degree of course structure  ...  Hiltz also found that levels of maturity, degree of effort, levels of academic ability and motivation all correlate positively with learning outcomes.  ...  V) Data Analysis : The author used multiple regression procedure to test the simultaneous influence of the hypothesized independent variables on students' academic outcomes and to validate the stated hypotheses  ... 
dblp:conf/amcis/Taskov07 fatcat:wzqjgda3wne5znhuamc5eqksdq


2002 Personnel Psychology  
In web-based training, for example, individuals can use hyper-links and menus to customize the material to which they attend, determine the sequence by which they learn, and control the amount of time  ...  Yet, today's technologically based training systems often provide individuals with significant control over their learning (Brown, 2001) .  ...  Consistent with the idea of sequencing individuals' learning process, trainees with a stronger foundation of basic skills and knowledge early in training displayed higher levels of strategic skills and  ... 
doi:10.1111/j.1744-6570.2002.tb00111.x fatcat:whrtvbviq5g5xmo7w3wzqz6j3y

Semi-Supervised Speech Recognition via Graph-based Temporal Classification [article]

Niko Moritz, Takaaki Hori, Jonathan Le Roux
2021 arXiv   pre-print
Semi-supervised learning has demonstrated promising results in automatic speech recognition (ASR) by self-training using a seed ASR model with pseudo-labels generated for unlabeled data.  ...  In this setup, GTC is used to learn not only a temporal alignment, similarly to CTC, but also a label alignment to obtain the optimal pseudo-label sequence from the weighted graph.  ...  Note that self-training with lattice-based supervision was also proposed in [22] using a hybrid ASR system and the LF-MMI objective in order to incorporate frame-level confidence scores and alternate  ... 
arXiv:2010.15653v2 fatcat:dqdntftdrzfrnh6upmc5yjsnz4

Capturing Sequences of Learners' Self-Regulatory Interactions With Instructional Material During Game-Based Learning Using Auto-Recurrence Quantification Analysis

Daryn A. Dever, Mary Jean Amon, Hana Vrzáková, Megan D. Wiedbusch, Elizabeth B. Cloude, Roger Azevedo
2022 Frontiers in Psychology  
learning technologies to scaffold self-regulation during game play.  ...  Through hierarchical modeling, analyses suggested that greater dwell times and learning gains were associated with more predictable sequences of interaction with instructional materials.  ...  For this third research question, we hypothesized that learners with more repetitive eye gaze sequences would be present in learners with restricted agency and related with higher learning gains.  ... 
doi:10.3389/fpsyg.2022.813677 pmid:35712220 pmcid:PMC9197103 fatcat:h6fs6mjazzebtmmvomwxk25htm

Instance-Aware Predictive Navigation in Multi-Agent Environments [article]

Jinkun Cao, Xin Wang, Trevor Darrell, Fisher Yu
2021 arXiv   pre-print
In this work, we aim to achieve efficient end-to-end learning of driving policies in dynamic multi-agent environments.  ...  We design a sequential action sampling strategy to better leverage predicted states on both scene-level and instance-level.  ...  MEP predicts the inter-agent events along with the visual structure as well as their uncertainty scores to enable multiple hypotheses forecasting.  ... 
arXiv:2101.05893v1 fatcat:tyogyo2dlbbadbzauqiq6t6g5y

Do Video Modeling and Metacognitive Prompts Improve Self-Regulated Scientific Inquiry?

Yoana Omarchevska, Andreas Lachner, Juliane Richter, Katharina Scheiter
2022 Educational Psychology Review  
Our findings show that video modeling examples are a promising instructional method for supporting inquiry learning on both the process and the learning outcomes level.  ...  Process mining revealed that in the VM conditions these processes occurred in unique sequences and that self-regulation processes had many self-loops.  ...  In line with Bannert et al. (2015) , we hypothesized that the VMP condition would outperform the VM condition (H3b).  ... 
doi:10.1007/s10648-021-09652-3 fatcat:panq42qkj5h4lilenszxh4t5ra

Progressive Joint Modeling in Unsupervised Single-Channel Overlapped Speech Recognition

Zhehuai Chen, Jasha Droppo, Jinyu Li, Wayne Xiong
2018 IEEE/ACM Transactions on Audio Speech and Language Processing  
The improvement comes from better model generalization, training efficiency and the sequence level linguistic knowledge integration.  ...  learning and a discriminative training criterion.  ...  ACKNOWLEDGMENT We thank Chris Basoglu, Frank Seide for their invaluable assistance with CNTK; Mike Seltzer, Takuya Yoshioka, Hakan Erdogan and Andreas Stolcke for many helpful conversations.  ... 
doi:10.1109/taslp.2017.2765834 fatcat:2e7p7bwqsvhk5beu6lei762ivi

Learning Representations for Predicting Future Activities [article]

Mohammadreza Zolfaghari, Özgün Çiçek, Syed Mohsin Ali, Farzaneh Mahdisoltani, Can Zhang, Thomas Brox
2019 arXiv   pre-print
In this work, we address future prediction at the abstract level of activities. We propose a network module for learning embeddings of the environment's dynamics in a self-supervised way.  ...  To take the ambiguities and high variances in the future activities into account, we use a multi-hypotheses scheme that can represent multiple futures.  ...  Therefore, we propose learning multiple hypotheses with their uncertainties, similar to multihypotheses networks (MHN) [16, 25, 6, 36] .  ... 
arXiv:1905.03578v1 fatcat:6kfyojpbkzedxjm2dz3pyxi3hi
« Previous Showing results 1 — 15 out of 129,730 results