Filters








8,455 Hits in 4.9 sec

On Length Divergence Bias in Textual Matching Models [article]

Lan Jiang, Tianshu Lyu, Yankai Lin, Meng Chong, Xiaoyong Lyu, Dawei Yin
2022 arXiv   pre-print
In this work, we provide a new perspective to study this issue -- via the length divergence bias.  ...  Despite the remarkable success deep models have achieved in Textual Matching (TM) tasks, it still remains unclear whether they truly understand language or measure the semantic similarity of texts by exploiting  ...  In this work, we investigate the length divergence heuristic that textual matching models are prone to learn.  ... 
arXiv:2109.02431v3 fatcat:nfliyes3xveddputrkk5fw3lce

On Length Divergence Bias in Textual Matching Models

Lan Jiang, Tianshu Lyu, Yankai Lin, Meng Chong, Xiaoyong Lyu, Dawei Yin
2022 Findings of the Association for Computational Linguistics: ACL 2022   unpublished
In this work, we provide a new perspective to study this issuevia the length divergence bias.  ...  Despite the remarkable success deep models have achieved in Textual Matching (TM) tasks, it still remains unclear whether they truly understand language or measure the semantic similarity of texts by exploiting  ...  In this work, we investigate the length divergence heuristic that textual matching models are prone to learn.  ... 
doi:10.18653/v1/2022.findings-acl.330 fatcat:xhoajn55jjcuhc36y7h3hgmfva

Contextualizing Variation in Text Style Transfer Datasets [article]

Stephanie Schoch, Wanyu Du, Yangfeng Ji
2021 arXiv   pre-print
This understanding, however, is likely to have implications for selecting multiple data sources for model training.  ...  Based on our results, we propose a categorization of stylistic and dataset properties to consider when utilizing or comparing text style datasets.  ...  For example, machine translation models have been found to exhibit stylistic bias through reflecting demographically-biased training data (Hovy et al., 2020) .  ... 
arXiv:2108.07871v1 fatcat:qfs6w2zmmzdzbbdvc5bdycdnje

Can Shuffling Video Benefit Temporal Bias Problem: A Novel Training Framework for Temporal Grounding [article]

Jiachang Hao, Haifeng Sun, Pengfei Ren, Jingyu Wang, Qi Qi, Jianxin Liao
2022 arXiv   pre-print
These methods do not reason the target moment locations based on the visual-textual semantic alignment but over-rely on the temporal biases of queries in training sets.  ...  The cross-modal matching task leverages the content consistency between shuffled and original videos to force the grounding model to mine visual contents to semantically match queries.  ...  Temporal bias problem refers to that a method reasons the target moment locations not based on the visual-textual semantic matching but over-relies on the temporal biases of queries in training set [37  ... 
arXiv:2207.14698v1 fatcat:2srhrtljovby7ijznxgbvjpvwu

Adversarial Representation Learning for Text-to-Image Matching [article]

Nikolaos Sarafianos, Xiang Xu, Ioannis A. Kakadiaris
2019 arXiv   pre-print
In addition, we demonstrate that BERT, a publicly-available language model that extracts word embeddings, can successfully be applied in the text-to-image matching domain.  ...  The proposed approach achieves state-of-the-art cross-modal matching performance on four widely-used publicly-available datasets resulting in absolute improvements ranging from 2% to 5% in terms of rank  ...  Acknowledgments This work has been funded in part by the UH Hugh Roy and Lillie Cranz Cullen Endowment Fund.  ... 
arXiv:1908.10534v1 fatcat:vvi2onangjbo7h4bsjsozwhfay

Domain-Aware Service Recommendation for Service Composition

Bofei Xia, Yushun Fan, Cheng Wu, Keman Huang, Wei Tan, Jia Zhang, Bing Bai
2014 2014 IEEE International Conference on Web Services  
Therefore, how to mine matching patterns between services in relevant domains and compositions becomes crucial to service recommendation for composition.  ...  In this paper, a novel approach is proposed to offer domain-aware service recommendation. First,  ...  Second, on top of vKNN, a probabilistic matching model -Domain Router (DR) model -is designed to decompose a textual requirement to relevant domains.  ... 
doi:10.1109/icws.2014.69 dblp:conf/icws/XiaFWHTZB14 fatcat:cgshtgqn6rfd3avao3xgjwmj6m

The influence of firm's feedbacks on user-generated content's linguistic style matching–An explanation based on communication accommodation theory

Dewen Liu, Chenyiming Gong, Sikang Zhang, Yongbin Ma
2022 Frontiers in Psychology  
In addition, users' textual sentiment and the published text length moderate the impact of firms' feedbacks (e.g., frequency, text length) on LSM.  ...  Specifically, the users' textual sentiment valence increases the positive effect of firms' feedbacks frequency and weakens the negative effect of firms' feedbacks text length on LSM.  ...  According to model 1, the firm's feedbacks frequency and text length did not significantly affect the LSM, indicating that the firm's feedbacks model (model 1) alone did not have predictive power on users  ... 
doi:10.3389/fpsyg.2022.949968 fatcat:uwh6hooiebdcfdc54sk7je5ema

Synopses of Movie Narratives: a Video-Language Dataset for Story Understanding [article]

Yidan Sun, Qin Chao, Boyang Li
2022 arXiv   pre-print
We establish benchmarks on video-text retrieval and zero-shot alignment on movie summary videos. With SyMoN, we hope to lay the groundwork for progress in multimodal story understanding.  ...  Differing from most existing video-text datasets, SyMoN features large semantic gaps between the visual and the textual modalities due to the prevalence of reporting bias and mental state descriptions.  ...  Understanding Video-Text Divergence by Sequencing Videos As discussed earlier, SYMON are characterized by large gaps between the textual and visual modalities due to the reporting bias, or the tendency  ... 
arXiv:2203.05711v1 fatcat:y4lqpr6hjzhv5fo5vbfaootnrq

How the structure of Wikipedia articles influences user navigation

Daniel Lamprecht, Kristina Lerman, Denis Helic, Markus Strohmaier
2016 New Review of Hypermedia and Multimedia  
Wikipedia and then evaluate link selection models based on article structure and other potential influences in navigation, such as the generality of an article's topic.  ...  In free-form Wikipedia navigation, covering all Wikipedia usage scenarios, we find that click choices can be best modeled by a bias towards article structure, such as a tendency to click links located  ...  In fact, the order of best fits of the navigational influences for wikigames exactly matches the one for Wikipedia.  ... 
doi:10.1080/13614568.2016.1179798 pmid:28670171 pmcid:PMC5468769 fatcat:e5edgboi2rcfhmixinrr2l2qqe

The Elements of Temporal Sentence Grounding in Videos: A Survey and Future Directions [article]

Hao Zhang, Aixin Sun, Wei Jing, Joey Tianyi Zhou
2022 arXiv   pre-print
As the background, we present a common structure of functional components in TSGV, in a tutorial style: from feature extraction from raw video and language query, to answer prediction of the target moment  ...  Connecting computer vision and natural language, TSGV has drawn significant attention from researchers in both communities.  ...  [165] disentangles bias from TSGV model by adjusting losses to compensate for biases dynamically.  ... 
arXiv:2201.08071v1 fatcat:2k2if6dsyveinec2dmmujcmhkq

Visual Commonsense in Pretrained Unimodal and Multimodal Models [article]

Chenyu Zhang, Benjamin Van Durme, Zhuowan Li, Elias Stengel-Eskin
2022 arXiv   pre-print
Moreover, increasing model size does not enhance performance, suggesting that the key to visual commonsense lies in the data.  ...  Our results indicate that multimodal models better reconstruct attribute distributions, but are still subject to reporting bias.  ...  Let the length of N sm be m and the length of N lg be n.  ... 
arXiv:2205.01850v1 fatcat:anteyantxrfnneby7ug4gcxohi

Unsupervised, Efficient and Semantic Expertise Retrieval

Christophe Van Gysel, Maarten de Rijke, Marcel Worring
2016 Proceedings of the 25th International Conference on World Wide Web - WWW '16  
It yields a statistically significant improved ranking over vector space and generative models in most cases, matching the performance of supervised methods on various benchmarks.  ...  We exclusively employ textual evidence and avoid explicit feature engineering by learning distributed word representations in an unsupervised way.  ...  The document-centric language models (Model 2) perform slightly better than our method on two (out of six) benchmarks in terms of MAP and NDCG@100: (1) For the CERC 2007 assessment we match performance  ... 
doi:10.1145/2872427.2882974 dblp:conf/www/GyselRW16 fatcat:5iuangtpjngzreapd2oky5i5nq

Adversarial Attentive Multi-modal Embedding Learning for Image-Text Matching

Kaimin Wei, Zhibo Zhou
2020 IEEE Access  
Matching the image and text with deep models has been extensively studied in recent years.  ...  In this work, we propose a novel model named Adversarial Attentive Multi-modal Embedding Learning (AAMEL) for image-text matching.  ...  In our model, the generator consists of two components for multi-modal embedding generating, i.e., visual attention model and textual attention model.  ... 
doi:10.1109/access.2020.2996407 fatcat:ilgrbt4k2zebbe3nlf532c3jma

ICRC-HIT: A Deep Learning based Comment Sequence Labeling System for Answer Selection Challenge

Xiaoqiang Zhou, Baotian Hu, Jiaxin Lin, Yang xiang, Xiaolong Wang
2015 Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015)  
In this paper, we present a comment labeling system based on a deep learning strategy.  ...  Good class in this answer selection challenge.  ...  Acknowledgments This work was supported in part by the National N atural Science Foundation of China (61272383, 61 173075 and 61203378), the Strategic Emerging In dustry Development Special Funds of Shenzhen  ... 
doi:10.18653/v1/s15-2037 dblp:conf/semeval/ZhouHLXW15 fatcat:nuzetkcpmbfh3mwwd5ytr5kr54

Critical Search: A Procedure for Guided Reading in Large-Scale Textual Corpora

Jo Guldi
2018 Journal of Cultural Analytics  
Consider the case of searching for change in the language of property in the parliamentary debates.  ...  Displacement is a theme in modern Britain since the beginnings of the enclosure of traditional peasant commons in the middle ages, 2 Eric Stokes, The English Utilitarians and India (Oxford: Clarendon Press  ...  Statistical inspection of the results reveals the bias implicit in particular models.  ... 
doi:10.22148/16.030 fatcat:5huexof5lnfkzatzv42ygr6jve
« Previous Showing results 1 — 15 out of 8,455 results