Filters








46,848 Hits in 8.1 sec

Uncertainty in Neural Network Word Embedding: Exploration of Threshold for Similarity [article]

Navid Rekabsaz, Mihai Lupu, Allan Hanbury
2018 arXiv   pre-print
Word embedding, specially with its recent developments, promises a quantification of the similarity between terms.  ...  We first observe and quantify the uncertainty factor of the word embedding models regarding to the similarity value.  ...  Uncertainty of Similarity In this section we make a series of practical observations on word embeddings and the similarities computed based on them.  ... 
arXiv:1606.06086v2 fatcat:pgzs6wouh5cdtm56cchesjsrqu

A Financial Service Chatbot based on Deep Bidirectional Transformers [article]

Shi Yu, Yuxin Chen, Hussain Zaidi
2020 arXiv   pre-print
We investigated two uncertainty metrics, information entropy and variance of dropout sampling in BERT, followed by mixed-integer programming to optimize decision thresholds.  ...  The proposed approach can be useful for industries seeking similar in-house solutions in their specific business domains.  ...  We thank colleagues in Vanguard Retail Group (IT/Digital, Customer Care) for their pioneering effort collecting and curating all the data used in our approach.  ... 
arXiv:2003.04987v1 fatcat:pa6brq5avnb33hkvs74kpbdsgu

Explaining Financial Uncertainty through Specialized Word Embeddings

Christoph Kilian Theil, Sanja Štajner, Heiner Stuckenschmidt
2020 ACM/IMS Transactions on Data Science  
As a baseline, we use an existing dictionary of financial uncertainty triggers; furthermore, we retrieve related terms in specialized word embedding models to automatically expand this dictionary.  ...  To explore this field, we use term weighting methods to detect linguistic uncertainty in a large dataset of financial disclosures.  ...  ACKNOWLEDGMENTS We thank the anonymous reviewers for their helpful comments.  ... 
doi:10.1145/3343039 fatcat:noxcctneczf3noe56oxxjez5ti

AVA: A Financial Service Chatbot Based on Deep Bidirectional Transformers

Shi Yu, Yuxin Chen, Hussain Zaidi
2021 Frontiers in Applied Mathematics and Statistics  
We investigated two uncertainty metrics, information entropy and variance of dropout sampling, in BERT, followed by mixed-integer programming to optimize decision thresholds.  ...  The proposed approach can be useful for industries seeking similar in-house solutions in their specific business domains.  ...  We use mixed-integer optimization to find a threshold for human escalation of a user query based on the mean prediction and the uncertainty of the prediction.  ... 
doi:10.3389/fams.2021.604842 fatcat:2c3olripsndttkiuwasgutivza

Exploring Confidence Measures for Word Spotting in Heterogeneous Datasets [article]

Fabian Wolf, Philipp Oberdiek, Gernot A. Fink
2019 arXiv   pre-print
In this paper, we explore different metrics for quantifying the confidence of a CNN in its predictions, specifically on the retrieval problem of word spotting.  ...  We investigate four different approaches that are either based on the network's attribute estimations or make use of a surrogate model.  ...  INTRODUCTION Word spotting is a powerful tool for exploring handwritten document collections.  ... 
arXiv:1903.10930v1 fatcat:3fmpjfx4wbh5dkyztuyitflxhq

Clustering Chinese Product Features with Multilevel Similarity [chapter]

Yu He, Jiaying Song, Yuzhuang Nan, Guohong Fu
2015 Lecture Notes in Computer Science  
To handle different levels of connections between co-referred product features, we consider three similarity measures, namely the literal similarity, the word embedding-based semantic similarity and the  ...  This paper presents an unsupervised hierarchical clustering approach for grouping co-referred features in Chinese product reviews.  ...  To approach this, we explore three levels of similarities, namely the literal similarity, the semantic similarity based on word embeddings and the contextual similarity based on explanatory evaluations  ... 
doi:10.1007/978-3-319-25816-4_28 fatcat:dtxb4q5xmbc4vokvk2o6ixve3u

Word2Box: Learning Word Representation Using Box Embeddings [article]

Shib Sankar Dasgupta, Michael Boratko, Shriya Atmakuri, Xiang Lorraine Li, Dhruvesh Patel, Andrew McCallum
2021 arXiv   pre-print
Learning vector representations for words is one of the most fundamental topics in NLP, capable of capturing syntactic and semantic relationships useful in a variety of downstream NLP tasks.  ...  We demonstrate improved performance on various word similarity tasks, particularly on less common words, and perform a qualitative analysis exploring the additional unique expressivity provided by Word2Box  ...  In this work, we introduce WORD2BOX, a region-based embedding for words where each word is represented by an n-dimensional hyperrectangle or "box".  ... 
arXiv:2106.14361v1 fatcat:lzpcx7qgzrhqxf2plmin6jgray

Multimodal Word Distributions [article]

Ben Athiwaratkun, Andrew Gordon Wilson
2019 arXiv   pre-print
Word embeddings provide point representations of words containing useful semantic information.  ...  We introduce multimodal word distributions formed from Gaussian mixtures, for multiple word meanings, entailment, and rich uncertainty information.  ...  Acknowledgements We thank NSF IIS-1563887 for support.  ... 
arXiv:1704.08424v2 fatcat:xtuohjrfyrhq7nzsgrriztz4hm

Multimodal Word Distributions

Ben Athiwaratkun, Andrew Wilson
2017 Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)  
Word embeddings provide point representations of words containing useful semantic information.  ...  We introduce multimodal word distributions formed from Gaussian mixtures, for multiple word meanings, entailment, and rich uncertainty information.  ...  Acknowledgements We thank NSF IIS-1563887 for support.  ... 
doi:10.18653/v1/p17-1151 dblp:conf/acl/AthiwaratkunW17 fatcat:42me3nofqra35bab5pluqqoatu

Fluent: An AI Augmented Writing Tool for People who Stutter [article]

Bhavya Ghai, Klaus Mueller
2021 arXiv   pre-print
Fluent embodies a novel active learning based method of identifying words an individual might struggle pronouncing. Such words are highlighted in the interface.  ...  On hovering over any such word, Fluent presents a set of alternative words which have similar meaning but are easier to speak. The user is free to accept or ignore these suggestions.  ...  Phonetic embeddings map each word to its corresponding vector representation based on the constituting phonemes. Words with similar pronunciation will be closer to each other in the embedding space.  ... 
arXiv:2108.09918v1 fatcat:4dfrbmkphnhw5ermsmn4zkrs3a

Detecting Emerging Symptoms of COVID-19 using Context-based Twitter Embeddings [article]

Roshan Santosh, H. Andrew Schwartz, Johannes C. Eichstaedt, Lyle H. Ungar, Sharath C. Guntuku
2020 arXiv   pre-print
In this paper, we present an iterative graph-based approach for the detection of symptoms of COVID-19, the pathology of which seems to be evolving.  ...  More generally, the method can be applied to finding context-specific words and texts (e.g. symptom mentions) in large imbalanced corpora (e.g. all tweets mentioning #COVID-19).  ...  If all words for given depth are explored, the top m words corresponding to that depth are selected based on similarity to CEmb.  ... 
arXiv:2011.03983v1 fatcat:dg6dxblwyvagxhxdmolw4eidoy

Out-of-Distribution Detection using Multiple Semantic Label Representations [article]

Gabi Shalev, Yossi Adi, Joseph Keshet
2019 arXiv   pre-print
Deep Neural Networks are powerful models that attained remarkable results on a variety of tasks.  ...  However, it is not clear how a network will act when it is fed with an out-of-distribution example. In this work, we consider the problem of out-of-distribution detection in neural networks.  ...  Our model is based on word embedding representations.  ... 
arXiv:1808.06664v3 fatcat:zwtomwj54vbsfifzdljb7ph67i

Simultaneous Learning of Pivots and Representations for Cross-Domain Sentiment Classification

Liang Li, Weirui Ye, Mingsheng Long, Yateng Tang, Jin Xu, Jianmin Wang
2020 PROCEEDINGS OF THE THIRTIETH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE AND THE TWENTY-EIGHTH INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE  
A series of approaches depend on the pivot features that behave similarly for polarity prediction in both domains.  ...  Cross-domain sentiment classification aims to leverage useful knowledge from a source domain to mitigate the supervision sparsity in a target domain.  ...  Acknowledgments This work was supported by the National Key R&D Program of China (2017YFC1502003) and Natural Science Foundation of China (61772299 and 71690231).  ... 
doi:10.1609/aaai.v34i05.6336 fatcat:3wbotv24w5gs5oeucrowtbspju

Frequency discrimination in budgerigars (Melopsittacus undulatus): Effects of tone duration and tonal context

Micheal L. Dent, Robert J. Dooling, Alisa S. Pierce
2000 Journal of the Acoustical Society of America  
FDLs in budgerigars for 20-ms tones embedded in a sequence of six other tones were similar to FDLs measured for tones of the same frequency presented in isolation.  ...  Moreover, there was no effect of introducing trial-by-trial variation in the location of the frequency change in the seven-tone complexes for budgerigars, a condition for which humans showed a large decrement  ...  In experiment 3, we explored the effects of a surrounding tonal context on discrimination of frequency change in a ͑24-ms͒ tone burst.  ... 
doi:10.1121/1.428651 pmid:10830387 fatcat:lr3hntbnmzccfegvegr2fhx46m

Analyzing the Role of Model Uncertainty for Electronic Health Records [article]

Michael W. Dusenberry, Dustin Tran, Edward Choi, Jonas Kemp, Jeremy Nixon, Ghassen Jerfel, Katherine Heller, Andrew M. Dai
2019 arXiv   pre-print
In light of this, we investigate the role of model uncertainty methods in the medical domain.  ...  Meanwhile, the presence of significant variability in patient-specific predictions and optimal decisions motivates the need for capturing model uncertainty.  ...  For this analysis, we focus on the free-text clinical notes found in the EHR. For each word in the notes vocabulary, we have an associated embeddings distribution formulated as a multivariate Normal.  ... 
arXiv:1906.03842v2 fatcat:qlaxzgwl6raitbgqt7l6jhg4rm
« Previous Showing results 1 — 15 out of 46,848 results