Filters








760 Hits in 4.2 sec

Unsupervised Paraphrasing with Pretrained Language Models [article]

Tong Niu, Semih Yavuz, Yingbo Zhou, Nitish Shirish Keskar, Huan Wang, Caiming Xiong
2021 arXiv   pre-print
Our recipe consists of task-adaptation, self-supervision, and a novel decoding algorithm named Dynamic Blocking (DB).  ...  To address this drawback, we adopt a transfer learning approach and propose a training pipeline that enables pre-trained language models to generate high-quality paraphrases in an unsupervised setting.  ...  E Paraphrasing in German We pair BART directly with Dynamic Blocking to generate paraphrases in German.  ... 
arXiv:2010.12885v2 fatcat:wcronrkhx5cidasbpv7uvmmbdu

Learning to Selectively Learn for Weakly-supervised Paraphrase Generation [article]

Kaize Ding, Dingcheng Li, Alexander Hanbo Li, Xing Fan, Chenlei Guo, Yang Liu, Huan Liu
2021 arXiv   pre-print
Specifically, we tackle the weakly-supervised paraphrase generation problem by: (1) obtaining abundant weakly-labeled parallel sentences via retrieval-based pseudo paraphrase expansion; and (2) developing  ...  Though unsupervised endeavors have been proposed to address this issue, they may fail to generate meaningful paraphrases due to the lack of supervision signals.  ...  control (Huang and Chang, 2021), or dynamic blocking (Niu et al., 2020) .  ... 
arXiv:2109.12457v1 fatcat:6f7ikuvz7ncehdu3erbwpgk3ou

MultiGranCNN: An Architecture for General Matching of Text Chunks on Multiple Levels of Granularity

Wenpeng Yin, Hinrich Schütze
2015 Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)  
We present MultiGranCNN, a general deep learning architecture for matching text chunks.  ...  We demonstrate stateof-the-art performance of MultiGranCNN on clause coherence and paraphrase identification tasks.  ...  Dynamic 2D Pooling The match feature models generate an s 1 × s 2 matrix.  ... 
doi:10.3115/v1/p15-1007 dblp:conf/acl/YinS15 fatcat:d3jiup34jbghtnrntx4csxfwyy

A Hybrid Model for Paraphrase Detection Combines pros of Text Similarity with Deep Learning

Mohamed I., Wael H., Hawaf Abdalhakim
2019 International Journal of Computer Applications  
Paraphrase detection (PD) is a very essential and important task in Natural language processing.  ...  The goal of paraphrase detection is to check whether two statements written in natural language have the identical semantic or not.  ...  For the sentence embedding, according to their purposes, they generally fall into two categories: task-specific sentence embeddings and general-purpose sentence embeddings [26] .  ... 
doi:10.5120/ijca2019919011 fatcat:jjl3modl6vbn7kfuddibziantu

Noun2Verb: Probabilistic frame semantics for word class conversion [article]

Lei Yu, Yang Xu
2022 arXiv   pre-print
We evaluate an incremental set of probabilistic models that learn to interpret and generate novel denominal verb usages via paraphrasing.  ...  However, existing natural language processing systems are impoverished in interpreting and generating novel denominal verb usages.  ...  (listener) and generation (speaker) model simultaneously via generative modeling.  ... 
arXiv:2205.06321v1 fatcat:hk4uhpuqtberhkg2yakzyipibq

On the difficulty of a distributional semantics of spoken language [article]

Grzegorz Chrupała, Lieke Gelderloos, Ákos Kádár, Afra Alishahi
2018 arXiv   pre-print
We evaluate two simple unsupervised models which, to varying degrees of success, learn semantic representations of speech fragments.  ...  We conjecture that unsupervised learning of the semantics of spoken language becomes feasible if we abstract from the surface variability.  ...  Related work Studies of unsupervised learning from speech typically aim to discover the phonemic or lexical building blocks of the language signal.  ... 
arXiv:1803.08869v2 fatcat:qzq7vuiocfesdh4c3nqnv5zi3y

An Empirical Study on Learning and Improving the Search Objective for Unsupervised Paraphrasing [article]

Weikai Steven Lu
2022 arXiv   pre-print
Research in unsupervised text generation has been gaining attention over the years.  ...  In this dissertation, we address the research problem of smoothing the noise in the heuristic search objective by learning to model the search dynamics.  ...  Results of Unsupervised Paraphrase Generation Table 4 .1 presents the result of automatic evaluation for unsupervised paraphrase generation.  ... 
arXiv:2203.12106v1 fatcat:bpbusikrefdwnhbss2tt3rkbgq

Multi-Task Learning in Natural Language Processing: An Overview [article]

Shijie Chen, Yu Zhang, Qiang Yang
2021 arXiv   pre-print
We first review MTL architectures used in NLP tasks and categorize them into four classes, including the parallel architecture, hierarchical architecture, modular architecture, and generative adversarial  ...  [126] trains multi-role dialogue representations via unsupervised multi-task pre-training on reference prediction, word prediction, role prediction, and sentence generation.  ...  For the sentence simplification task, [37] uses paraphrase generation and entailment generation as two auxiliary tasks.  ... 
arXiv:2109.09138v1 fatcat:hlgzjykuvzczzmsgnl32w5qo5q

Certainty factor model in paraphrase detection

Senem Kumova Metin, Bahar Karaoğlan, Tarık Kışla, Katira Soleymanzadeh
2021 Pamukkale University Journal of Engineering Sciences  
A set of succeeding paraphrase detection features (generic and distance based features) is built by filtering and this set is used as evidences in CF model.  ...  The proposed CF model in paraphrase identification is realized by utilizing the renowned paraphrase corpus of Microsoft Research (MSRP) [2] that is stated to be a standard resource in paraphrase identification  ...  by dynamic pooling.  ... 
doi:10.5505/pajes.2020.75350 fatcat:aw5geqmftnfrbpl4nbdpq3nrgq

Unsupervised Text Segmentation via Deep Sentence Encoders: a first step towards a common framework for text-based segmentation, summarization and indexing of media content

Iacopo Ghinassi
2021 Zenodo  
Our method shows improvement over other unsupervised methods and it gives results that are competitive with supervised approaches without the need for any training data.  ...  dynamic programming [14, 26] .  ...  left block.  ... 
doi:10.5281/zenodo.4744398 fatcat:ropxpxkcynaqdnv5jswuv6j3ji

Detecting Hallucinated Content in Conditional Neural Sequence Generation [article]

Chunting Zhou, Graham Neubig, Jiatao Gu, Mona Diab, Paco Guzman, Luke Zettlemoyer, Marjan Ghazvininejad
2021 arXiv   pre-print
We also apply our method to word-level quality estimation for MT and show its effectiveness in both supervised and unsupervised settings.  ...  Neural sequence models can generate highly fluent sentences, but recent studies have also shown that they are also prone to hallucinate additional content not supported by the input.  ...  For MT, we create paraphrased targets via knowledge distillation (Kim and Rush, 2016) where we use the output from TranS2S conditioned on the source sentence in the bi-text corpus as the paraphrased  ... 
arXiv:2011.02593v3 fatcat:p4ldz52bhvgqnn2fpfntgqop5u

Joint Space Neural Probabilistic Language Model for Statistical Machine Translation [article]

Tsuyoshi Okita
2017 arXiv   pre-print
This paper uses Latent Dirichlet Allocation (LDA) [5, 46, 6, 45, 33] to obtain the genre ID via (unsupervised) document classification since our interest here is on the genre of sentences in testset.  ...  This is since the observations can be regarded as being generated from a dynamic mixture model [19] as in (1) , the Dirichlet priors on the rows have a shared parameter. 0 G G G u 0 d d 0 0 u k H α  ... 
arXiv:1301.3614v3 fatcat:gwxlwzqf6zajlhsyd5mb6qkr6u

Short Answer Grading Using String Similarity And Corpus-Based Similarity

Wael, Aly A.
2012 International Journal of Advanced Computer Science and Applications  
These systems work in a supervised manner where predefined patterns and scoring rules are generated.  ...  The best correlation values achieved using Character-based and term-based were 0.435 and 0.382 using N-gram and Block distance respectively.  ...  Needleman-Wunsch algorithm is an example of dynamic programming, and was the first application of dynamic programming to biological sequence comparison.  ... 
doi:10.14569/ijacsa.2012.031119 fatcat:xnxdemczlzbi3o3xf4hq7llbte

Generic resources are what you need: Style transfer tasks without task-specific parallel training data [article]

Huiyuan Lai, Antonio Toral, Malvina Nissim
2021 arXiv   pre-print
First, we strengthen the model's ability to rewrite by further pre-training BART on both an existing collection of generic paraphrases, as well as on synthetic pairs created using a general-purpose lexical  ...  We propose a novel approach to this task that leverages generic resources, and without using any task-specific parallel (source-target) data outperforms existing unsupervised approaches on the two most  ...  Ethics Statement All work that automatically generates and/or alters natural text could unfortunately be used maliciously.  ... 
arXiv:2109.04543v1 fatcat:dnks6ss4ljakflw2a6f6zh5cbi

Don't Settle for Average, Go for the Max: Fuzzy Sets and Max-Pooled Word Vectors [article]

Vitalii Zhelezniak, Aleksandar Savkov, April Shen, Francesco Moramarco, Jack Flann, Nils Y. Hammerla
2019 arXiv   pre-print
Finally, we propose DynaMax, a completely unsupervised and non-parametric similarity measure that dynamically extracts and max-pools good features depending on the sentence pair.  ...  We show that max-pooled word vectors are only a special case of fuzzy BoW and should be compared via fuzzy Jaccard index rather than cosine similarity.  ...  Finally, we propose DynaMax, a completely unsupervised and non-parametric similarity measure that dynamically extracts and max-pools good features depending on the sentence pair.  ... 
arXiv:1904.13264v1 fatcat:msehfx2psng4zgu66q5utxc5sa
« Previous Showing results 1 — 15 out of 760 results