Filters








1,939 Hits in 4.0 sec

Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems [article]

Tsung-Hsien Wen, Milica Gasic, Nikola Mrksic, Pei-Hao Su, David Vandyke, Steve Young
2015 arXiv   pre-print
The LSTM generator can learn from unaligned data by jointly optimising sentence planning and surface realisation using a simple cross entropy training criterion, and language variation can be easily achieved  ...  This paper presents a statistical language generator based on a semantically controlled Long Short-term Memory (LSTM) structure.  ...  It can learn from unaligned data by jointly optimising its sentence planning and surface realisation components using a simple cross entropy training criterion without any heuristics, and good quality  ... 
arXiv:1508.01745v2 fatcat:bon3kfeakzhvdbtpigkj45hqle

Semantically Conditioned LSTM-based Natural Language Generation for Spoken Dialogue Systems

Tsung-Hsien Wen, Milica Gasic, Nikola Mrkšić, Pei-Hao Su, David Vandyke, Steve Young
2015 Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing  
The LSTM generator can learn from unaligned data by jointly optimising sentence planning and surface realisation using a simple cross entropy training criterion, and language variation can be easily achieved  ...  This paper presents a statistical language generator based on a semantically controlled Long Short-term Memory (LSTM) structure.  ...  It can learn from unaligned data by jointly optimising its sentence planning and surface realisation components using a simple cross entropy training criterion without any heuristics, and good quality  ... 
doi:10.18653/v1/d15-1199 dblp:conf/emnlp/WenGMSVY15 fatcat:lvi5w4o7rnd2tfuaapmopuoray

Multi-domain Neural Network Language Generation for Spoken Dialogue Systems

Tsung-Hsien Wen, Milica Gašić, Nikola Mrkšić, Lina M. Rojas-Barahona, Pei-Hao Su, David Vandyke, Steve Young
2016 Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies  
In this procedure, a model is first trained on counterfeited data synthesised from an out-of-domain dataset, and then fine tuned on a small set of in-domain utterances with a discriminative objective function  ...  Corpus-based evaluation results show that the proposed procedure can achieve competitive performance in terms of BLEU score and slot error rate while significantly reducing the data needed to train generators  ...  Furthermore, Cuayhuitl et al. (2014) trained statistical surface realisers from unlabelled data by an automatic slot labelling technique.  ... 
doi:10.18653/v1/n16-1015 dblp:conf/naacl/WenGMRSVY16 fatcat:hqqujijvpjdzjcjglsg6pa6ena

Multi-domain Neural Network Language Generation for Spoken Dialogue Systems [article]

Tsung-Hsien Wen, Milica Gasic, Nikola Mrksic, Lina M. Rojas-Barahona, Pei-Hao Su, David Vandyke, Steve Young
2016 arXiv   pre-print
In this procedure, a model is first trained on counterfeited data synthesised from an out-of-domain dataset, and then fine tuned on a small set of in-domain utterances with a discriminative objective function  ...  Corpus-based evaluation results show that the proposed procedure can achieve competitive performance in terms of BLEU score and slot error rate while significantly reducing the data needed to train generators  ...  Furthermore, Cuayhuitl et al. (2014) trained statistical surface realisers from unlabelled data by an automatic slot labelling technique.  ... 
arXiv:1603.01232v1 fatcat:nz5r2fzckradbg3nu7huef4ncy

Point at the Triple: Generation of Text Summaries from Knowledge Base Triples

Pavlos Vougiouklis, Eddy Maddalena, Jonathon Hare, Elena Simperl
2020 The Journal of Artificial Intelligence Research  
Our approach is based on a pointer-generator network, which, in addition to generating regular words from a fixed target vocabulary, is able to verbalise triples in several ways.  ...  We undertake an automatic and a human evaluation on single and open-domain summaries generation tasks. Both show that our approach significantly outperforms other data-driven baselines.  ...  After an entity is annotated in the text, its realisation is replaced by a surface form tuple that consists from this realisation and the name of the entity.  ... 
doi:10.1613/jair.1.11694 fatcat:3ikrw3lzunhanlo6xumhbyv3pu

Underspecified Universal Dependency Structures as Inputs for Multilingual Surface Realisation

Simon Mille, Anja Belz, Bernd Bohnet, Leo Wanner
2018 Proceedings of the 11th International Conference on Natural Language Generation  
Background With the advent of large-scale treebanks and statistical NLG, surface realisation research turned to the use of treebank annotations, processed in various ways, as inputs to surface realisation  ...  with a surface realiser as the second component.  ... 
doi:10.18653/v1/w18-6527 dblp:conf/inlg/MilleBBW18 fatcat:ceooc6ifyrarnl2sgr4kcxnc3y

Optimization Learning: Perspective, Method, and Applications

Risheng Liu
2020 Proceedings of the Twenty-Ninth International Joint Conference on Artificial Intelligence  
Numerous tasks at the core of statistics, learning, and vision areas are specific cases of ill-posed inverse problems.  ...  We move beyond these limits and propose a theoretically guaranteed optimization learning paradigm, a generic and provable paradigm for nonconvex inverse problems, and develop a series of convergent deep  ...  were realised in the text with their most frequent surface form.  ... 
doi:10.24963/ijcai.2020/711 dblp:conf/ijcai/VougiouklisMHS20 fatcat:diq6cuqkxrfstge5i3g7ydthb4

Stochastic Language Generation in Dialogue using Recurrent Neural Networks with Convolutional Sentence Reranking

Tsung-Hsien Wen, Milica Gasic, Dongho Kim, Nikola Mrksic, Pei-Hao Su, David Vandyke, Steve Young
2015 Proceedings of the 16th Annual Meeting of the Special Interest Group on Discourse and Dialogue  
The natural language generation (NLG) component of a spoken dialogue system (SDS) usually needs a substantial amount of handcrafting or a well-labeled dataset to be trained on.  ...  This paper presents a statistical language generator based on a joint recurrent and convolutional neural network structure which can be trained on dialogue act-utterance pairs without any semantic alignments  ...  To form a training corpus, dialogues from a set of 3577 dialogues collected in a user trial of a statistical dialogue manager proposed by Young et al. (2013) were randomly sampled and shown to workers  ... 
doi:10.18653/v1/w15-4639 dblp:conf/sigdial/WenGKMSVY15 fatcat:yals6sftofcqdktpwijkyzfzxe

Stochastic Language Generation in Dialogue using Recurrent Neural Networks with Convolutional Sentence Reranking [article]

Tsung-Hsien Wen, Milica Gasic, Dongho Kim, Nikola Mrksic, Pei-Hao Su, David Vandyke, Steve Young
2015 arXiv   pre-print
The natural language generation (NLG) component of a spoken dialogue system (SDS) usually needs a substantial amount of handcrafting or a well-labeled dataset to be trained on.  ...  This paper presents a statistical language generator based on a joint recurrent and convolutional neural network structure which can be trained on dialogue act-utterance pairs without any semantic alignments  ...  To form a training corpus, dialogues from a set of 3577 dialogues collected in a user trial of a statistical dialogue manager proposed by Young et al. (2013) were randomly sampled and shown to workers  ... 
arXiv:1508.01755v1 fatcat:7trc544s7bepfkbsxqmwbxy22m

A Domain Agnostic Approach to Verbalizing n-ary Events without Parallel Corpora

Bikash Gyawali, Claire Gardent, Christophe Cerisara
2015 Proceedings of the 15th European Workshop on Natural Language Generation (ENLG)  
We present a method for automatically generating descriptions of biological events encoded in the KB BIO 101 Knowledge base.  ...  We evaluate our approach on a corpus of 336 event descriptions, provide a qualitative and quantitative analysis of the results obtained and discuss possible directions for further work.  ...  Surface Realisation In our approach, surface realisation takes as input an even description.  ... 
doi:10.18653/v1/w15-4703 dblp:conf/enlg/GyawaliGC15 fatcat:naialintrjdrbkdb7yengtneaa

Stochastic Language Generation in Dialogue using Factored Language Models

François Mairesse, Steve Young
2014 Computational Linguistics  
Most previous work on trainable language generation has focused on two paradigms: (a) using a statistical model to rank a set of pre-generated utterances, or (b) using statistics to determine the generation  ...  Two data-driven methods for generating paraphrases in dialogue are presented: (a) by sampling from the Nbest list of realisations produced by BAGEL's FLM reranker; and (b) by learning a structured perceptron  ...  However, learning to map a single input to a set of surface realisations is a nontrivial machine learning problem.  ... 
doi:10.1162/coli_a_00199 fatcat:ehegb4qj3bgm3hguv3av6cvtce

DR 4.4: Natural Multimodal Interaction Final Pro- totype

Bernd Kiefer, Ivana Kruijff-Korbayova, Anna Welker, Rifca Peters, Sarah McLeod
2019 Zenodo  
self disclosure and a social talk part.  ...  For less experienced users, a graphical editor and compiler for hierarchical state machines has been implemented and is now ready for use.  ...  For the name slot (used to label a hotel name or airline name, etc.) and the ref anaphora slots performance is worse.  ... 
doi:10.5281/zenodo.3443669 fatcat:ez7jk76vmncshlzoivgqftx4ji

Crossing the Conversational Chasm: A Primer on Natural Language Processing for Multilingual Task-Oriented Dialogue Systems [article]

Evgeniia Razumovskaia, Goran Glavaš, Olga Majewska, Edoardo M. Ponti, Anna Korhonen, Ivan Vulić
2021 arXiv   pre-print
In task-oriented dialogue (ToD), a user holds a conversation with an artificial agent to complete a concrete task.  ...  We find that the most critical factor preventing the creation of truly multilingual ToD systems is the lack of datasets in most languages for both training and evaluation.  ...  Acknowledgments Evgeniia Razumovskaia is supported by a scholarship from Huawei.  ... 
arXiv:2104.08570v2 fatcat:bi5xizz4wzct5fpiuk3ikotjta

Evaluating the State-of-the-Art of End-to-End Natural Language Generation: The E2E NLG Challenge [article]

Ondřej Dušek, Jekaterina Novikova, Verena Rieser
2019 arXiv   pre-print
Introducing novel automatic and human metrics, we compare 62 systems submitted by 17 institutions, covering a wide range of approaches, including machine learning architectures -- with the majority implementing  ...  However, vanilla seq2seq models often fail to correctly express a given meaning representation if they lack a strong semantic control mechanism applied during decoding.  ...  The authors would like to thank Lena Reed and Shereen Oraby for help with computing the slot error rate. We would also like to thank Prof.  ... 
arXiv:1901.07931v2 fatcat:qlcgv2r66bfu3cp2foifzfaxbq

A Generative Model for Joint Natural Language Understanding and Generation [article]

Bo-Hsiang Tseng, Jianpeng Cheng, Yimai Fang, David Vandyke
2020 arXiv   pre-print
A key to success in either task is parallel training data which is expensive to obtain at a large scale.  ...  We also show that the model can be trained in a semi-supervised fashion by utilising unlabelled data to boost its performance.  ...  Both models are trained on 5% of the training data. from either predicting not_mention label for certain slots in ground truth semantics; predicting arbitrary values on slots not present in the ground  ... 
arXiv:2006.07499v1 fatcat:6x5aakdhvbfr7j5etcu7dhmjzy
« Previous Showing results 1 — 15 out of 1,939 results