引言分析.docx

1、引言分析APA STYLETEXT 1TEXT 2TEXT 3Jianpeng Cheng,Mirella Lapata(2016).Neural Summarization by Extracting Sentences and WordIpsj Sig Notes,2016,2016:31-36O.Adams,G.Neubig,T.Cohn,S.Bird,Q.T.Do, and S.Nakamura(2016).Learning a Lexicon and Translation Models from phoneme LatticesSubmitted on 23 Mar 2016 (v

2、1), last revised 1 Jul 2016 this version, v3.Conference on Empirical Methods in Natural Lang.,2016:2377-2382S.Cadoni,E.Chouzenoux,J.C.Pesquet, and Caroline Chaux(2016). A block parallel majorize-minimize memory gradient algorithm IEEE International Conference on Image Processing,2016:3194-3198Why it

3、 was chosen?Why it was chosen?Why it was chosen?1.Recent publication(2016-5)2.Published in a scientific journal of ACL3.It is related to the subjects we are learning about.Based on the different classes of summarization models.4.The method of the model is classic and has achived desired effect.1.Rec

4、ent publication(2016-5)2.Published in a scientific journal of ACL3.Experiments demonstrate phoneme error rate improvements against two baselines and the models ability to learn useful bilingual lexical. 4.It is related to the subjects we are learning about.1.Recent publication(2016-3)2.IEEE Internat

5、ional Conference on Image Processing3.This is an application of mathematical knowleage to solve the problem of 3D image.4.The Block Parallel Majorize-Minimize Memory Gradient (BP3MG) algorithm proposed in this paper solves the optimization problem effectively.Abstract1.Classification(1) Report abstr

6、act :This tape of abstract needs reflect purposes, methods, important results and conclusions.报道摘要：这一类摘要反映了文章的目的、方法、重要结果和结论。(2) Indicative abstract: It is an abstract of this thesis and level of the result obtained.指示性摘要：是描述论文的主题，所得结果的水平。(3) Report -indicative abstract:In the form of a reported abst

7、ract, the most valuable part of the thesis is expressed, and the remainder is expressed in an indicative abstract.报道-指示性摘要：以报道性摘要的形式表述论文中价值最高的那部分内容，其余部分则以指示性摘要形式表达。2.Basic elements Abstracts should state the objectives of the project,describe the methods used,summarize the significant findings and s

8、tate the implications of the findings. Elements of Abstract； a. Purpose; （目的） b. Methods; （方法） c.Results; （结果） d.Conclusion.（结论）3.Common tenses Through the study of ten documents, we find that the tenses used in the abstract are the present tense and the past tense.The present perfect tense is occas

9、ionally used.(1) Simple present tense(一般现在时)Used to show a general truth,or to indicate a state,or regular actions or process,it is most commonly used in these papers.Examples: 1. Language documentation begins by gather-ing speech. 2. This architecture allows us to develop different classes of summa

10、rization models which can extract sentences or words. We train our models on large scale corpora containing hundreds of thousands of document-summary pairs. 3. Experimental results on two summarization datasets demonstrate that our models obtain results comparable to the state of the art without any

11、 access to linguistic annotation.(2) Simple past tense（一般过去时）Used to describe the discovery or process of a certain moment in the past.Examples:We used less than 10 hours of EnglishJapanese data from the BTEC corpus (Takezawa et al., 2002), com-prised of spoken utterances paired with textual trans-l

12、ations.(3)present perfect tense（现在完成时）The present perfect tense is the link between the past and the present.Example:The need to access and digest large amounts of textual data has provided strong impetus to de-velop automatic summarization systems aiming to create shorter versions of one or more do

13、cuments, whilst preserving their information content.4.Sentence patterns 1)It is.Smartphone apps for rapid collection of bilin-gual data have been increasingly investigated (De Vries et al., 2011; De Vries et al., 2014; Reiman, 2010; Bird et al., 2014; Blachon et al., 2016). It is common for these a

14、pps to collect speech segments paired with spoken translations in another language, making spoken translations quicker to obtain than phonemic transcriptions.2)There be.In this work we propose a data-driven approach to summarization based on neural networks and continuous sentence features. There ha

15、s been a surge of interest recently in repurposing sequence transduction neural network architectures for NLP tasks such as machine translation (Sutskever et al., 2014), question answering (Hermann et al., 2015), and sentence compression (Rush et al., 2015). 3) Experimental results. Experimental res

16、ults on two summarization datasets demonstrate that our models obtain results comparable to the state of the art without any access to linguistic annotation.4) We present a method to .5) This architecture allows us to.Learning a Lexicon and Translation Model from Phoneme LatticesTitleLanguage docume

17、ntation begins by gathering speech.Provides backgroundManual or automatic transcription at the word level is typically not possible because of the absence of an orthography or prior lexicon, and though manual phonemic transcription is possible, it is prohibitively slow. On the other hand, translatio

18、ns of the minority language into a major language are more easily acquired.Problem descriptionWe propose a method to harness such translations to improve automatic phoneme recognition.Design thinkingThe method assumes no prior lexicon or translation model, instead learning them from phoneme lattices

19、 and translations of the speech being transcribed.Design innovationThe method assumes no prior lexicon or translation model, instead learning them from phoneme lattices and translations of the speech being transcribed.Presents the significance and achievement of the studyNeural Summarization by Extr

20、acting Sentences and WordsTitleTraditional approaches to extractive summarization rely heavily on human-engineered features.Provides background and problem descriptionIn this work we propose a data-driven approach based on neural networks and continuous sentence features. We develop a general frame-

21、work for single-document summarization composed of a hierarchical document encoder and an attention-based extractor.Design techniquesThis architecture allows us to develop different classes of summarization models which can extract sentences or words.Advantages of themethodWe train our models on lar

22、ge scale corpora containing hundreds of thousands of document-summary pairs.Design methodExperimental results on two summarization datasets demonstrate that our models obtain results comparable to the state of the art without any access to linguistic annotation.Design resultsA Block Parallel Majoriz

23、e-minimize Memory Gradient Algorithm TitleIn the field of 3D image recovery, huge amounts of data need to be processed. Provides background and problem descriptionParallel optimization methods are then of main interest since they allow to overcome memory limitation issues, while benefiting from the

24、intrinsic acceleration provided by recent multicore computing architectures. In this context, we propose a Block Parallel Majorize Minimize Memory Gradient (BP3MG) algorithm for solving large scale optimization problems.Design techniquesThis algorithm combines a block coordinate strategy with an eff

25、icient parallel update.Design innovationThe proposed method is applied to a 3D microscopy image restoration problem involving a depth-variant blur, where it is shownto lead to significant computational time savings with respect to a sequential approach. Advantages of theMethod and applicationLearnin

26、g a Lexicon and Translation Model from Phoneme LatticesIntroductionAnalysis Most of the worlds languages are dying out and have little recorded data or linguistic documentation (Austin and Sallabank, 2011).It is important to ad-equately document languages while they are alive so that they may be inv

27、estigated in the future. Language documentation traditionally involves one-on-one elicitation of speech from native speakers in or-der to produce lexicons and grammars that describe the language. However, this does not scale: lin-guists must first transcribe the speech phonemically as most of these

28、languages have no standardized orthography. This is a critical bottleneck since it takes a trained linguist about 1 hour to transcribe the phonemes of 1 minute of speech (Do et al., 2014).(1) This sentence explains the social context of the recorded language and the current research situation(2)The

29、sentence reflects the importance of recording the language(3)Attributive clause: That (4)Cause adverbial clause: Since .Smartphone apps for rapid collection of bilin-gual data have been increasingly investigated (De Vries et al., 2011; De Vries et al., 2014; Reiman, 2010; Bird et al., 2014; Blachon

30、et al., 2016). It is common for these apps to collect speech segments paired with spoken translations in another language, making spoken translations quicker to obtain than phonemic transcriptions.(5) Passive sentence: Have been done .(6) Formal subject: It is .We present a method to improve automat

31、ic phoneme transcription by harnessing such bilingual data to learn a lexicon and translation model directly from source phoneme lattices and their written tar-get translations, assuming that the target side is a major language that can be efficiently transcribed.A Bayesian non-parametric model expr

32、essed with a weighted finite-state transducer (WFST) framework represents the joint distribution of source acoustic features, phonemes and latent source words given the target words. Sampling of alignments is used to learn source words and their target translations, which are then used to improve transcription of the source audio they were learnt from. Importantly, the model assumes no p