论文翻译
Unsupervised Object Discovery and Localization in the Wild:
Part-based Matching with Bottom-up Region Proposals
Minsu Cho
1, Suha Kwak 1, Cordelia Schmid 1,y Jean Ponce 2, 1 Inria 2 ′
Ecole Normale Superieure′ / PSL Research University 野外无监督的目标发现与定位：一部分基于自下而上的区域匹配
Abstract
This paper addresses unsupervised discovery and localization of dominant objects from a noisy image collection with multiple object classes. The setting of this problem is fully unsupervised, without even imagelevel annotations or any assumption of a single dominant class . This is far more general than typical colocalization, cosegmentation, or weakly-supervised localization tasks. We tackle the discovery and localization problem using a partbased region matching approach: We use off-the-shelf region proposals to form a set of candidate bounding boxes for objects and object parts. These regions are efficiently matched across images using a probabilistic Hough transform that evaluates the confidence for each candidate correspondence considering both appearance and spatial consistency. Dominant objects are discovered and localized by comparing the scores of candidate regions and selecting those that stand out over other regions containing them. Extensive experimental evaluations on standard benchmarks demonstrate that the proposed approach significantly outperforms the current state of the art in colocalization, and achieves robust object discovery in challenging mixed-class datasets. 摘要
本文针对无监督的发现及从多对象类噪声图像采集中对主要目标的定位。这个问题的设置是完全无监督的，甚至没有图像级的注释或任何一个单一的显性类的假设。这是远比典型的联合定位，联合分割，弱监督定位任务更普遍。我们处理的发现和定位问题，采用基于区域匹配方法：我们使用现成的区域提案，以形成一组候选边界框的对象和对象的部分。这些区域被有效地匹配在图像使用概率霍夫变换，计算每个候选对应信心考虑外观、空间一致性。主要目标的发现和定位是通过比较候选区的得分并选择那些站在其他区域以外的候选区。广泛的实验评估标准的基准测试表明，该方法明显优于现在的联合定位，在具有挑战性的混合类数据实现鲁棒的目标发现。
1. Introduction
Object localization and detection is highly challenging because of intraclass variations, background clutter, and occlusions present in realworld images. While significant progress has been made in this area over the last decade, as shown by recent benchmark results [11, 16], most state-of-the-art methods still rely on strong supervision in the form of manually-annotated bounding boxes on target instances. Since those detailed annotations are expensive to acquire and also prone to unwanted biases and errors, recent work
has explored the problem of weakly-supervised object discovery where instances of an object class are found in a collection of images without any box-level annotations. Typically, weakly-supervised localization [9, 35, 36, 43, 45, 56] requires positive and negative image-level labels for a target object class. On the other hand, cosegmentation [25, 29, 40] and colocalization [12, 27, 51] assume less supervision and only require the image collection to contain a single dominant object class, allowing noisy images to some degree.
1.简介
目标定位和检测是非常具有挑战性的，因为类内的变化，背景杂乱，和在现实世界中的图像遮挡。虽然在过去十年中已经取得了重大进展，如最近的基准测试结果[ 16，11 ]，大多数国家最先进的方法仍然依赖于强大的监督，在目标的情况下，手动标注边界框的形式。由于这些详细的注释是需要很大代价，也容易出现不必要的偏差和错误，最近的工作研究了弱监督的目标发现问题，在采集的图像中对象类的实例中发现的图像的集合，而无需任何边框注释。典型地，弱监督定位[ 9，35，36，43，45，56 ]需要正面和负面的目标对象类图像的标签。另一方面，cosegmentation [ 25，29，40 ]和[ 12，27，51共定位，承担更少的监督和]只需要图像采集包含一个单一的显性对象类，允许一定程度的噪声图像。
This paper addresses unsupervised object localization in a far more general scenario where a given image collection contain multiple dominant object classes and even noisy im-ages without any target objects. As illustrated in Fig. 1, the setting of this problem is fully unsupervised, without any image-level annotations, an assumption of a single domi-nant class, or even a known number of object classes. In spite of this generality, the proposed method markedly out-performs the state of the arts in colocalization [27, 51] on standard benchmarks [16, 40], and closely competes with current weakly-supervised localization [9, 43, 56].
本文讨论了无监督的对象定位在一个更一般的场景，一个给定的图像采集包含多个主要的对象类，甚至嘈杂的我年龄没有任何目标对象。如图1所示，此问题的设置是完全无监督的，没有任何形象层面的注释，一个单一的显性类的一个假设，甚至是一个已知的对象类。在这种普遍性的怨恨，该方法明显优于共存[ 27国家的艺术，51 ]标准的基准测试[ 16，40 ]，并密切竞争与电流弱监督定位[ 9，43，56 ]。
We advocate a part-based matching approach to unsuper-vised object discovery using bottom-up region proposals. Multi-scale region proposals have been widely used before to restrict the search space for object bounding boxes in ob-ject recognition [9, 20, 53] and localization [9, 27, 51, 54]. We go further and propose here to use these regions to form a set of candidate regions not only for objects, but also for object parts. We use a probabilistic Hough transorm [2] to match those candidate regions across images, and assign them confidence scores reflecting both appearance and spa-tial consistency. This can be seen as an unsupervised and ef-ficient variant of both deformable part models [18, 19] and graph matching methods [5, 14]. Objects are discovered and localized by selecting the most salient regions that con-tain corresponding parts. To this end, we introduce a score that measures how much a region stands out over other re-gions containing it. The proposed algorithm alternates be-tween part-based region matching and foreground localiza-tion, improving both over iterations. 我们提倡一部分以进行修正使用自下而上的区域建议对象发现匹配方法。多尺度区域的建议已经被广泛使用之前，限制搜索空间的对象包围盒在转播—
对象识别[ 9，20，53 ]和[ 27，9定位，51，54 ]。我们进一步提出使用这些区域，以形成一
组候选区域，不仅为对象，但也为对象的部分。我们用一个概率Hough变换[ 2 ]相匹配的候选区域的图像，并给他们信心分数反映的外观和空间一致性。这可以看出，作为一种无监督和EF的变形的零件模型[ 18的变体，19 ]和图匹配的方法[ 5，14 ]。对象的发现和选择所包含的相应部分的最显着的地区定位。为此，我们引入了一个分数，衡量区域站在其他区域包含它。该算法在基于区域匹配之间的前景和定位，提高迭代。
The main contributions of this paper can be summarized as follows: (1) A part-based region matching approach to unsupervised object discovery is introduced. (2) An effi-cient and robust matching algorithm based on a probabilis-tic Hough transform is proposed. (3) A standout score for robust foreground localization is introduced. (4) Object dis-covery and localization in a fully unsupervised setup is ex-plored on challenging benchmark datasets [16, 40].
本文的主要贡献概括如下：（1）介绍了一种基于区域匹配方法的无监督目标发现。（2）一个高效鲁棒匹配算法基于随机Hough变换算法。（3）介绍了一种强大的前景定位出色的成绩。（4）在一个完全无监督的设置对象的发现和定位的探索具有挑战性的基准数据集[ 16，40 ]。
2. Related work
Unsupervised object discovery has long been attempted in computer vision. Sivic et al. [48] and Russell et al. [42] apply statistical topic discovery models. Grauman and Darrel [21] use partial correspondence and clustering of lo-cal features. Kim and Torralba [28] employ a link analy-sis technique. Faktor and Irani [17] propose clustering by composition. Unsupervised object discovery, however, has proven extremely difficult “in the wild”; all of these previ-ous approaches have been successfully demonstrated in a restricted setting with a few distinctive object classes, but their localization results turn out to be far behind weakly-supervised results on challenging benchmarks [12, 28, 51].
2.相关的工作
在计算机视觉中，无监督的对象的发现一直被尝试。Sivic等人。[ 48 ]和罗素等人。[ 42 ]适用于统计专题发现模型。洛杉矶和DAR关系[ 21 ]利用部分对应和局部特征聚类。基姆和济南[ 28 ]利用链接分析技术。该和伊朗[ 17 ]提出了聚类组成。无监督的对象发现，然而，已被证明非常困难的“野生”；所有这些已有的方法已被证明成功的限制和一些独特的对象类的设置，但其定位结果将远弱监督结果上具有挑战性的基准[ 12，28，51 ]。
Given the difficulty of fully unsupervised discovery, re-cent work has more focused on weakly-supervised ap-proaches from different angles. Cosegmentation is the prob-lem of segmenting common foreground regions out of a set of images. It has been first introduced by Rother et al. [38] who fuse Markov random fields with color his-togram matching to segment objects common to two images. Since then, this approach has been improved in nu-merous ways [4, 6, 23, 55], and extended to handle more general cases [7, 25, 40, 54]. Given the same type of in-put as cosegmentation, colocalization seeks to localize ob-jects with bounding boxes instead of pixel-wise segmen-tations. Tang et al. [51] use the discriminative cluster-ing framework of [25] to localize common objects in a set of noisy images, and Joulin et al. [27] extend it to colo-calization of video frames. Weakly-supervised localiza-tion [9, 12, 35, 36, 46, 49] shares the same type of output as colocalization, but assumes a more supervised scenario with image-level labels that indicate whether a target ob-ject class appears in the image or not. These labels enable to learn more discriminative localization methods, e.g., by mining negative images [9]. Recent work on
discriminative patch discovery [15, 44, 50] learns mid-level visual repre-sentations in a weakly-supervised mode, and use them for object recognition [15, 44] and discovery [13, 50].
给出了完全无监督发现的困难，重新分工作多集中在弱监督方法从不同的角度。cosegmentation是分割共同的前景区域的一组图像的问题。它首先介绍了洛特等人的。【38】融合马尔可夫随机域颜色直方图匹配的两个图像常见的对象分割。从那时起，这种方法已经在众多的方法[ 4，6，23，55 ]改进，并扩展到处理更一般的情况下，[ 7，25，40，54 ]。相同类型的放式cosegmentation寻求定位，定位对象与包围盒代替像素点”。唐等。[ 51 ]使用[ 25 ]判别聚类框架定位在一组图像常见的物体，和Joulin等人。[ 27 ]扩展到彩色视频帧定位。弱监督定位方法[ 9，12，35，36，46，49 ]股输出共存同一类型，但具有更监督场景图像水平的标签表明确定目标对象类出现在图像或不。这些标签可以学习更多的歧视定位方法，例如，通过挖掘负的图像[ 9 ]。最近的工作在一个弱监督的模式识别碎片发现学习中级的视觉陈述，并使用它们的对象识别和发现
Region proposals have been used in many of the methods discussed so far, but most of them [12, 27, 28, 42, 51, 54] use relatively a small number of the best proposals (typ-ically, less than 100 for each image) to form whole ob-ject hypotheses, often together with generic objectness mea-sures [1]. In contrast, we use a large number of region pro-posals (typically, between 1000 and 4000) as primitive el-ements for matching without any objectness priors. While many other approaches [7, 40, 41] also use correspondences between image pairs to discover object regions, they do not use an efficient part-based matching approach such as ours. Many of them [7, 21, 40] are driven by correspondence techniques, e.g., the SIFT flow [32], based on generic local regions. In the sense that semi-local or mid-level parts are crucial for representing generic objects [18, 30], we believe segment-level regions are more adequate for object match-ing and discovery. The work of Rubio et al. [41] introduces such a segment-level matching term in their cosegmentation formulation. Unlike ours, however, it requires a reason-able initialization by an objectness measure [1], and does not scale well with a large number of segments and images.
区的建议已被应用在许多讨论的方法到目前为止，但是他们中的大多数人[ 12，27，28，42，51，54 ]使用相对少量的最好的建议（典型地，小于100为每个图像）形成整体目标的假设，通常与通用的对象性措施措施[ 1 ]。相反，我们使用了大量的区域问题（通常，之间的1000和4000）作为原始元素的匹配没有任何对象的先验知识。虽然许多其他的方法也使用图像对发现对象区域之间的对应关系，他们不使用有效的部分匹配的方法，如我们的。他们中的许多是由通信技术，例如，筛流，基于通用的本地区。在这个意义上，中层为代表的局部或部分的通用对象是至关重要的，我们相信段水平地区为对象的匹配和发现更充分。卢比奥等人的工作。引入这样一个段的匹配项在其共同的分割公式。不像我们，然而，它需要一个对象的措施合理的初始化，并没有大量的片段和图像尺度。
3. Proposed approach
For unsupervised object discovery, we combine an effi-cient part-based matching technique with a foreground lo-calization scheme. In this section we first introduce the two main components of our approach, and then describe the overall algorithm for unsupervised object discovery.
3。提出的方法
对于无监督的对象发现，我们结合了一个高效的基于部分的匹配技术与前景定位方案。在这一节中，我们首先介绍我们的方法的主要组成部分，然后描述了整体算法的无监督对象发现。
3.1. Part-based region matching
For part-based matching in an unsupervised setting, we use off-the-shelf region proposals [34] as candidate regions for objects and object parts: Diverse multi-scale proposals include meaningful parts of objects as well as objects them-selves. Let us assume that two sets of region proposals R and R0 have been extracted from two images I and I0, re-spectively. Let r = (f; l) 2 R denote a region with fea-ture f observed at location l. We use 8 8 HOG descrip-tors [10, 22] for f to describe the region patches, and center position and scale for l to specify the location. A match between r0 and r is represented by (r; r0). For the sake ofbrevity, we use short notations D for two region sets, and m for a match: D = (R; R0), m = (r; r0) in R R0. Then,our probabilistic model of a match confidence for m is represented by p(mjD). Assuming a common object appears in I and I0, let the offset x denote its pose displacement between I and I0, related to properties such as position and scale change. p(xjD) becomes the probability of the com-mon object being located with offset x. Now, the match confidence is decomposed in a Bayesian manner：where we suppose that appearance matching ma is indepen-dent of geometry matching mg and an object location offset x. Appearance likelihood p(ma) is simply computed as the similarity between f and f0. Geometry likelihood p(mgjx) is estimated by comparing displacement l0 l to the given offset x. For p(mgjx), we construct three-dimensional off-set bins for translation and scale change, and use a Gaussian distribution centered on offset x.
在无监督的环境中的一部分为基础的匹配，我们使用现成的，现成的地区的建议，作为候选区域的对象和对象的部分：不同的多尺度建议包括有意义的部分对象，以及对象的他们自己。让我们假设两套区域建议R和R0已经从两图像I和I0分别提取。让R =（F；L）2 R表示一个区域的特征在位置L我们使用8 8猪描述等[ 10 ]观察，22 F来描述区域的补丁，和L指定位置中心的位置和规模。R0和R之间的比赛是由（R；R0）。为了简洁，我们使用短符号D两区域设置，和M的比赛：D =（R；R0），M =（R；R0）R R0。然后，我们的概率为M的比赛信心模型为代表的P（MJD）。假设一个共同的对象出现在I和I0，让偏移x表示姿态位移I和I0之间的相关特性，如位置和规模的变化。P（XJD）成为普遍的对象位于偏移X的概率，对手的信心是在贝叶斯方式分解：我们假设外观匹配马是独立的几何匹配MG和对象位置偏移X的外观可能P （MA）是计算F和F0之间的相似性。几何概率P（mgjx）是通过位移l0 l给定偏移量X P估计（mgjx），我们构建了三维平移和尺度变化的集箱，并使用一个高斯分布中心偏移X。
The main issue is how to estimate geometry prior p(xjD) without any information about objects and their lo-cations. Inspired by the generalized Hough transform [2] and its extensions [31, 57], we propose the Hough space score h(xjD), that is the sum of individual probabilities p(m; xjD) over all possible region matches m 2 R R0. The voting is done with an initial assumption of a uniform prior over x:
which predicts a pseudo likelihood of common objects at offset x. Assuming p(xjD) / h(xjD)1, we rewrite Eq.(1)to define the Hough match confidence as：
主要的问题是如何估计几何先验P（XJD）没有关于对象和地点的任何信息。受广义Hough变换[ 2 ]和它的扩展[ 31，57 ]，我们提出了Hough空间分H（XJD），即个体概率的总和（M；XJD）在所有可能的区域匹配M 2 r R0。投票是通过一个统一的事先假设过的：
它预示着一个共同的目标在伪似然偏移X假设P（XJD）/小时（XJD）1，我们重写等式（1）定义Hough比赛的信心:
Interestingly, this formulation can be seen as a combination of bottom-up and top-down processes: The bottom-upprocess aggregates individual votes into the Hough spacescores (Eq.(2)), and the top-down process evaluates eachmatch confidence based on those scores (Eq.(3)).Wecall this algorithm Probabilistic Hough Matching (PHM).Leveraging the Hough space score as a spatial prior, it provides robust match confidences for candidate matches. Inparticular, given multi-scale region proposals, different region matches on the same object cast votes for each other,and make all the region matches
on the object obtain highconfidences. This is an efficient part-based matching procedure with complexity of O(nn0), where n and n0 are thenumber of regions in R and R0, respectively. As shownin Fig. 2c, reliable matches can be obtained when a propermapping constraint (e.g., one-to-one, one-to-many, etc.) isenforced on the confidence as a post-processing。
有趣的是，这一提法可以看作一个组合—自下而上和自上而下的过程：自下而上过程聚集个人票到Hough 空间分数（方程（2）），和自上而下的过程评估比赛的信心基于分数（方程（3））。我们调用该算法匹配概率Hough（PHM）。利用Hough空间分为空间之前，它的亲—为候选人提供强大的比赛信心的比赛。在特定的，给定的多尺度区域的建议，不同的重新—祗园比赛相同的对象互相投票，并使所有的区域上的匹配对象获得高悄悄话。这是一个高效的基于部分匹配的专业—程序复杂度为O（nn0），其中n和N0是在R和R0的地区号码，分别。如图所示图2c，可靠的比赛时可以得到适当的映射约束（例如，一对一，一对多等）是强制执行的信心，作为后处理。
We define the region confidence as a max-pooled matchconfidence for r in R with respect to R0:
which derives from the best matches from R0 to R under one-to-many mapping constraints. High region confidences guarantee that corresponding regions have at least single good matches in consideration of both appearance and spa-tial consistency. As shown in Fig. 2d, the region confidence provides a useful measure for common regions between im-ages, thus functioning as a driving force in object discovery.
我们定义区域的信心，作为一个最大池匹配R R与R0的信心：这源于从R0最匹配的R在一对多映射约束。高区域的信心保证相应区域的外观和空间一致性的考虑，至少有个很好的比赛。如图2所示，该地区的信心提供了一个有用的措施，在我年龄之间的共同区域，从而发挥作用的驱动力在对象发现。
3.2. Foreground localization
Foreground objects do not directly emerge from partbased region matching: A region with the highest confidence is often a salient part of a common object while good localization is supposed to tightly bound the entire object region. We need a principled and unsupervised way to tackle the intrinsic ambiguity in separating the foreground objects from the background, which is one of the main challenges in unsupervised object discovery. In Gestalt principles of visual perception [39] and design [24], regions that “stand out”are more likely to be seen as a foreground. A high contrast lies between the foreground and background, and a lower contrast between foreground parts or background parts. Inspired by these figure/ground principles, we evaluate a foreground score of a region by its perceptual contrast standing out of its potential backgrounds. To measure the contrast, we leverage on the region confidence from part-based matching, which is well supported by the work of Peterson and Gibson, demonstrating the role of object recognition or matching in the figure/ground process [37].
3.2。前景定位
前景物体不直接从部分地区出现：一个匹配可信度最高的地区往往是一个共同的目标一突出部，而良好的定位应该是紧密结合的整个目标区域。我们需要一个有原则和无监督的方式来解决前景对象分离的本质模糊从背景中，这是一个主要的挑战在无监督的对象发现。格式塔原则视觉感知[ 39 ]和设计[ 24 ]，区域“更可能被视为一个前景，一个高对比的是前景和背景，以及前景部分或背景的低对比度部分。受这些图形/背景原则，我们评价—吃了一个前景得分的地区，其感性对比站在它的潜在背景。测量相比之下，我们充分利用了区域信心的一部分—基于匹配，它是由体育工作的支持—Terson和吉普森，展示对象识别的作用—在图/地面的过程[ 37 ]识别或匹配。
First, we generalize the notion of the region confidence to exploit multiple images. Let us assume I as a target image, and I0 as a source image. The region confidence of Eq.(4) is a function of region r in target R with its best correspondence r0 in source R0 as a latent variable. Given multiple source images, it can be naturally extended with more latent variables, meaning the best correspondences from the source images to r. Let us define neighbor images N of target image I as an index set of source images where an object in I may appear.
Generalizing Eq.(4), the region confidence can be rewritten as
首先，我们概括的概念，该地区的信心利用多幅图像。让我们假设我是一个目标—年龄和I0为源图像。区域信心方程（4）是其最好的肺靶R区域函数R—响应R0源R0为潜变量。鉴于多—多个源图像，它可以自然地扩展更多的潜变量，意味着最好的对应关系源图像到R，让我们定义相邻图像的目标图像，我作为一个指标集的源图像中的一个对象在我可能会出现。概括式（4），该地区信心可以改写为（R）=最大C（R；R0）（R；R0）fri0gi2n i2n我我X=最大C（R；R0）（R；R0）；（5）我r02ri0 [J].i2nX
which reduces to the aggregated confidence from the neighbor images. More images may give better confidences.Given regions R with their region confidences, we evaluate a perceptual contrast for region r 2R by computing the increment of its confidence from its potential backgrounds.To this end, we define the standout score as
where r ( rb means region r is contained in region rb.The idea is illustrated in Fig. 3b. Imagine a region gradually shrinking from a whole image region, to a tight object region, to a part region. Significant increase in region confidence is most likely to occur at the point of taking the tight object region. In practice, we decide the inclusive relation r ( rb by two simple criteria: (1) The box area of r is less than 50% of the box area of rb. (2) 80% of the box area of r overlaps with the box area of rb.The standout score reflects the principle that we perceive a lower contrast between parts of the foreground than that between the background and the foreground. As shown in the example of Fig. 3b, we can localize potential object regions by selecting regions with top standout scores.
这减少了聚集的信心从嘶鸣—博尔图像。更多的图像可以提供更好的信心。鉴于区域R与区域的信心，我们评价—吃的区域R 2R的感知对比度的计算来自其潜在背景的信心的增加。为此，我们定义了出色的成绩:其中R（RB是R区域包含在区域Rb。这个想法是在图3b所示。想象一个区域研究—从一个整体图像区域收缩，到一个紧密的对象区域，以局部区域。在区域配置显著增加—都是最有可能发生在以紧点目标区域。在实践中，我们决定了包容关系R（RB的两个简单的准则：（1）R箱面积少比RB的箱区50%。（2）80%的箱面积R与RB的禁区。出色的成绩反映了我们感知原理前景部分的低对比度比背景与前景。如图所示图3b的例子，我们可以重新定位的潜在对象—通过选择最出色的得分区域区域。
3.3. Object discovery algorithm
For unsupervised object discovery, we combine partbased region matching and foreground localization in a coordinate descent-style algorithm. Given a collection of images C, our algorithm alternates between matching image pairs and re-localizing potential object regions. Instead of matching all possible pairs over the images, we retrieve k neighbors for each image and perform part-based matching only from those neighbor images. To make the algorithm robust to localization failure in precedent iterations,we maintain five potential object regions for each image.Both the neighbor images and the potential object regions are updated over iterations.
The algorithm starts with an entire image region as an initial set of potential object regions Oi for each image Ii, and performs the following three steps at each iteration.
3.3。对象发现算法
对于无监督的对象发现，我们结合部分—基于区域匹配和前景定位的协同—坐标下降型算法。给定一个集合的即时通讯—年龄C，我们之间交替的图像匹配算法对和重新定位潜在的对象区域。而不是匹配所有可能对的图像，我们检索每一个图像的邻居和执行基于部分的匹配—只有从那些邻居的图像。为了使算法—在算法的鲁棒迭代定位失败的先例，我们维持五个潜在的目标区域的每个图像。无论是相邻的图像和潜在的对象区域被更新的迭代。
该算法从整个图像区域的潜在目标区域的OI为每个图像II的初始设置，并执行以下三个步骤，在每一次迭代。
Neighbor image retrieval. For each image Ii, k nearest neighbor images fIj j i 2 Nig are retrieved based on the similarity between Oi and Oj. We use 10 neighbor im-ages (k = 10).3 At the first iteration, as the potential ob-ject regions become entire image regions, nearest-neighbor matching with the GIST descriptor [52] is used. From the second iteration, we perform PHM with re-localized object regions. For efficiency, we only use the top 20 region pro-posals according to region confidences, which are contained in the potential object regions. The similarity for retrieval is computed as the sum of those region confidences.
邻域图像检索。每个图像二、K近邻图像基金我2正是基于爱和橙汁之间的相似性检索。我们用10邻域图像（K = 10）。3在第一次迭代中，作为潜在的目标区域成为整个图像区域，近邻与GIST描述子[ 52 ]匹配使用。从第二迭代，我们进行重新定位目标区域的PHM。为了提高效率，我们只使用前20区的设想根据区域的信心，所包含的潜在目标区域。检索相似度计算的那些地区的信心和。
Part-based region matching. Part-based matching by PHM is performed on Ii from its neighbor images fIj j j 2 Nig. To exploit current localization in a robust way, an asymmetric matching strategy is adopted: We use all re-gions proposals in Ii, whereas for the neighbor image Ij we take regions only contained in potential object regions Oj. This matching strategy does not restrict potential ob-ject regions in target Ii while effectively utilizing localized object regions at the precedent step.
基于零件的区域匹配。部分匹配的PHM进行二邻图像j 2该基金。利用一个强大的电流定位，采用不对称匹配策略：我们使用的所有区域建议II，而相邻图像IJ我们采取区域只包含在潜在目标区域的OJ。这种匹配策略不限制潜在目标区域的目标而有效利用定位目标区域的先例步骤。
Foreground localization. For each image Ii, the stand-out scores are computed so that the set of potential object regions Oi is updated to that of regions with top standout scores. This re-localization improves both neighbor image retrieval and region matching at the subsequent iteration.
前景定位。每个图像II，脱颖而出的分数计算，潜在目标区域的设置更新，顶级出色的得分区域。这种重新定位在随后的迭代提高了邻域图像检索和区域匹配。
These steps are repeated for a few iterations until near-convergence. As will be shown in our experiments, 5 itera-tions are sufficient as no significant change occurs in more iterations. Final localization is done by selecting the most standing-out region at the end. The algorithm is designed based on the idea that better localization makes better re-trieval and matching, and vice versa. As each image is inde-pendently processed at each iteration, the algorithm is eas-ily parallelizable in computation. Object discovery on 500 images takes less than an hour with a 10-core desktop com-puter, using our current parallel MATLAB implementation.
这些步骤是重复几个迭代，直到接近收敛。将在我们的实验显示，5，有足够的迭代没有显着的变化发生在更多的迭代。最后的定位是通过选择最终末的区域在结束。该算法是基于的想法，更好的定位使得更好的检索和匹配设计，反之亦然。为每个图像独立处理在每次迭代中，该算法易于并行计算。500图像对象发现以小于10核桌面电脑一小时，利用我们现有的并行的MATLAB实现。
4. Experimental evaluation
The degree of supervision used in visual learning tasks varies from strong (supervised localization [18, 20]) to weak (weakly-supervised localization [9, 46]), very weak (colocalization [27, 51] and cosegmentation [40]), and null (fully-unsupervised discovery). To evaluate our approach for unsupervised object discovery, we conduct two types of experiments: separate-class and mixed-class experiments. Our separate-class experiments test performance of our ap-proach in a very weakly supervised mode. Our mixed-class experiments test object discovery ”in the wild” (in a fully-unsupervised mode), by mixing all images of all classes in a dataset, and evaluating performance on the whole dataset. To the best of our knowledge, this type of localization experiments has never been fully attempted be-fore on
challenging real-world datasets. We conduct ex-periments on two realistic benchmarks, the Object Discov-ery [40] and the PASCAL VOC 2007 [16], and compare the results with those of the current state of the arts in coseg-mentation [29, 25, 26, 40], colocalization [8, 12, 42, 27, 51], and weakly-supervised localization [9, 12, 35, 36, 46, 56].
PASCAL VOC(pattern analysis,statistical modelling and computational learning visual object classes)模式分析，统计建模，计算学习视觉物体分类。
4。实验评价
用于视觉学习任务的监督程度由强（监督定位[ 18，20 ]）弱（弱监督定位[ 9，46 ]），很弱（共存[ 27，51 ]和[ 40 ]，并cosegmentation）空（完全无监督的发现）。为了评估我们的方法的无监督对象发现，我们进行了两个实验类型：单独的类和混合类实验。我们单独的类实验性能测试我们的方法在一个非常弱的监督模式。我们的混合类实验测试对象发现“在野外”（在一个完全无监督模式），通过混合所有类别的所有图像的数据集，并在整个数据集的性能评价。我们的知识，这种类型的本地化实验从来没有被充分地尝试在挑战性的真实世界的数据集。我们进行实验的两个现实的基准，对象发现[ 40 ]和PASCAL VOC 2007 [ 16 ]，并比较结果与艺术的当前状态在CoSeg心理[ 29，25，26，40 ]，与[ 8，12，42，27，51 ]，弱监督定位[ 9，12，35，36，46，56 ]。
4.1. Evaluation metrics
The correct localization (CorLoc) metric is an evalua-tion metric widely used in related work [12, 27, 46, 51], and defined as the percentage of images correctly localized
according to the PASCAL criterion: area(bp\bgt) > 0:5,
area(bp[bgt)
where bp is the predicted box and bgt is the ground-truth box. The metric is adequate for a conventional separate-class setup: As a given image collection contains a single target class, only object localization is evaluated per image. In a mixed-class setup, however, we have another dimen-sion involved: As different images may contain different object classes, associative relations across the images need to be evaluated. As such a metric orthogonal to CorLoc, we propose the correct retrieval (CorRet) evaluation metric de-fined as follows. Given the k nearest neighbors identified by retrieval for each image, CorRet is defined as the mean per-centage of these neighbors that belong to the same (ground-truth) class as the image itself. This measure depends on k, fixed here to a value of 10. CorRet may also prove useful in other applications that discover the underlying “topology”(nearest-neighbor structure) of image collections. CorRet and CorLoc metrics effectively complement each other in the mixed-class setup: CorRet reveals how correctly an im-age is associated to other images, while CorLoc measures how correctly an object is localized in the image.
4.1。评价指标
正确的定位（corloc）度量是一个评价指标广泛应用于相关工作[ 12，27，46，51 ]，和的百分比定义为图像的正确定位
根据帕斯卡标准：面积（BP BGT）> 0:5，
地区（BP [ BGT）
在BP预测箱和BGT是地面真理盒。的度量是足够的常规的单独的类设置：作为一个给定的图像采集包含一个单一的目标类，只有对象的定位是每个图像的评价。然而在混合班的设置，我们有另一个维度：不同的图像可能包含不同的对象类之间的关联关系的图像需要进行评估。作为这样一个度量corloc正交，我们
提出正确的检索（正确）的评价指标定义如下。鉴于K近邻经检索每个图像，准确的定义为每个这些邻居，属于相同的百分比平均（真实）类图像本身。这项措施取决于钾，固定在这里的值为10。正确的也可能有助于其他应用程序，发现潜在的“拓扑”（近邻结构）的图像集合。纠正和corloc度量有效互补的混合类设置：正确揭示了如何正确的图像与其他图像相关，而corloc措施如何正确的对象定位在图像。
4.2. The Object Discovery dataset
The Object Discovery dataset [40] was collected by the Bing API using queries for airplane, car, and horse, re-sulting in image sets containing outlier images without the query object. We use the 100 image subsets [40] to enable comparisons to previous state of the art in cosegmentation and colocalization. In each set of 100 images, airplane, car, horse have 18, 11, 7 outlier images, respectively. Following [51], we convert the ground-truth segmentations and coseg-mentation results of [29, 25, 26, 40] to localization boxes.
We conduct separate-class experiments as in [12, 51], and a mixed-class experiment on a collection of 300 images from all the three classes. The mixed-class image collection contains 3 classes and 36 outlier images. Figure 4 shows theaverage CorLoc and CorRet over iterations, where we seethe proposed algorithm quickly improves both localization(CorLoc) and retrieval (CorRet) in early iterations, and thenapproaches a steady state. In the separate-class setup, Cor-Ret is always perfect because no other object class existsin the retrieval. As we have found no significant change inboth localization and retrieval after 4-5 iterations in all ourexperiments, we measure all performances of our method inthis paper after 5 iterations. The separate-class results arequantified in Table 1, and compared to those of state-of-the-art cosegmentation [29, 25, 26] and colocalization [40, 51]methods. Our method outperforms all of them in this setup.
The mixed-class result is in Table 2, and examples of localization are shown in Fig. 5. Remarkably, our localizationperformance in the mixed-class setup is almost the same asthat in the separate-class setup. Localized objects are visualized in red boxes with five most confident regions insidethe object, indicating parts most contributing to discovery.Table 2 and Fig. 4 show that our localization is robust tonoisy neighbor images retrieved from different classes.
我们进行单独的类实验，如在[ 12，51 ]，和一个混合的类实验上收集的300幅图像，从所有的三个类。混合类图像集合包含3个类和36个异常图像。图4显示了在迭代的平均corloc和正确的，在这里我们看到了算法快速提高定位（corloc）和检索（正确的）在早期的迭代，然后趋于一个稳定状态。在单独的类中设置，COR RET永远是完美的因为没有其他对象类中存在的检索。我们已经经过4-5迭代在我们所有的实验发现在定位和检索无显著变化，我们测量了所有的表演我们的方法，经过5次迭代。单独的类结果在表1被量化，并与国家的最先进的cosegmentation [ 29，25，26 ]和[ 40，51 ]方法共存。我们的方法优于所有这些在这个设置。
混合类的结果是在表2中，和本地化的例子如图5所示。值得注意的是，我们的本地化性能的混合类设置几乎是相同的，在单独的类设置。本地化的对象是可视化的红色框中有五个最有信心的区域内的对象，显示部分最有助于发现。表4和图2显示我们的定位是强大的，从不同类别的噪声邻居图像检索。
4.3. PASCAL VOC 2007 dataset
The PASCAL VOC 2007 [16] contains realistic imagesof 20 object classes. Compared to the Object Discoverydataset, it is significantly more challenging due to considerable clutter, occlusion, and diverse viewpoints. To facilitate a scale-level analysis and comparison to other methods,we conduct experiments on two subsets of different sizes:PASCAL07-6x2 and PASCAL07-all. The PASCAL07-6x2subset [12] consists of all images from 6 classes (aeroplane, bicycle, boat, bus, horse, and motorbike) of train+valdataset from the left and right
aspect each. Each of the12 class/viewpoint combinations contains between 21 and50 images for a total of 463 images. For a large-scale experiment with all classes following [9, 12, 36], we take alltrain+val dataset images discarding images that only contain object instances marked as “difficult” or “truncate”.
Each of the 20 classes contains between 49 and 1023 imagesfor a total of 4548 images. We refer to it as PASCAL07-all.Experiments on PASCAL07-6x2. In the separate-classsetup, we evaluate performance for each class in Table 3,where we also analyze each component of our method by
removing it from the full version: ‘w/o MOR’ eliminatesthe use of multiple object regions over iterati ons, thus main-taining only a single potential object region for each image. ‘w/o PHM’ substitutes PHM with appearance-basedmatching without any geometric consideration. ‘w/o STO’replaces the standout score with the maximum confidence.As expected, we can see that the removal of each component damages performance substantially. In particular, itclearly shows both part-based matching (using PHM) andforeground localization (using the standout score) are crucial for robust object discovery. In Table 5, we quantitatively compare ours to previous results [8, 12, 47, 51]on PASCAL07-6x2. Our method significantly outperformsthose with a large margin. Note that our method doesnot incorporate any form of object priors such as off-the-shelf objectness measures [12, 47, 51], and only use positive images (P) without more training data, i.e., negativeimages (N) [12, 47]. For the mixed-class experiment, werun our method on a collection of all class/view images inPASCAL07-6x2, and evaluate its CorLoc and CorRet peformance in Table 4. To better understand our retrieval performance per class, we measure CorRet for classes (regardless of views) in the third row, and analyze it by increasingthe numbers of iterations and neighbor images in Fig. 6.This shows that our method achieves better localization andretrieval simultaneously, and benefits from each other. InFig. 7, we show example results of our mixed-class experiment on PASCAL07-6x2. In spite of a small size of objectseven partially occluded, our method is able to localize instances from cluttered scenes, and discovers confident object parts as well. From Table 5, we see that even withoutusing the separate-class setup, the method localizes targetobjects markedly better than recent colocalization methods.
4.3。PASCAL VOC 2007数据集
PASCAL VOC 2007 [ 16 ]包含20类对象逼真的图像。与对象的发现数据集相比，它是显着更具挑战性的，由于相当大的杂波，闭塞，和不同的观点。促进规模水平分析与其他方法相比，我们的行为在两个大小不同的亚群的实验：pascal07-6x2和pascal07所有。的pascal07-6x2子集[ 12 ]包括所有图片来自6班（飞机，车，船，车，马，和摩托车）火车+ VAL数据集从左侧和右侧各方面。12类/视点组合中的每一个包含21和50幅图像，共463幅图像。对于所有的类如下[ 9，12，36 ]一个大规模的实验中，我们把所有的火车+瓦尔数据丢弃的图像，图像只包含对象实例标记为“困难”或“截断”。
20个类中的每一个包含49和1023幅图像，共4548幅图像。我们称它为pascal07-all.experiments在pascal07-6x2。在单独的类设置中，我们评估每一类的性能表3中，我们也分析了我们的方法的每个组件
把它从完整版：“W／O MOR的消除了多目标区域在迭代的使用，因此主要的泰宁只有一个潜在的目标区域的每一个图像。W / O PHM的代用品PHM的外观匹配没有任何几何的考虑。“W / O STO”取代以最大的信心的出色的成绩。正如预期的那样，我们可以看到，每个组件的性能大大损害去除。特别是，它清楚地表明这两部分匹配（使用PHM）和前景定位（使用出色的得分）是强大的对象发现的关键。在表5中，我们定量比较我们以前的结果[ 8，12，47，51 ]对pascal07-6x2。我们的方法显着优于那些大幅度。请注意，我们的方法不包含任何形式的对象的先验如现成对象的措施[ 12，47，51 ]，只使用正面图像（P）没有更多的训练数据，即负面形象（N）[ 12，47 ]。对于混合班的实验，我们对所有类/查看图像在pascal07-6x2采集方法，并评价其corloc和准确性能表4。为了更好地理解我们的检索性能每班，我们衡量正确的类（无论观点）排在第三，并通过增加迭代次数和图6中相邻图像的数字。这表明，我们的方法达到更好的定位和检索的同时，从对方的好处。在图7中，我们展示了我们的实验结果pascal07-6x2混合班。尽管一个小规模的对象，甚至部分遮挡，我们的方法是能够本地化的情况下，从混乱的场景，并发现有信心的对象的部分。从表5中，我们看到，即使不使用单独的类设置，定位目标对象的方法明显优于最近的定位方法。
Experiments on PASCAL07-all.Here we tackle a muchmore challenging and larger-scale discovery task, usingall the images from the PASCAL07 dataset. We firstreport separate-class results, and compare our results tothose of the state of the arts in weakly-supervised localization [9, 36, 43, 47, 45, 46, 56] and colocalization [27] in Table 6. Note that weakly-supervised methods use more training data, i.e., negative images (N). Also note that the bestperforming method [56] uses CNN features pretrained onthe ImageNet dataset [11], thus additional supervised data(A). Surprisingly, the performance of our method is veryclose to the best of weakly-supervised localization [9] not
using such additional data.In the mixed-class setting, we face an issue linked to thepotential presence of multiple dominant labeled (ground-truth) objects in each image. Basically, both CorLoc andCorRet are defined as a per-image measure, e.g., CorLocassigns an image true if any true localization is done in theimage. For images with multiple class labels in the mixed-class setup, which is the case of PASCAL-all with highlyoverlapping class labels (e.g., persons appear in almost 1/3of images), CorLoc needs to be extended in a natural manner. To measure a class-specific average CorLoc in such amulti-label and mixed-class setup, we take all images containing the object class and measure their average CorLocfor the class. The upper bound of this class-specific average CorLoc may be less than 100% because only one localization exists for each image in our setting. To complement this, as shown at the last column of Table 7, weadd the ‘any’-class average CorLoc, where we assign animage true if any true localization of any class exists inthe image. The similar evaluation is also done for CorRec. Both ‘any’-class CorLoc and CorRet have an upperbound of 100% even when images have multiple class labels, whereas those in ‘Av.’ (average) may not. Notethat the mixed-class PASCAL07-all dataset has a very imbalanced class distribution: the 20 classes have very different numbers of images, from 49 (sheep) to 1023 (person). Yet, as quantified in Table 7, our method still performs well even in this unsupervised mixed-class setting,and its localization performance is comparable to that in the
separate-class setup. To better understand this, we visualize in Fig. 8 a confusion matrix of retrieved neighbor images based on the mixed-class result, where each row corresponds to the average retrieval ratios (%) by each class.Note that the matrix reflects class frequency so that the person class appears dominant. We clearly see that despiterelatively low retrieval accuracy, many of retrieved imagescome from other classes with partial similarity, e.g., bicycle- motorbike, bus - car, etc. Figure 9 shows a typical example of such cases. These results strongly suggest that ourpart-based approach to object discovery effectively benefitsfrom different but similar classes without any class-specificsupervision. Interestingly, the significant difference in retrieval performance (CorRet) from 100% in the separate-class setup influences much less on localization (CorLoc).Further analysis of our experiments also reveals that in thecase of an imbalanced distribution of classes, a class withlower frequency is harder to be localized than a class withhigher frequency. To see this, consider ‘the highest’ (person, car, chair, dog, cat) and ‘the lowest’ (sheep, cow,boat, bus, dinningtable) in class frequency. We have measured how much the average performance changes betweenthe separate-class (clean) and mixed-class (imbalanced) settings. The average CorLoc of ‘the highest’ only drops by1.2%, while that of ‘the lowest’ drops by 9.4%. This clearlyindicates that a class with lower class frequency is harder tolocalize in the mixed-class setting. Retr ieval performanceof ‘the lowest’ (CorRet 11.0%) is also much worse than
that of ‘the highest’ (CorRet 30.7%). For more information, see our project webpage: http://www.di.ens.fr/willow/research/objectdiscovery/.
实验pascal07-all.here我们应对更多的挑战和更大规模的发现任务，使用从pascal07数据集的所有图像。我们首先报告单独的类的结果，和比较我们的结果与国家的艺术在弱监督定位[ 9，36，43，47，45，46，56 ]和[ 27 ]表6共存。请注意，弱监督的方法使用更多的训练数据，即，负的图像（氮）。还请注意，最好的方法[ 56 ]采用美国有线电视新闻网特点pretrained在ImageNet数据集[ 11 ]，因此额外的监控数据（一）。
令人惊讶的是，我们的方法的性能是非常接近的最好的弱监督本地化[ 9 ]不
使用这些附加数据，在混合类设置中，我们面临着一个问题，在每个图像中的潜在存在的多个标记的标记（地面）对象的存在。基本上，都corloc和正确的定义为每一个图像的措施，例如，corloc指定一个图像如果真有真正的定位是图像做。在混合的班级设置多类标签的图像，这是高度重叠的类标签Pascal所有案例（例如，人几乎出现在1 / 3的图像），corloc需要以自然的方式扩展。衡量一个特定类的平均corloc这么多标签和混合的班级设置，我们将所有包含该对象的类的图像和测量他们的班级平均corloc。这类特定的平均corloc上界可以小于100%因为我们设置每个图像只有一个存在局限性。作为补充，在表7的最后一栏中，我们添加任何阶级的平均corloc，我们指定一个图像如果真有真实的定位任何一类存在于图像。类似的评价也做了修正。无论任何阶级的corloc和正确的有100%的上限，即使图像有多个类的标签，而在“AV”（平均值）可能不。注意，混合班pascal07所有的数据集有一个非常不平衡类分布：20类有不同数量的图像，从49（羊）到1023（人）。然而，在表7中的量化，我们的方法仍然表现良好，即使在这种无监督的混合类设置，其定位性能与在
单独的类设置。为了更好地理解这一点，我们在图8中的混淆矩阵的基础上的混合类结果，其中每一行对应的平均检索率（%）由每个类的检索相邻图像，注意，矩阵反映类的频率，使人类出现占主导地位。我们清楚地看到，尽管检索精度相对较低，许多的检索图像来自其他类的部分相似，例如，自行车-摩托车，公共汽车-汽车等。图9显示了一个典型的例子，这种情况。这些结果有力地表明，我们的部分为基础的方法，有效地从不同的，但类似的类，没有任何类特定的监督，有效的利益。有趣的是，在检索性能的显著差异（正确的）在单独的类中设置100%个影响定位少得多（corloc）。我们的实验进一步分析还发现，在类分布不平衡的情况下，频率较低的一类很难被本地化比频率较高的一类。看到这一点，认为“最高”（人、车、椅子、狗、猫）和“最低”（羊、牛、船、汽车、dinningtable）上课频率。我们测量的平均性能变化之间的单独的类（清洁）和混合类（不平衡）设置。平均corloc“最高”只下降了1.2%，而'最低的下降了9.4%。这清楚地表明，一类具有较低的类频率是难以本地化的混合类设置。检索性能的“最低”（对11%）也比差远了
这“最高”（修正30.7%）。更多信息，看我们的项目网站：HTTP：/ / www.di。ENS。FR /柳/研究/ objectdiscovery /。
5. Discussion and conclusion
We have demonstrated unsupervised object localizationin the challenging mixed-class setup, which has never beenfully attempted before on a challenging dataset such as [16].The result shows that the effective use of part-based matching is a crucial factor for object discovery. In the future,we will advance this direction and further explore how tohandle multiple object instances per image as well as buildvisual models for classification and detection. In this paper,our aim has been to evaluate our unsupervised algorithmperse, and have thus abstained from any form of additionalsupervision such as off-the-shelf saliency/objectness measures, negative data, and pretrained features. The use ofsuch information will further improve our results.
5。讨论和结论
我们已经证明无监督的对象定位在具有挑战性的混合类设置中，从未有过充分尝试之前在一个具有挑战性的数据集，如[ 16 ]。结果表明，基于部分匹配的有效使用—是对象发现的一个重要因素。在未来，我们将推进这一方向，并进一步探讨如何处理每个图像的多个对象实例以及构建用于分类和检测的视觉模型。在本文中，我们的目标是评估我们的无监督算法
本身，并因此避免任何形式的附加监督如现成的显著性/对象的方法—措施，负面数据，和pretrained特征。使用这些信息将进一步提高我们的研究结果。
有关英汉翻译论文
有关英汉翻译论文 范文一：生态学视域下的英汉翻译 一、引言:英汉翻译研究生态学视域的成立 生态学是研究物质与周围环境关系的一门科学，其发轫之初很长时间里被用于自然科 学研究。而随着生态学的发展和交叉、跨学科研究日趋成熟，社会科学的研究者们发现了 生态学与人文社会科学的相容性，认为人文社会领域和自然一样，事物与事物、事物与环 境间都发生着联系，没有独立于环境之外的事物和现象，故而认为生态学研究的理论、方 法同样适用于社会科学研究。生态学研究与社会科学研究遂真正结合起来。20世纪50年 代以来，人们对文本的认识由仅仅关注认识文本内部自足性问题发展到研究文本内外部关 联问题，研究的视野也随之由微观文字向宏观文化转向，人们对文本的认识，已无法将文 本语言与该语言所蕴含或显现的某国或某民族的文化相割断。对于成功的翻译者来说，他 不仅要作为两种语言的专家，还要是两种语言各自代表的文化的熟知者。对于英汉翻译的 研究者来说，他们需要不断探寻着解释原文本与译文本之间形成文化差异的依据———这 些依据就是社会各层面因素影响文本文字的关系。生态学强调事物与其环境协调、互动、 互相促进。在英汉翻译领域，一个成功或优秀的译作应当是不仅内部自足和谐发展，而且 内部与外部互惠互利、共生共栖的文本。其动态性体现在这种交互作用是永无止息、不断 旋螺式发展的，而非一成不变或者无意义的重复。用生态学视角看待英汉翻译，为的是建 立一种整体性思维。有学者为“整体性”的意义进行了阐释:“整体性是生命的基本属性，整体性不是部分的简单累加，也不是由外力推动而形成的原子集合体，它有着超越部分之 和的更为丰富的内涵和属性。”对翻译及其研究工作来说，把握“整体性”是关键。翻译 活动涵盖原文本和译文本，涉及原文作者、译者、译文读者3方，包括语言和文化2个层面，同时受到政治环境、经济环境、主流意识形态等的影响。梳理和廓清这些形成翻译活 动的要素的动态平衡发展，有助于建构英汉翻译研究的生态学视域。 二、原文本与译文本的交互解读 原文本一经生成，不是一成不变，而是在各种读者解读下成为“一百个读者有一百个 哈姆雷特”的消费品。如果没有读者的阅读，则文本只是完成了生产过程，惟有读者的阅 读才能使文本的意义得以生成，使文本价值得以实现。译文本的产生就是原文本蕴涵得到 复活的一个过程，译文本的产生同时是对原文本的“消费”，促成原文本价值的重新生成 和意义再现。译文本对原文本的意义即在于此:扩大或者新生成原文本意义。译者与原文 作者的生活背景、个人知识、经验、思维方式等都有很大差异，译者对原文本的意义进行 解读的过程中有译者的前结构参与，这种前结构大大影响了原文本意义的重现:原文作者 将生活的客观信息内化于文本，译者在翻译之前先作为原文本的读者，在阅读过程中以自 己的前结构对原文本进行内化，但这次内化的主体是译者而非原文作者，由此不同主体所 作的“内化”经由不同文本载体产生出不同的文本接受。原文本意义和译者根据前结构而 解读出来的意义不同，所以可以说，译者通过内化能扩大原文本的意义，使原文本意义更 加多样。接受美学和读者反应批评两种理论都强调读者阅读对文本的意义，认为读者也是
翻译类论文模板
摘要（三号宋体，加粗，居中） 语言和文化的关系近年来已成为研究热点。二十世纪八、九十年代，文化教学研究在西方世界已获得大力发展。受此影响，国内的外语文化研究也取得一些进展, 研究人员日益关注如何把对文化的理解和掌握贯穿于第二语言和外语的教与学中。语言学家伊利欣克尔教授(Eli Hinkel)主编的《第二语言教与学的文化因素》一书讲述了关于第二语言文化的研究。该书重点讲述文化因素在语言研究方面的应用及其研究成果在语言教学中的应用，并阐明了第二语言的学习以及人们的世界观是如何在第二语言和第二文化的背景中影响其学习、理解及交流的。 在该书引言部分，原作者简要介绍了文化研究与文化教学的重要概念和理论基础，并对每个章节加以评述，从而揭示了文化因素是如何影响第二语言学习的各个方面的。第一章主要介绍了如何理解文化和语言的关系，并通过对香港大学学生课堂行为的研究, 追溯到东西方课堂文化差异的渊源。另外，对苏格拉底与孔子有关教育和生活、教育目标、教育的作用、课堂活动的框架等内容进行了讨论。本论文作者在对以上部分进行翻译与思考之后，得出这样一个观点和结论：学生的课堂行为与教师的期望存在差距，是源自不同的教学思想和哲学根源。这就给外语教学提出了一个亟待重视和解决的问题，即要把对文化的理解和掌握始终贯穿在外语教学中。 （中文摘要内容不少于500汉字，为小四号宋体） 关键词（小四号宋体、加粗）：（冒号）翻译；第二语言；教与学；文化因素；讨论（小四号宋体，关键词之间用分号分隔）
（此处另起一页，方法：点出“插入”---分隔符---分页符---确定即可） Abstract（三号，Times New Roman，居中）In recent years, the relationship between language and culture has become hot in research, and culture in teaching research has achieved great development. Researchers have become increasingly aware of how to promote cultural understanding and mastery in foreign language teaching and learning. The language professor Eli Hinkel has published a book named Culture in Second Language Teaching and Learning which gives advice about the research of foreign language culture. This book focuses on culture as it is applied to language research and the applications of its findings in language pedagogy. In addition, it focuses on second language learning and the ways in which people’s worldviews affect their learning, understanding and communication. Then, the author continues to introduce the main content of the two chapters translated. The introduction part mainly talks about some important conceptions and fundamental theories about culture and language in teaching research, and the original author’s comment. It points out how culture factors have influenced many different aspects of second language learning and using. Then Chapter One touches briefly on just a few of the studies that have contributed to an understanding of the relationship between culture and language, Through researches on students’ classroom behavior in Hong Kong University, the origin of the cultural differences has been explored. Besides, Chapter One also details the viewpoints of Confucius and Socrates in education and life, the role of education, teacher’s role, and arrangement of classroom activities. The conclusion can be made by the author of the thesis that t he reason why students’ classroom behaviors and teachers’ expectations disparate is that they have different teaching ideological and philosophical roots. This has caused a serious problem with the foreign language teaching --- culture has to run through foreign language teaching and learning practice all along. （英文摘要内容应与中文摘要内容相对应，小四号Times New Roman） Key words（小四号Times New Roman，加粗）:（冒号）translating; second language; teaching and learning; culture; discourse （与中文关键词对应，小四号Times New
文学翻译论文
从形合与意合辩证法角度观古诗翻译策略 (潘静20084033042 英语师范三班) 摘要：形合( hypotaxis ) 和意合（parataxis）是英语和汉语篇章组织的常用手段，也是英汉两种语言的重要区别所在。本文提出应当运用哲学中的辩证法原理看待语言中的形合与意合，真正做到学术研究全面、客观、严谨以及科学。一方面，要重视句子及其以下层面的语言特点：英语侧重形合，汉语侧重意合；另一方面，更要研究英汉语的意合与形合在句子层面以上（段落、篇章）所具备的不同作用：英语篇章使用词汇手段等意合组织方式，而汉语篇章中则采用重复和骈偶等形合手段【1】。本文利用对《枫桥夜泊》这首古诗的英译研究分析，强调辩证地、动态地看待英汉语中的形意合问题，改进汉语古诗翻译策略，使译作体现并契合原诗作者的思想与内容，使译作读者明白并理解原诗的内涵与文化。 关键词：形合意合辩证法古诗翻译 Abstract: The very common means of English and Chinese discourse are hypotactic and paratactic, as well as the important differences. This paper argues that hypotaxis and parataxis in a language should be researched in a dialectic way. Only in this way, can academic research become more complete and objective. On the one hand, our attention should be paid to the characteristics of English and Chinese sentences and structures below the sentences leve l–English is more hypotactic while Chinese is more paratactic. On the other hand, it is very important and necessary to study the different functions of hypotactic and paratactic structures at the level of paragraphs in the two languages. Lexically paratactic structures are often found in English paragraphs, and hypotactic structures in the form of repetition and parallelism are frequently employed in Chinese. The author attaches the high importance to research languages in a dialectic way by the analysis of Mooring at night by Maple Bridge. So that we can find some strategies about the poem translation and make the successful exchange of two kinds of people with various background. Key words: hypotaxis, parataxis, dialectic way, poem translation 1.引文 形合与意合是英汉对比研究和翻译理论研究的重要课题之一，也是英汉语篇组织的重要手段之一。所谓形合就是依仗形式（包括词的变化形态，词汇的衔接等）将语言符号由“散”（个体的词）到“集”（组织乃至语篇）的语言组织手段，具有外显性和刚性；而意合则是依仗意义，即内在的逻辑关系组织语言的手段，具有内隐性和柔性。
英汉翻译学期论文
英汉翻译学期论文 国际商务中信用证翻译特点及策略学期论文题目: __ 指导教师： 学号： 姓名： 外国语学院（部）英语专业2012 届 2012年4月16日
Translation Characteristics and Strategies of Letters of Credit in international Business By Submitted to School of Foreign Languages in Partial Fulfillment of the Requirements for the Term Paper For English-Chinese Translation at Under the Supervision of
Translation Characteristics and Strategies of Letters of Credit in international Business ABSTRACT Payment in international trade is crucial for both the seller and the buyer. With the development of international trade, Letters of Credit payment gradually came into being as banks and financial institutions were involved in international trade settlement. With the help of this method of payment, the payment responsibility of the importer is transferred to bank to ensure the safety, as a result, the rapid payment is received by the seller, and meanwhile, the regular shipping documents are given to the buyer conveniently. Therefore, with the benefit of Letters of Credit, the distrust contradiction between the seller and the buyer is solved to a certain extent. Nowadays, payment by Letters of Credit has become the widely adopted method. Letters of Credit English for specific purposes has its own unique vocabulary features. This paper summarizes characteristics of its translation from the credit classification, use of the process, and starting letter words in English, and explores the English translation strategy. It is divided into three chapters: The first chapter probes into the definition of the Letter of Credit, its classification and its use process in international trade; the second describes the characteristics and common errors of the credit translation; The last one is the focus of this article, discussing Letters of Credit translation strategy based on these two chapters, from the perspective of the translation Communicative Theory, Skopos Theory, Functional Equivalence etc. Keywords: Letters of Credit Translation; characteristics; strategies; Functional Theory
英语翻译论文开题报告精编版
英语翻译论文开题报告 精编版 MQS system office room 【MQS16H-TTMS2A-MQSS8Q8-MQSH16898】
英语翻译论文开题报告 时间:2015-03- 12Bell.?Translation?and?Translating:?Theory?and?Practice.?Beijing:?Forei gn?Language?Teaching?and?Research?Press,?2006. 崔长青,?张碧竹.?翻译的要素[M].?苏州:?苏州大学出版社,?2007. 李琏.?英式显性词性转换与英语写作[J].?新疆教育学学 报,?2003,?19(1):?85-89. 李连生.?英汉互译中的词性转换[J].?武汉交通管理干部学院学 报,?1996,?(1):102-107. 项伙珍.?谈翻译中的转性译法[J].?长江职工大学学报,?2000,?17(3):?46-48. 叶海燕.?翻译中的词性转换及换形[J].?安徽工业大学学报(社会科学 版),?2005,?22(3):60-61. (责任编辑：1025) 三、对英文翻译中词类转换的引入 从语言的角度来分析，对一门语言中的词性以及在语句中的成分进行分析，?从转喻理论的相关知识出发，来探讨英汉语言翻译中词类转换的概念和知识，并就其异同和特性进行对比，不仅是当前国内外学者研究的重点问题，也是大学英语教学中始终关注的热点。?为?此?，从英文翻译的实践中，?通过例证或典型问题的互译，从语言结构及表达习惯上进行探讨英文翻译中的转换方法，以增强语言翻译的可读性和准确性，并从中探讨出词类转换的规律，帮助更多的学生从中获得有益的指
合肥工业大学论文翻译
论文翻译 设计题目: 统计力学坐标膜致密的自我避免 学生姓名：刘全乐 学号：20100489 专业班级：机械设计制造及其自动化10-4班 指导老师：翟华 时间：2013年06月13日
统计力学集体坐标阶段中膜致密的自我避免 D.R. Daniels? 英国华威大学物理系 收于2004年7月23 于2005年1月30日前可用在线摘要 我们一个提出范围涵盖很广的统计的技术理论,密集的填充有限度的过滤膜的自我回避继而接近均匀分布的阶段。我们也给出一个理论的关于小角度的分散光谱会类似这样聚集的预期，此外这还表明这可以有效地筛选出过滤膜密集的相位。这是一个定量的关于什么时候这个工作中残留的可以应用的可以被表达的理论的方法概括的讨论。在这个工作中得到的结果很可能与物质的或者是生物的过滤膜可能的新奇的稠密的阶段相关联。 关键词:统计力学;膜;自我回避 介绍 自我回避系统过滤膜是两个空间在三维空间混合的有联系的啮合嵌入，就是这样过滤膜很自然的在红细胞的细胞支架中出现。碳纤维氧化涂层聚合覆盖或者是单层出现以及交联的DNA网络这描绘过滤膜表现的统计力学更多的是人们众所周知的一个二维普遍化的单向度聚合物情况. 了解自我避免限度式过滤膜的性能。因此剩下来就研究理论上的以及实验上的两者的一个有效地区域。为了进一步说明我们引用了被给出的广泛认同的评论的读物,例如《推荐》期刊。在很多的理论上的工作在很好的溶解条件下单一的隔绝的范围薄膜的案例已经被提出。目前的研究文献提出这样一个在预定的低温的阶段和高温的褶皱的情况阶段范围薄膜被崩溃的跃迁所扭曲的的特征。这个低温的相位被称为平相拥有很大的持久性长度。它拥有很大的持久性长度以及在嵌入的体积空间中有一个自发的方向。这个高温的褶皱的相位特征在于一个有限的持久性长度以及在嵌入的空间体积内没有自发的方向。 然而, 所有先前的理论上的关于自我避免式的交互单个的孤立的限度过滤膜的推理已经被证实非常的复杂。在文献中这个分析的工作在相关的关键的范例以及全局的阶段图解中的一系列的结果也没确切表明。例如，此外,现有的数字的工作表明一个单个的孤立的限度薄膜在褶皱的相位(当然是扭曲的跃迁)是完全会消失的,以及个别人自我避免的限度式过滤膜是经常出现在延伸的相位。 在这个工作中(反对大多数以前工作的结论) 在这样宽松的特征正如稀释的条件我们将不能够处理一个在良好的溶剂中单一的孤立的自我隔离限度薄膜根据。恰恰相反, 在这工作中我们将学习到理论上的在以前大多数工作基础上的精确地反物理的限制，也就是那种吧密集的集中地大量填充的，因此大量的自我避免式的限度过滤膜也近乎同类的的系列。在这个工作中我们的假设的方法是对于相对于以前取得巨大成功的高分子聚合物产品和高分子融合物受启发的通过类似的方法。这在里我们应该将工作理解为它并不是为了让我描述我们熟知的情况一个有规律的安排好的在层叠状态下的过滤膜的堆栈。此外,显式薄膜的影响是可以忽略不计的,如典型的情况式过滤膜的自我描述。 我们将在以下的一个部分中继续进行被用到的理论大纲。
本科及研究生学术论文参考文献格式
本科及研究生学术论文参考文献格式 （Bibliography, Works Citied, References） 一.文学、翻译方向：MLA （一）说明 1. 文献类别标识 （1）常用文献类型用单字母标识 专著——[M]（monograph）；期刊文章——[J]（journal）；专著、论文集中的析出文献——[A]（article）；论文集——[C]（collected papers）；学位论文——[D]（dissertation）；专利——[P]（patent）；技术标准——[S]（standardization）；报纸——[N]（newspaper article）；科技报告——[R]（report）；其他未说明的文献类型，如字典——[Z] （2）电子文献载体类型用双字母标识 磁带——[MT]（magnetic tape）；磁盘——[DK]（disk）；光盘——[CD]（CD-ROM）；联机网络——[OL]（online）； （3）电子文献载体类型的参考文献类型标识（文献类型标识/载体类型标识） 联机网上数据库——[DB/OL]（data base online）；磁带数据库——[DB/MT]（data base on magnetic tape）；光盘图书——[M/CD]（monograph on CD-ROM）；磁盘软件——[CP/DK]（computer program on disk）；）网上期刊——[J/OL]（serial online）；网上电子公告——[EB/OL]（electronic bulletin board online） 2. 文献排序 （1）先外文，后中文。 （2）所有文献均按作者的姓（或机构名称）的字母顺序排列。
英汉对比与翻译期末论文 (2)
英汉对比与翻译——词汇三境 [摘要]：本文将英汉语言中词语分布做了对比，并就词汇翻译提出了“意译”、“义译”、“化境”三个要求，从而得知，词汇翻译的地道性很大程度上取决于语境的还原程度。 关键词: 译意、译义、臻化、词汇、语境 正文：英国语言学家George W.W ilk ins在《语言教学中的语言学》一书中指出:“没有语法,表达甚微;没有词汇,表达为零。”由此可见词汇在英语中所居的重要地位。在翻译时，即使一个译者已经掌握了大量的词汇与句法，还是会遇到总不能够突破的瓶颈，这个障碍往往是由词汇，甚至是简单词汇所引起的。例如：“He is a perfect stranger in the city.”这里对于什么是一个“perfect stranger”的理解与处理会造成译文层次的不同，同时也是考量一个译者的水平。如果仅翻译字面意思,翻译成“完美的陌生人”，译文显得晦涩难懂；当我们进一步理解“perfect”这个词的含义，并进入意义层面, “perfect”就是“完完全全”的意思，那么“perfect stranger”该译为“完完全全的陌生人”，“他是这城市中完完全全的陌生人”意思准确但是和全句契合不够完美，同“in the city”在搭配上略有滞涩；所以，我们将其转化，将“完完全全的陌生人”这一名词词组转化为动词词组，意为“对……完全陌生”，这句话就可译为“他对这城市完全陌生”，这样的搭配显得协调，不但译出了源语言想表达的意思，“疏离”这一
语境也得以完整地保留了。 以上例子就体现了英汉互译中一个很重要的规律，即词性转换。世界上的语言，不管其各有多少词类，名词和动词都占最大的比例；不管有多少句型，都能归纳为静态与动态两类。总的来说，汉语呈动态特征，英文呈静态特征。在词汇的使用上，汉语的动态表现为：汉语的动词丰富，使用频率高；而英语中，名词、介词等非动词性此类的使用频率很高。据周志培（2003:390）统计显示截选如下：（1）“邓小平在武昌、深圳、珠海、上海等地的谈话要点”及英译 语言总 数名 词 介 词 动 词 助 动 非限 定动 词 代 词 形容 词 副 词 数 词 连 词 量 词 冠 词 汉语词数636 167 11 186 27 41 70 70 39 17 8 % 26% 2% 29% 4% 7% 11% 11% 6% 3% 1% 英语词数907 238 99 105 19 30 98 104 43 28 5 84 % 26% 11% 12% 2% 3% 11% 11% 5% 3% 7% 9% （2）“英国首相撒切尔夫人在中国欢迎宴会上的讲话”及汉译文 语言总 数名词介词动词助动非限定 动词 代词形容 词 副词数 词 连词量词冠 词 汉语词数324 99 9 72 5 36 22 42 15 13 2 % 31% 3% 22% 2% 11% 7% 13% 4% 4% 0.6% 英语词数366 91 49 41 1 9 39 50 22 13 23 28 % 25% 14% 11% 0.2% 2% 10% 14% 6% 4% 6% 8% 上述结果表明，若要在英汉互译中，若要翻译出地道的T arget Text，就应当尽量符合Target Language的表达习惯。这必然包含必 要的词性转换。在进行英译汉时，根据需要将英文中的名词，介词等
英语专业毕业论文翻译类论文
英语专业毕业论文翻译 类论文 Document number：NOCG-YUNOO-BUYTT-UU986-1986UT
毕业论文（设计）Title:The Application of the Iconicity to the Translation of Chinese Poetry 题目:象似性在中国诗歌翻译中的应用 学生姓名孔令霞 学号 BC09150201 指导教师祁晓菲助教 年级 2009级英语本科（翻译方向）二班 专业英语 系别外国语言文学系
黑龙江外国语学院本科生毕业论文（设计）任务书 摘要
索绪尔提出的语言符号任意性，近些年不断受到质疑，来自语言象似性的研究是最大的挑战。语言象似性理论是针对语言任意性理论提出来的，并在不断发展。象似性是当今认知语言学研究中的一个重要课题，是指语言符号的能指与所指之间的自然联系。本文以中国诗歌英译为例，探讨象似性在中国诗歌翻译中的应用，从以下几个部分阐述：（1）象似性的发展；（2）象似性的定义及分类；（3）中国诗歌翻译的标准；（4）象似性在中国诗歌翻译中的应用，主要从以下几个方面论述：声音象似、顺序象似、数量象似、对称象似方面。通过以上几个方面的探究，探讨了中国诗歌翻译中象似性原则的重大作用，在诗歌翻译过程中有助于得到“形神皆似”和“意美、音美、形美”的理想翻译效果。 关键词：象似性；诗歌；翻译
Abstract The arbitrariness theory of language signs proposed by Saussure is severely challenged by the study of language iconicity in recent years. The theory of iconicity is put forward in contrast to that of arbitrariness and has been developing gradually. Iconicity, which is an important subject in the research of cognitive linguistics, refers to a natural resemblance or analogy between the form of a sign and the object or concept. This thesis mainly discusses the application of the iconicity to the translation of Chinese poetry. The paper is better described from the following parts: (1) The development of the iconicity; (2) The definition and classification of the iconicity; (3) The standards of the translation to Chinese poetry; (4) The application of the iconicity to the translation of Chinese poetry, mainly discussed from the following aspects: sound iconicity, order iconicity, quantity iconicity, and symmetrical iconicity. Through in-depth discussion of the above aspects, this paper could come to the conclusion that the iconicity is very important in the translation of poetry. It is conductive to reach the ideal effect of “the similarity of form and spirit” and “the three beauties”. Key words: the iconicity; poetry; translation
文献综述,外文翻译,论文网站
文献综述怎么写 1) 什么是文献综述? 文献综述是研究者在其提前阅读过某一主题的文献后，经过理解、整理、融会贯通，综合分析和评价而组成的一种不同于研究论文的文体。 2) 文献综述的写作要求 1、文献综述的格式 文献综述的格式与一般研究性论文的格式有所不同。这是因为研究性的论文注重研究的方法和结果，而文献综述介绍与主题有关的详细资料、动态、进展、展望以及对以上方面的评述。因此文献综述的格式相对多样，但总的来说，一般都包含以下四部分：即前言、主题、总结和参考文献。撰写文献综述时可按这四部分拟写提纲，再根据提纲进行撰写工作。 前言，要用简明扼要的文字说明写作的目的、必要性、有关概念的定义，综述的范围，阐述有关问题的现状和动态，以及目前对主要问题争论的焦点等。前言一般200-300字为宜，不宜超过500字。 正文，是综述的重点，写法上没有固定的格式，只要能较好地表达综合的内容，作者可创造性采用诸多形式。正文主要包括论据和论证两个部分，通过提出问题、分析问题和解决问题，比较不同学者对同一问题的看法及其理论依据，进一步阐明问题的来龙去脉和作者自己的见解。当然，作者也可从问题发生的历史背景、目前现状、发展方向等提出文献的不同观点。正文部分可根据内容的多少可分为若干个小标题分别论述。 小结，是结综述正文部分作扼要的总结，作者应对各种观点进行综合评价，提出自己的看法，指出存在的问题及今后发展的方向和展望。内容单纯的综述也可不写小结。 参考文献，是综述的重要组成部分。一般参考文献的多少可体现作者阅读文献的广度和深度。对综述类论文参考文献的数量不同杂志有不同的要求，一般以30条以内为宜，以最近3-5年内的最新文献为主。 2、文献综述规定 1. 为了使选题报告有较充分的依据，要求硕士研究生在论文开题之前作文献综述。 2. 在文献综述时，研究生应系统地查阅与自己的研究方向有关的国内外文献。通常阅读文献不少于30篇，且文献搜集要客观全面 3. 在文献综述中，研究生应说明自己研究方向的发展历史，前人的主要研究成果，存在的问题及发展趋势等。 4. 文献综述要条理清晰，文字通顺简练。 5. 资料运用恰当、合理。文献引用用方括号[ ]括起来置于引用词的右上角。 6. 文献综述中要有自己的观点和见解。不能混淆作者与文献的观点。鼓励研究生多发现问题、多提出问题、并指出分析、解决问题的可能途径，针对性强。 7. 文献综述不少于3000字。 3、注意事项 ⒈搜集文献应尽量全。掌握全面、大量的文献资料是写好综述的前提，否则，随便搜集一点资料就动手撰写是不可能写出好的综述。 ⒉注意引用文献的代表性、可靠性和科学性。在搜集到的文献中可能出现观点雷同，有的文献在可靠性及科学性方面存在着差异，因此在引用文献时应注意选用代表性、可靠性和科学性较好的文献。 ⒊引用文献要忠实文献内容。由于文献综述有作者自己的评论分析，因此在撰写时应分清作者的观点和文献的内容，不能篡改文献的内容。引用文献不过多。文献综述的作者引用间接文献的现象时有所见。如果综述作者从他人引用的参考文献转引过来，这些文献在他人
英汉互译课程论文原文(5)Translation Theory
嘉兴学院课程论文 2012—2013学年第1学期 课程名称：英汉互译 班级：姓名：学号： 要求：阅读下文，并组织4个同学将其译为汉语（请注明每位同学翻译的部分） A Short History of Western Translation Theory Introduction Antoine Berman has argued that because "reflection on translation has become an internal necessity of translation itself…(t)he construction of a history of translation is the first task of a modern theory of translation" (Berman 1992: 1). This paper gives a very brief overview of the history of western theories of translation, from the perspective of the end of the twentieth century. The framework for my discussion will be a discourse analysis approach to the history of ideas developed by the French historian Michel Foucault. Adapting Steiner (1998: 248-9), I will divide the history of discourse on translation into four periods: (1) a "traditional" period, from the beginning of the Christian era to the end of the eighteenth century, which is a period of "immediate empirical focus", (2) a period of "theory and hermeneutic inquiry", growing out of German Romanticism around the beginning of the nineteenth century, (3) a "modern" period, reaching well into the twentieth century, in which the influence of General Linguistics is increasingly dominant, and (4) the contemporary period, subsequent to the publication of Steiner"s book, which has taken to itself the name of "Translation Studies". An Approach to the History of Ideas My sources to the end of the nineteenth century will be drawn from Douglas Robinson"s Western Translation Theory(Robinson1997a). In studying the history of ideas about translation theory, we are seeking to make sense of a series of existing written texts - what they have to say about the act of translation and translated texts, and how they say what they have to say. Clearly there is a vast amount of discourse about the practice and significance of the act of translation. The 334 pages of Robinson"s book include 124 texts by 90 authors. The historical "sequel", which is the basis of my later remarks, Venuti"s Translation Studies Reader (2000), includes 30 essays by 30 authors and covers 524 pages. This mass of documents is far from being a random collection of statements. As Foucault has noted: "we know perfectly well that we are not free to say just anything, that we cannot simply speak of anything, when we like or where we like; not just anyone, finally, may speak of just anything" (Foucault 1972: 216). Specific fields of discourse are organised in regular ways and grow out of definable social contexts. In all fields of discourse, including those related to translation, "discursive practices are characterised by the delimitation of a field of objects, the definition of a legitimate perspective
学术论文参考文献格式
学术论文参考文献格式 （Bibliography, Works Citied, References） 一.文学、翻译方向：MLA （一）说明 1. 文献类别标识 （1）常用文献类型用单字母标识 专著——[M]（monograph）；期刊文章——[J]（journal）；专著、论文集中的析出文献——[A]（article）；论文集——[C]（collected papers）；学位论文——[D]（dissertation）；专利——[P]（patent）；技术标准——[S]（standardization）；报纸——[N]（newspaper article）；科技报告——[R]（report）；其他未说明的文献类型，如字典——[Z] （2）电子文献载体类型用双字母标识 磁带——[MT]（magnetic tape）；磁盘——[DK]（disk）；光盘——[CD]（CD-ROM）；联机网络——[OL]（online）； （3）电子文献载体类型的参考文献类型标识（文献类型标识/载体类型标识） 联机网上数据库——[DB/OL]（data base online）；磁带数据库——[DB/MT]（data base on magnetic tape）；光盘图书——[M/CD]（monograph on CD-ROM）；磁盘软件——[CP/DK]（computer program on disk）；）网上期刊——[J/OL]（serial online）；网上电子公告——[EB/OL]（electronic bulletin board online） 2. 文献排序 （1）先外文，后中文。 （2）所有文献均按作者的姓（或机构名称）的字母顺序排列。 3. 多加注意标点符号与字句之间的空格、标点符号类型 （二）外文著录格式 为对应起见，本格式全部规则说明中的标点符号均为Times New Roman格式。为醒目之便，这里用红色标识出来。 1. 独著 姓, 名. 书名. 出版地: 出版社, 出版年代. Adams, Hazard. Critical Theory Since Plato. Boston: Heinle & Heinle Publisher, 2004. Showalter, Elaine. A Literature of Their Own: British Women Novelists from Bronte to Lessing. London: Virago Press Ltd, 2009. 2. 两至三名作者 姓, 名, 名姓, 名姓, and 名姓. 书名. 出版地: 出版社, 出版年代. Atwan, Robert, Donald Mcquade, and John Wright. Edsels, Luckies, and Frigidaires: Advertising the American Way. New York: Dell, 1979. Horton, Rod W., and Herbert W. Edwards. Backgrounds of American Literary Thought. New York: