Reliability and effectiveness of clickthrough data for automatic image annotation

Automatic image annotation using supervised learning is performed by concept classifiers trained on labelled example images. This work proposes the use of clickthrough data collected from search logs as a source for the automatic generation of concept training data, thus avoiding the expensive manual annotation effort. We investigate and evaluate this approach using a collection of 97,628 photographic images. The results indicate that the contribution of search log based training data is positive despite their inherent noise; in particular, the combination of manual and automatically generated training data outperforms the use of manual data alone. It is therefore possible to use clickthrough data to perform large-scale image annotation with little manual annotation effort or, depending on performance, using only the automatically generated training data. An extensive presentation of the experimental results and the accompanying data can be accessed at .     

Utilizing similarity relationships among existing data for high accuracy processing of content-based image retrieval

Retrieving similar images based on its visual content is an important yet difficult problem. We propose in this paper a new method to improve the accuracy of content-based image retrieval systems. Typically, given a query image, existing retrieval methods return a ranked list based on the similarity scores between the query and individual images in the database. Our method goes further by relying on an analysis of the underlying connections among individual images in the database to improve this list. Initially, we consider each image in the database as a query and use an existing baseline method to search for its likely similar images. Then, the database is modeled as a graph where images are nodes and connections among possibly similar images are edges. Next, we introduce an algorithm to split this graph into stronger subgraphs, based on our notion of graph’s strength, so that images in each subgraph are expected to be truly similar to each other. We create for each subgraph a structure called integrated image which contains the visual features of all images in the subgraph. At query time, we compute the similarity scores not only between the query and individual database images but also between the query and the integrated images. The final similarity score of a database image is computed based on both its individual score and the score of the integrated image that it belongs to. This leads effectively to a re-ranking of the retrieved images. We evaluate our method on a common image retrieval benchmark and demonstrate a significant improvement over the traditional bag-of-words retrieval model.     

Featured correspondence topic model for semantic search on social image collections

Nowadays, due to the rapid growth of digital technologies, huge volumes of image data are created and shared on social media sites. User-provided tags attached to each social image are widely recognized as a bridge to fill the semantic gap between low-level image features and high-level concepts. Hence, a combination of images along with their corresponding tags is useful for intelligent retrieval systems, those are designed to gain high-level understanding from images and facilitate semantic search. However, user-provided tags in practice are usually incomplete and noisy, which may degrade the retrieval performance. To tackle this problem, we present a novel retrieval framework that automatically associates the visual content with textual tags and enables effective image search. To this end, we first propose a probabilistic topic model learned on social images to discover latent topics from the co-occurrence of tags and image features. Moreover, our topic model is built by exploiting the expert knowledge about the correlation between tags with visual contents and the relationship among image features that is formulated in terms of spatial location and color distribution. The discovered topics then help to predict missing tags of an unseen image as well as the ones partially labeled in the database. These predicted tags can greatly facilitate the reliable measure of semantic similarity between the query and database images. Therefore, we further present a scoring scheme to estimate the similarity by fusing textual tags and visual representation. Extensive experiments conducted on three benchmark datasets show that our topic model provides the accurate annotation against the noise and incompleteness of tags. Using our generalized scoring scheme, which is particularly advantageous to many types of queries, the proposed approach also outperforms state-of-the-art approaches in terms of retrieval accuracy.     

融合多类特征的Web查询意图识别

识别搜索引擎用户的查询意图在信息检索领域是备受关注的研究内容。文中提出一种融合多类特征识别Web查询意图的方法。将Web查询意图识别作为一个分类问题,并从不同类型的资源包括查询文本、搜索引擎返回内容及Web查询日志中抽取出有效的分类特征。在人工标注的真实Web查询语料上采用文中方法进行查询意图识别实验,实验结果显示文中采用的各类特征对于提高查询意图识别的效果皆有一定帮助,综合使用这些特征进行查询意图识别,88。5%的测试查询获得准确的意图识别结果。     

Impact of query intent and search context on clickthrough behavior in sponsored search

Implicit feedback techniques may be used for query intent detection, taking advantage of user behavior to understand their interests and preferences. In sponsored search, a primary concern is the user’s interest in purchasing or utilizing a commercial service, or what is called online commercial intent. In this paper, we develop a methodology for employing the content of search engine result pages (SERPs), along with the information obtained from query strings, to study characteristics of query intent, with a particular focus on sponsored search. Our work represents a step toward the development and evaluation of an ontology for commercial search, considering queries that reference specific products, brands, and retailers. Characteristics of query categories are studied with respect to aggregated user clickthrough behavior on advertising links. We present a model for clickthrough behavior that considers the influence of such factors as the location of ads and the rank of ads, along with query category. We evaluate our work using a large corpus of clickthrough data obtained from a major commercial search engine. In addition, the impact of query intent is studied on clickthrough rate, where a baseline model and the query intent model are compared for the purpose of calculating an expected ad clickthrough rate. Our findings suggest that query-based features, along with the content of SERPs, are effective in detecting query intent. Factors such as query category, the rank of an ad, and the total number of ads displayed on a result page relate to the context of the ad, rather than its content. We demonstrate that these context-related factors can have a major influence on expected clickthrough rate, suggesting that these factors should be taken into consideration when the performance of an ad is evaluated.     

The effects of multiple query evidences on social image retrieval

System performance assessment and comparison are fundamental for large-scale image search engine development. This article documents a set of comprehensive empirical studies to explore the effects of multiple query evidences on large-scale social image search. The search performance based on the social tags, different kinds of visual features and their combinations are systematically studied and analyzed. To quantify the visual query complexity, a novel quantitative metric is proposed and applied to assess the influences of different visual queries based on their complexity levels. Besides, we also study the effects of automatic text query expansion with social tags using a pseudo relevance feedback method on the retrieval performance. Our analysis of experimental results shows a few key research findings: (1) social tag-based retrieval methods can achieve much better results than content-based retrieval methods; (2) a combination of textual and visual features can significantly and consistently improve the search performance; (3) the complexity of image queries has a strong correlation with retrieval results’ quality—more complex queries lead to poorer search effectiveness; and (4) query expansion based on social tags frequently causes search topic drift and consequently leads to performance degradation.     

基于模糊集的主题提取和层次发现算法

从语义相关性角度分析超链归纳主题搜索(HITS)算法,发现其产生主题漂移的原因在于页面被投影到错误的语义基上,提出了一种基于模糊集的主题提取和层次发现算法(FSTH),通过用户日志扩展查询词,构造符合用户需要的个性化根集和基础集合,达到防止主题漂移的目的。FSTH采用模糊集划分方法,层次地发现与用户查询相关的主题页面集合,利用HITS算法分别计算每个主题页面集合中页面的权威值,返回与查询相关的其他主题权威页面。在14个查询上的实验结果表明,与HITS算法相比,FSTH算法不仅可以减少7%~53%的主题漂移率,而且可以发现与查询相关的多个主题.     

应用多索引加法量化编码的近邻检索算法

目的 海量图像检索技术是计算机视觉领域研究热点之一,一个基本的思路是对数据库中所有图像提取特征,然后定义特征相似性度量,进行近邻检索。海量图像检索技术,关键的是设计满足存储需求和效率的近邻检索算法。为了提高图像视觉特征的近似表示精度和降低图像视觉特征的存储空间需求,提出了一种多索引加法量化方法。方法 由于线性搜索算法复杂度高,而且为了满足检索的实时性,需把图像描述符存储在内存中,不能满足大规模检索系统的需求。基于非线性检索的优越性,本文对非穷尽搜索的多索引结构和量化编码进行了探索新研究。利用多索引结构将原始数据空间划分成多个子空间,把每个子空间数据项分配到不同的倒排列表中,然后使用压缩编码的加法量化方法编码倒排列表中的残差数据项,进一步减少对原始空间的量化损失。在近邻检索时采用非穷尽搜索的策略,只在少数倒排列表中检索近邻项,可以大大减少检索时间成本,而且检索过程中不用存储原始数据,只需存储数据集中每个数据项在加法量化码书中的码字索引,大大减少内存消耗。结果 为了验证算法的有效性,在3个数据集SIFT、GIST、MNIST上进行测试,召回率相比近几年算法提升4%~15%,平均查准率提高12%左右,检索时间与最快的算法持平。结论 本文提出的多索引加法量化编码算法,有效改善了图像视觉特征的近似表示精度和存储空间需求,并提升了在大规模数据集的检索准确率和召回率。本文算法主要针对特征进行近邻检索,适用于海量图像以及其他多媒体数据的近邻检索。     

Complementary relevance feedback-based content-based image retrieval

We propose a complementary relevance feedback-based content-based image retrieval (CBIR) system. This system exploits the synergism between short-term and long-term learning techniques to improve the retrieval performance. Specifically, we construct an adaptive semantic repository in long-term learning to store retrieval patterns of historical query sessions. We then extract high-level semantic features from the semantic repository and seamlessly integrate low-level visual features and high-level semantic features in short-term learning to effectively represent the query in a single retrieval session. The high-level semantic features are dynamically updated based on users’ query concept and therefore represent the image’s semantic concept more accurately. Our extensive experimental results demonstrate that the proposed system outperforms its seven state-of-the-art peer systems in terms of retrieval precision and storage space on a large scale imagery database.     

Mobile multi-view object image search

High user interaction capability of mobile devices can help improve the accuracy of mobile visual search systems. At query time, it is possible to capture multiple views of an object from different viewing angles and at different scales with the mobile device camera to obtain richer information about the object compared to a single view and hence return more accurate results. Motivated by this, we propose a new multi-view visual query model on multi-view object image databases for mobile visual search. Multi-view images of objects acquired by the mobile clients are processed and local features are sent to a server, which combines the query image representations with early/late fusion methods and returns the query results. We performed a comprehensive analysis of early and late fusion approaches using various similarity functions, on an existing single view and a new multi-view object image database. The experimental results show that multi-view search provides significantly better retrieval accuracy compared to traditional single view search.     

Discovering and understanding word level user intent in Web search queries

Identifying and interpreting user intent are fundamental to semantic search. In this paper, we investigate the association of intent with individual words of a search query. We propose that words in queries can be classified as either content or intent, where content words represent the central topic of the query, while users add intent words to make their requirements more explicit. We argue that intelligent processing of intent words can be vital to improving the result quality, and in this work we focus on intent word discovery and understanding. Our approach towards intent word detection is motivated by the hypotheses that query intent words satisfy certain distributional properties in large query logs similar to function words in natural language corpora. Following this idea, we first prove the effectiveness of our corpus distributional features, namely, word co-occurrence counts and entropies, towards function word detection for five natural languages. Next, we show that reliable detection of intent words in queries is possible using these same features computed from query logs. To make the distinction between content and intent words more tangible, we additionally provide operational definitions of content and intent words as those words that should match, and those that need not match, respectively, in the text of relevant documents. In addition to a standard evaluation against human annotations, we also provide an alternative validation of our ideas using clickthrough data. Concordance of the two orthogonal evaluation approaches provide further support to our original hypothesis of the existence of two distinct word classes in search queries. Finally, we provide a taxonomy of intent words derived through rigorous manual analysis of large query logs.     

Simulated evaluation of faceted browsing based on feature selection

In this paper we explore the limitations of facet based browsing which uses sub-needs of an information need for querying and organising the search process in video retrieval. The underlying assumption of this approach is that the search effectiveness will be enhanced if such an approach is employed for interactive video retrieval using textual and visual features. We explore the performance bounds of a faceted system by carrying out a simulated user evaluation on TRECVid data sets, and also on the logs of a prior user experiment with the system. We first present a methodology to reduce the dimensionality of features by selecting the most important ones. Then, we discuss the simulated evaluation strategies employed in our evaluation and the effect on the use of both textual and visual features. Facets created by users are simulated by clustering video shots using textual and visual features. The experimental results of our study demonstrate that the faceted browser can potentially improve the search effectiveness.     

一种基于稀疏典型性相关分析的图像检索方法

图像语义检索的一个关键问题就是要找到图像底层特征与语义之间的关联,由于文本是表达语义的一种有效手段,因此提出通过研究文本与图像两种模态之间关系来构建反映两者间潜在语义关联的有效模型的思路,基于该模型,可使用自然语言形式(文本语句)来表达检索意图,最终检索到相关图像.该模型基于稀疏典型性相关分析(sparse canonical correlation analysis,简称sparse CCA),按照如下步骤训练得到:首先利用隐语义分析方法构造文本语义空间,然后以视觉词袋(bag of visual words)来表达文本所对应的图像,最后通过Sparse CCA算法找到一个语义相关空间,以实现文本语义与图像视觉单词间的映射.使用稀疏的相关性分析方法可以提高模型可解释性和保证检索结果稳定性.实验结果验证了Sparse CCA方法的有效性,同时也证实了所提出的图像语义检索方法的可行性.     

Relevance Feedback and Learning in Content-Based Image Search

A major bottleneck in content-based image retrieval (CBIR) systems or search engines is the large gap between low-level image features used to index images and high-level semantic contents of images. One solution to this bottleneck is to apply relevance feedback to refine the query or similarity measures in image search process. In this paper, we first address the key issues involved in relevance feedback of CBIR systems and present a brief overview of a set of commonly used relevance feedback algorithms. Almost all of the previously proposed methods fall well into such framework. We present a framework of relevance feedback and semantic learning in CBIR. In this framework, low-level features and keyword annotations are integrated in image retrieval and in feedback processes to improve the retrieval performance. We have also extended framework to a content-based web image search engine in which hosting web pages are used to collect relevant annotations for images and users' feedback logs are used to refine annotations. A prototype system has developed to evaluate our proposed schemes, and our experimental results indicated that our approach outperforms traditional CBIR system and relevance feedback approaches.     

A Multi-View Embedding Space for Modeling Internet Images,Tags, and Their Semantics

This paper investigates the problem of modeling Internet images and associated text or tags for tasks such as image-to-image search, tag-to-image search, and image-to-tag search (image annotation). We start with canonical correlation analysis (CCA), a popular and successful approach for mapping visual and textual features to the same latent space, and incorporate a third view capturing high-level image semantics, represented either by a single category or multiple non-mutually-exclusive concepts. We present two ways to train the three-view embedding: supervised, with the third view coming from ground-truth labels or search keywords; and unsupervised, with semantic themes automatically obtained by clustering the tags. To ensure high accuracy for retrieval tasks while keeping the learning process scalable, we combine multiple strong visual features and use explicit nonlinear kernel mappings to efficiently approximate kernel CCA. To perform retrieval, we use a specially designed similarity function in the embedded space, which substantially outperforms the Euclidean distance. The resulting system produces compelling qualitative results and outperforms a number of two-view baselines on retrieval tasks on three large-scale Internet image datasets.     

Narrowing the semantic gap - improved text-based web document retrieval using visual features

We present the results of our work that seek to negotiate the gap between low-level features and high-level concepts in the domain of web document retrieval. This work concerns a technique, called the latent semantic indexing (LSI), which has been used for textual information retrieval for many years. In this environment, LSI determines clusters of co-occurring keywords so that a query which uses a particular keyword can then retrieve documents perhaps not containing this keyword, but containing other keywords from the same cluster. In this paper, we examine the use of this technique for content-based web document retrieval, using both keywords and image features to represent the documents. Two different approaches to image feature representation, namely, color histograms and color anglograms, are adopted and evaluated. Experimental results show that LSI, together with both textual and visual features, is able to extract the underlying semantic structure of web documents, thus helping to improve the retrieval performance significantly, even when querying is done using only keywords.     

Learning non-metric visual similarity for image retrieval

Measuring visual similarity between two or more instances within a data distribution is a fundamental task in image retrieval. Theoretically, non-metric distances are able to generate a more complex and accurate similarity model than metric distances, provided that the non-linear data distribution is precisely captured by the system. In this work, we explore neural networks models for learning a non-metric similarity function for instance search. We argue that non-metric similarity functions based on neural networks can build a better model of human visual perception than standard metric distances. As our proposed similarity function is differentiable, we explore a real end-to-end trainable approach for image retrieval, i.e. we learn the weights from the input image pixels to the final similarity score. Experimental evaluation shows that non-metric similarity networks are able to learn visual similarities between images and improve performance on top of state-of-the-art image representations, boosting results in standard image retrieval datasets with respect standard metric distances.     

Genetic algorithm for text clustering based on latent semantic indexing

In this paper, we develop a genetic algorithm method based on a latent semantic model (GAL) for text clustering. The main difficulty in the application of genetic algorithms (GAs) for document clustering is thousands or even tens of thousands of dimensions in feature space which is typical for textual data. Because the most straightforward and popular approach represents texts with the vector space model (VSM), that is, each unique term in the vocabulary represents one dimension. Latent semantic indexing (LSI) is a successful technology in information retrieval which attempts to explore the latent semantics implied by a query or a document through representing them in a dimension-reduced space. Meanwhile, LSI takes into account the effects of synonymy and polysemy, which constructs a semantic structure in textual data. GA belongs to search techniques that can efficiently evolve the optimal solution in the reduced space. We propose a variable string length genetic algorithm which has been exploited for automatically evolving the proper number of clusters as well as providing near optimal data set clustering. GA can be used in conjunction with the reduced latent semantic structure and improve clustering efficiency and accuracy. The superiority of GAL approach over conventional GA applied in VSM model is demonstrated by providing good Reuter document clustering results.     

一种基于视觉词典优化和查询扩展的图像检索方法

视觉词典方法（Bag of visual words,BoVW）是当前图像检索领域的主流方法,然而,传统的视觉词典方法存在计算量大、词典区分性不强以及抗干扰能力差等问题,难以适应大数据环境.针对这些问题,本文提出了一种基于视觉词典优化和查询扩展的图像检索方法.首先,利用基于密度的聚类方法对SIFT特征进行聚类生成视觉词典,提高视觉词典的生成效率和质量;然后,通过卡方模型分析视觉单词与图像目标的相关性,去除不包含目标信息的视觉单词,增强视觉词典的分辨能力;最后,采用基于图结构的查询扩展方法对初始检索结果进行重排序.在Oxford5K和Paris6K图像集上的实验结果表明,新方法在一定程度上提高了视觉词典的质量和语义分辨能力,性能优于当前主流方法.