A query expansion framework in image retrieval domain based on local and global analysis

We present an automatic query expansion approach by generalizing the vector space model of information retrieval. In this framework, the images are presented by vectors of weighted concepts similar to the keyword-based representation in the text retrieval domain. The concepts comprise of color and texture patches from local image regions in a multi-dimensional feature space. To generate the concept vocabularies and represent the images, statistical model is built by utilizing a multi-class Support Vector Machine (SVM)-based classification technique. For automatic query expansion, the correlations between concepts are analyzed based on the neighborhood proximity between the concepts in encoded images by considering the local feedback information. The experimental results on a photographic image collection demonstrate the effectiveness of the proposed query expansion approaches.

A unified medical image retrieval framework integrating visual and text keywords using a novel multi-modal query expansion (QE) is presented. For the content-based image search, visual keywords are modeled using support vector machine (SVM)-based classification of local color and texture patches from image regions. For the text-based search, keywords from the associated annotations are extracted and indexed. The correlations between the keywords in both the visual and text feature spaces are analyzed for QE by considering local feedback information. The QE approach can propagate user perceived semantics from one modality to another and improve retrieval effectiveness when combined in multi-modal search. An evaluation of the method on imageCLEFmed'08 dataset and topics results in a mean average precision (MAP) score of 0.15 over comparable searches without QE or using only single modality.

A correlation-enhanced similarity matching framework for medical image retrieval is presented in a local concept-based feature space. In this framework, images are presented by vectors of concepts that comprise of local color and texture patches of image regions in a multi-dimensional feature space. To generate the concept vocabularies and represent the images, statistical models are built using a probabilistic multi-class support vector machine (SVM). For the similarity search, the concept correlations in the collection as a whole are analyzed as a global thesaurus-like structure and incorporated in a similarity matching function. The proposed scheme overcomes some limitations of the “bag of concepts” model, such as the assumption of feature independence. A systematic evaluation of image retrieval on a biomedical image collection of different modalities demonstrates the advantages of the proposed retrieval framework in terms of precision-recall.

Web search engines help users find relevant web pages by returning a result set containing the pages that best match the user’s query. When the identified pages have low relevance, the query must be refined to capture the search goal more effectively. However, finding appropriate refinement terms is difficult and time consuming for users, so researchers developed query expansion approaches to identify refinement terms automatically. There are two broad approaches to query expansion, automatic query expansion (AQE) and interactive query expansion (IQE) (Ruthven et al., 2003). AQE has no user involvement, which is simpler for the user, but limits its performance. IQE has user involvement, which is more complex for the user, but means it can tackle more problems such as ambiguous queries. Searches fail by finding too many irrelevant pages (low precision) or by finding too few relevant pages (low recall). AQE has a long history in the field of information retrieval, where the focus has been on improving recall (Velez et al., 1997). Unfortunately, AQE often decreased precision as the terms used to expand a query often changed the query’s meaning (Croft and Harper (1979) identified this effect and named it query drift). The problem is that users typically consider just the first few results (Jansen et al., 2005), which makes precision vital to web search performance. In contrast, IQE has historically balanced precision and recall, leading to an earlier uptake within web search. However, like AQE, the precision of IQE approaches needs improvement. Most recently, approaches have started to improve precision by incorporating semantic knowledge.

The tremendously huge volume of biomedical literature, scientists' specific information needs, long terms of multiples words, and fundamental problems ofsynonym and polysemy have been challenging issues facing the biomedical information retrieval community researchers. Search engines have significantlyimproved the efficiency and effectiveness of biomedical literature searching. The search engines, however, are known to return many results that are irrelevant to the intention of a user's query, in other words, perform not very sound in terms of precision and recall. To further improve precision and recall of biomedicalinformational retrieval, various query expansion strategies are widely used. In this thesis, we concentrate on empirical comparison, experiments and evaluations ininvestigating query expansion methods. We also use the findings as an empirical justification for cluster-based query expansion. We have investigated broadly many methods of query expansion such as local analysis, global analysis, ontology-based term reweighting across various search engines and obtained important insights. Among the findings, two-stage concept-based latent semantic analysis strategy and cluster-based query expansion have been presented and the Singular Value Decomposition (SVD) technique in the Latent Semantic Indexing (LSI) is utilized in the proposed method. In contrast to other query expansion methods, our strategy selects those terms that are most similar to the concepts of in the query as well as the related documents, rather than selects terms that are similar to the query terms only. Furthermore, we propose a novel framework for cluster-based query expansion. we have designed and implemented a novel and efficient computational approach to cluster-based query expansion using language modeling. Through our experiments in TREC genomic track ad-hoc retrieval task, we demonstrate that clusters which are created based on the whole collection or the initially returned document results of the original query can be utilized to perform query expansion and eventually improve the overall effectiveness and performance of information retrieval system in the biomedical literature retrieval. Lastly, we believe the principles of this strategy may be extended and utilized in other domains.

The objective of this paper is to provide a framework and computational model for automatic query expansion using psuedo relevance feedback. We expect that our model can be helpful in dealing with many important aspects in automatic query expansion in an efficient way. We have performed experiments based on our model using TREC data set. Results are encouraging as they indicate improvement in retrieval efficiency after applying query expansion.KeywordsAutomatic Query Expansion (AQE)Pseudo Relevance Feedback (PRF)Information Retrieval (IR)

One of the main issues associated with search engines is the query–document vocabulary mismatch problem, a long-standing problem in Information Retrieval (IR). This problem occurs when a user query does not match the content of stored documents, and it affects most search tasks. Automatic query expansion (AQE) is one of the most common approaches used to address this problem. Various AQE techniques have been proposed; these mainly involve finding synonyms or related words for the query terms. Word embedding (WE) is one of the methods that are currently receiving significant attention. Most of the existing AQE techniques focus on expanding the individual query terms rather the entire query during the expansion process, and this can lead to query drift if poor expansion terms are selected. In this article, we introduce Deep Averaging Networks (DANs), an architecture that feeds the average of the WE vectors produced by the Word2Vec toolkit for the terms in a query through several linear neural network layers. This average vector is assumed to represent the meaning of the query as a whole and can be used to find expansion terms that are relevant to the complete query. We explore the potential of DANs for AQE in Arabic document retrieval. We experiment with using DANs for AQE in the classic probabilistic BM25 model as well as for two recent expansion strategies: Embedding-Based Query Expansion approach (EQE1) and Prospect-Guided Query Expansion Strategy (V2Q). Although DANs did not improve all outcomes when used in the BM25 model, it outperformed all baselines when incorporated into the EQE1 and V2Q expansion strategies.

Structured thesauri encode equivalent, hierarchical, and associative relationships and have been developed as indexing/retrieval tools. Despite the fact that these tools provide a rich semantic network of vocabulary terms, they are seldom employed for automatic query expansion (QE) activities. This article reports on an experiment that examined whether thesaurus terms, related to query in a specified semantic way (as synonyms and partial-synonyms (SYNs), narrower terms (NTs), related terms (RTs), and broader terms (BTs)), could be identified as having a more positive impact on retrieval effectiveness when added to a query through automatic QE. The research found that automatic QE via SYNs and NTs increased relative recall with a decline in precision that was not statistically significant, and that automatic QE via RTs and BTs increased relative recall with a decline in precision that was statistically significant. Recall-based and a precision-based ranking orders for automatic QE via semantically encoded thesauri terminology were identified. Mapping results found between end-user query terms and the ProQuest® Controlled Vocabulary (1997) (the thesaurus used in this study) are reported, and future research foci related to the investigation are discussed.

This paper presents a medical image retrieval framework that uses visual concepts in a feature space employing statistical models built using a probabilistic multi-class support vector machine (SVM). The images are represented using concepts that comprise color and texture patches from local image regions in a multi-dimensional feature space. A major limitation of concept feature representation is that the structural relationship or spatial ordering between concepts are ignored. We present a feature representation scheme as visual concept structure descriptor (VCSD) that overcomes this challenge and captures both the concept frequency similar to a color histogram and the local spatial relationships of the concepts. A probabilistic framework makes the descriptor robust against classification and quantization errors. Evaluation of the proposed image retrieval framework on a biomedical image dataset with different imaging modalities validates its benefits.

PurposeQuery expansion and query limitation are two known techniques for assisting users to define efficient queries. The purpose of this article is to examine the effectiveness of the two methods.Design/methodology/approachThe research entailed an objective and subjective evaluation of the effectiveness of automatic and interactive query expansion and of two query limit options. The evaluation included both lab simulations and large‐scale user studies. The objective aspects were evaluated in lab simulations with experts judging user performance. The subjective analysis was carried out by having the participants evaluate the quality of, and express their satisfaction with, the retrieval process and its results, thus employing perceived‐value analysis.FindingsThe main findings reveal a difference between the perceived and real values of these techniques. While users expressed their satisfaction with interactive query expansion and its performance, the real‐value analysis of their performance did not show any significant difference between the retrieval modes.Originality/valueThe article evaluates the objective and subjective effectiveness of automatic and interactive query expansion and two query limit options.

Interactive searching and interface issues in the Okapi best match probabilistic retrieval system

Content Based Image Retrieval (CBIR) applies computer vision methods for image retreival purposes from the databases. It is majorly based on the user query, which is in visual form rather than the traditional text form. CBIR is applied in different fields extending from surveillance to remote sensing, E-purchase, medical image processing, security systems to historical research and many others. JPEG, a very commonly used method of lossy compression is used to reduce the size of the image before being stored or transmitted. Almost every digital camera in the market are storing the captured images in jpeg format. The storage industry has seen many major transformations in the past decades while the retrieval technologies are still developing. Though there are some breakthroughs happened in text retrieval, the same is not true for the image and other multimedia retrieval. Specifically image retreival has witnessed many algorithms in the spatial or the raw domain but since majority of the images are stored in the JPEG format, it takes time to decode the compressed image before extracting features and retrieving. Hence, in this research work, we focus on extracting the features from the compressed domain itself and then utilize support vector machines (SVM) for improving the retrieval results. Our proof of concept shows us that the features extracted in compressed domain helps retrieve the images 43% faster than the same set of images in the spatial domain and the accuracy is improved to 93.4% through SVM based feedback mechanism.

This paper reports our participation in the INEX 2008 Ad-Hoc Retrieval track. We investigated the effect of multiword terms on retrieval effectiveness in an interactive query expansion (IQE) framework. The IQE approach is compared to a state-of-the-art IR engine (in this case Indri) implementing a bag-of-word query and document representation, coupled with pseudo-relevance feedback (automatic query expansion(AQE)). The performance of multiword query and document representation was enhanced when the term structure was relaxed to accept the insertion of additional words while preserving the original structure and word order. The search strategies built with multiword terms coupled with QE obtained very competitive scores in the three Ad-Hoc tasks: Focused retrieval, Relevant-in-Context and Best-in-Context.

This paper focuses on the user and human‐computer interaction (HCI) aspects of the research based on the Okapi text retrieval system. Three experiments implementing different approaches to query expansion are described, highlighting the close relationship between the system’s functionality and different interface designs. The projects evaluated the retrieval effectiveness and usability of an automatic query expansion facility in a VT100 character‐based interface, and two different forms of interactive query expansion implemented in graphical user interface (GUI) environments with different windowing techniques. The experimental conditions, variables and constraints in undertaking operational user testing are discussed in relation to the interface features, and in terms of: the visibility of the system’s operations; the system/user control; and the cognitive load on the user. It is suggested that the quality and effectiveness of the search interaction for query expansion is dependent on resolving the tension between seemingly opposing interface and functional aspects, e.g. automatic vs interactive query expansion, explicit vs implicit use of a thesaurus, document vs query space.

The paper describes a user study examining methods for improving users queries, specifically interactive and automatic query expansion and advanced search options. The user study includes subjective and objective evaluation of the effect of the above methods and a comparison between the real and perceived effect.