Clustering Recall Research Articles

The effects of immediate recall and subsequent retrieval strategy on eyewitness memory

The current study investigated the effects of immediate recall and subsequent interview on mock witnesses’ memory of a crime. After watching a video of a crime, 111 participants were randomly allocated to an immediate recall condition (the iWitnessed mobile app, free recall). One week later, participants were randomly allocated to a retrieval condition (timeline aid, category clustering recall, free recall). Results showed that iWitnessed elicited more correct details than free recall at immediate recall (Time 1), without compromising accuracy. However, the immediate recall tool did not affect memory performance at a one-week delay (Time 2). At Time 2, participants with retrieval support (timeline aid, category clustering recall) reported more details than those with no retrieval support (free recall), without compromising accuracy. The findings highlight that high-quality immediate recall and retrieval support during a subsequent interview are necessary to elicit more complete and accurate accounts.

Public Housing Allocation Model in the Guangdong-Hong Kong-Macao Greater Bay Area under Clustering Algorithm

In the Guangdong-Hong Kong-Macao Greater Bay Area (Bay Area), the allocation methods of public rental housing are analyzed to achieve scientific and fair housing allocation as much as possible, so as to protect the housing demand of low-income and middle-income families. The housing model in the Bay Area is analyzed firstly, and the key points of public rental housing and allocation management models are discussed comprehensively. Furthermore, a method based on rough-based fuzzy clustering (RFC) is proposed to analyze the housing demands of security groups, and a public housing allocation model is constructed based on actual demand of residents. The housing allocation plan is given and decided by the decision-making department based on the demand of the security objects and the characteristics of public housing. The simulation experiments are performed on the clustering algorithm optimized based on rough set feature selection. On the Chess data set, the optimized clustering algorithm shows an obvious improvement in clustering accuracy and recall rate compared with the traditional clustering algorithms, which are 0.76 and 0.95, respectively. The bilateral matching method based on fuzzy axiom design can fully consider the actual needs of both the supply and demand of the housing security, which is beneficial to improve the rationality and correctness of public housing allocation. The allocation method of public housing based on demand clustering analysis focuses on improving the housing security level and strives to meet the higher-level housing improvement needs of housing security objects, so as to provide security objects with more expected living conditions and improve housing allocation effect.

Testing two retrieval strategies to enhance eyewitness memory: Category and location clustering recall.

Asking eyewitnesses to provide an initial account using free recall strategies is a crucial procedure used in most investigative interviews (e.g. Cognitive Interview or PEACE model) allowing eyewitnesses to recall accurate and detailed information about the crime. Recent studies suggest clustering retrieval strategies like Category Clustering Recall (CCR) can also be effective for obtaining accurate information from eyewitnesses, and perhaps more effective than the change order mnemonic, a witness-compatible questioning, or a free recall task. This study compared how effective three retrieval strategies were (free recall, CCR, and a new retrieval strategy: Location Clustering Recall - LCR) for obtaining and initial account from eyewitnesses when used as the sole form of retrieval but combined with compatible preliminary instructions, mnemonics, and procedures that constitute best practice. We employed a between-subjects experimental design where 69 participants watched a mock robbery video and were interviewed 48 hours later with either a free recall, CCR, or LCR. Participants interviewed with CCR or LCR recalled a higher number of correct details, particularly person-related details. Participants interviewed with LCR also recalled a higher number of environmental details. Report accuracy was high, and similar, for all groups. This study provides further evidence of the ecological validity of CCR as well as initial evidence suggesting LCR might be an effective retrieval strategy as well. These retrieval strategies (CCR and LCR) might be particularly advantageous in situations where there are multiple crime-scene locations and/or the interviewer is particularly interested in obtaining information concerning environmental and person-related details.

Testing a modified cognitive interview with category clustering recall in Iran

SummaryThe cognitive interview (CI) has been an effective method for interviewing eyewitnesses often leading to changes in legislation and practice in many countries. This study was the first to employ the CI in Iran and test whether category clustering recall (CCR) was superior to a free recall when incorporated within an investigative interview. A between‐subjects design assigned 66 participants to one of three interview conditions after they watched a mock robbery. The participants were interviewed 48 hr later using either a structured interview (SI), the CI, or a modified cognitive interview (MCI) that replaced free recall with CCR at the first retrieval attempt. Analysis of variance suggests CCR was more effective than free recall and the CI group recalled more information than the SI group, replicating the CI superiority effect. This has implications for law enforcement in Iran and worldwide by suggesting these techniques can be used to enhance recall.

Discovering event episodes from sequences of online news articles: A time-adjoining frequent itemset-based clustering method

Firms perform environmental surveillance to identify important events and their developments. To alleviate the stringent information processing and analysis requirements, automated methods are needed to discover from online news articles distinct episodes (stages) of an important event. We propose a time-adjoining frequent itemset-based method that incorporates essential temporal characteristics of news articles for event episode discovery. With a corpus of 1468 news articles that pertain to 248 episodes of 53 different events, we empirically evaluate the proposed method and include several prevalent techniques as benchmarks. The results show that our method outperforms the benchmark techniques consistently and significantly, attaining the cluster recall, cluster precision, and F-measure values at 0.706, 0.593, and 0.584, respectively.

DISC: Disambiguating homonyms using graph structural clustering

Author name ambiguity degrades information retrieval, database integration, search results and, more importantly, correct attributions in bibliographic databases. Some unresolved issues include how to ascertain the actual number of authors, how to improve the performance and how to make the method more effective in terms of representative clustering metrics (average cluster purity, average author purity, K-metric, pairwise precision, pairwise recall, pairwise-F1, cluster precision, cluster recall and cluster-F1). It is a non-trivial task to disambiguate authors using only the implicit bibliographic information. An effective method ‘DISC’ is proposed that uses graph community detection algorithm, feature vectors and graph operations to disambiguate homonyms. The citation data set is pre-processed and ambiguous author blocks are formed. A co-authors graph is constructed using authors and their co-author’s relationships. A graph structural clustering ‘gSkeletonClu’ is applied to identify hubs, outliers and clusters of nodes in a co-author’s graph. Homonyms are resolved by splitting these clusters of nodes across the hub if their feature vector similarity is less than a predefined threshold. DISC utilises only co-authors and titles that are available in almost all bibliographic databases. With little modifications, DISC can also be used for entity disambiguation. To validate the DISC performance, experiments are performed on two Arnetminer data sets and compared with five previous unsupervised methods. Despite using limited bibliographic metadata, DISC achieves on average K-metric, pairwise-F1, and cluster-F1 of 92%, 84% and 74%, respectively, using Arnetminer-S and 86%, 80% and 57%, respectively, using Arnetminer-L. About 77.5% and 73.2% clusters are within the range (ground truth clusters ± 3) in Arnetminer-S and Arnetminer-L, respectively.

Enhancing individual and collaborative eyewitness memory with category clustering recall

ABSTRACTMost crimes have multiple eyewitnesses. The police typically interview co-witnesses separately. In time-sensitive investigations, this could slow down evidence accumulation. Having co-witnesses collaboratively recall a crime could potentially expedite evidence accumulation. However, past research shows that collaborative group members often have conflicting retrieval strategies that disrupt each other, degrading overall recall. This cost could potentially be overcome by aligning group members’ retrieval strategies with category clustering recall (CCR), which is a retrieval strategy where information is recalled from a series of forensically relevant categories (e.g., recalling the protagonists’ appearance, then actions). This study examined the costs and benefits of collaborative eyewitness memory by having collaborative pairs of strangers, nominal pairs (i.e., two individuals whose recall is pooled) and lone individuals watch a crime and recall it using free recall or CCR. The collaborative pairs recalled the crime faster than the nominal pairs. They also recalled more correct information than individuals but less than nominal pairs, irrespective of the retrieval method. There is therefore a speed-recall completeness trade-off when collaborative groups recall crimes. Importantly, all participants recalled more correct information when using CCR. This provides initial evidence suggesting that CCR is superior to free recall. Further research examining CCR’s benefits is recommended.

Enhancing the cognitive interview with an alternative procedure to witness-compatible questioning: category clustering recall

ABSTRACTThe Cognitive Interview (CI) is one of the most widely studied and used methods to interview witnesses. However, new component techniques for further increasing correct recall are still crucial. We focused on how a new and simpler interview strategy, Category Clustering Recall (CCR), could increase recall in comparison with witness-compatible questioning and tested if a Revised Cognitive Interview (RCI) with CCR instead of witness-compatible questioning and without the change order and change perspective mnemonics would be effective for this purpose. Participants watched a mock robbery video and were interviewed 48 hours later with either the CI or the RCI. Recalled information was classified as either correct, incorrect or confabulation. Although exclusion of the change order and change perspective mnemonics in the RCI group might have caused a slight decrease in recall during the last interview phases, the RCI group generally produced more correct information than the CI group, with a lower number of confabulations. Further analyses revealed CCR was largely responsible for this increase in correct recall. CCR is a very promising interview technique which allowed the interviewer to obtain more detailed information without additional questions and may have, in certain situations, several practical advantages over a questioning phase.

Clustering Strategy in Intellectually Gifted Children

This study examined three aspects of the clustering strategy used by participants: the differences of clustering strategy between intellectually gifted and average children; the relationship between clustering strategy and recall performance in intellectually gifted and average children; and the differences in recall performance on collaborative recall task of intellectually gifted and average children. Thirty-eight intellectually gifted and 44 intellectually average children were recruited, and their collaborative and individual recall performances were tested. The results showed that intellectually gifted children had better clustering strategy than their peers when adjusted ratio of clustering scores were calculated ( p < .01); clustering strategy positively correlated with recall performance in intellectually average children ( p < .05), while no such correlation was found in intellectually gifted children ( p = .95); there was no collaborative inhibition in the gifted group ( p = .82). These findings partly resolve the conflicts in individual research and expand on the conclusions regarding collaborative memory.

Chinese Person Name Disambiguation Based on Two-Stage Clustering

Person name clustering disambiguation is the process that partitions name mentions according to corresponding target person entities in reality. The existed methods can not realize effective identification of important features to disambiguate person names. This paper presents a method of Chinese person name disambiguation based on two-stage clustering. This method adopts a stage-by-stage processing model to identify and utilize different types of important features. Firstly, we extract three kinds of core evidences namely direct social relation, indirect social relation and common description prefix, recognize document-pairs referring to the same person entity, and realize initial clustering of person names with high precision. Then, we take the result of initial clustering as new initial input, utilize the statistical properties of multi-documents to recognize and evaluate important features, and build a double-vector representation of clusters (cluster feature vector and important feature vector). Based on the processes above, the final clustering of person names is generated, and the recall of clustering is improved effectively. The experiments have been conducted on the dataset of CLP2010 Chinese person names disambiguation, and experimental results show that this method has good performance in person name clustering disambiguation.

Improving the Enhanced Cognitive Interview With a New Interview Strategy: Category Clustering Recall

SummaryIncreasing recall is crucial for investigative interviews. The enhanced cognitive interview (ECI) has been widely used for this purpose and found to be generally effective. We focused on further increasing recall with a new interview strategy, category clustering recall (CCR). Participants watched a mock robbery video and were interviewed 48 hours later with either the (i) ECI; (ii) revised enhanced cognitive interview 1 (RECI1) — with CCR instead of the change order mnemonic during the second recall; or (iii) revised enhanced cognitive interview 2 (RECI2) — also with CCR but conjunctly used with ‘eye closure’ and additional open‐ended follow up questions. Participants interviewed with CCR (RECI1 and RECI2) produced more information without compromising accuracy; thus, CCR was effective. Eye closure and additional open‐ended follow up questions did not further influence recall when using CCR. Major implications for real‐life investigations are discussed.Copyright © 2016 John Wiley & Sons, Ltd.

Result diversification in social image retrieval: a benchmarking framework

This article addresses the diversification of image retrieval results in the context of image retrieval from social media. It proposes a benchmarking framework together with an annotated dataset and discusses the results achieved during the related task run in the MediaEval 2013 benchmark. 38 multimedia diversification systems, varying from graph-based representations, re-ranking, optimization approaches, data clustering to hybrid approaches that included a human in the loop, and their results are described and analyzed in this text. A comparison of the use of expert vs. crowdsourcing annotations shows that crowdsourcing results have a slightly lower inter-rater agreement but results are comparable at a much lower cost than expert annotators. Multimodal approaches have best results in terms of cluster recall. Manual approaches can lead to high precision but often lower diversity. With this detailed results analysis we give future insights into diversity in image retrieval and also for preparing new evaluation campaigns in related areas.

Research on Concept Clustering of Order Task Ontology based on Modified PART

Clustering is the foundation of establishing the semantic model of order task ontology. The process of concept clustering of order manufacturing task was analyzed to solve the clustering problem. A three-index evaluation method of order task document was proposed to construct the corpus. And then SVD approach was employed to reduce the term-document matrix. On the basis of this, a modified PART clustering algorithm was proposed to reestablish similarity check function by integrating of latent semantic information of ontology concept to improve the clustering accuracy. Finally, the task documents in a building materials equipment enterprise were collected as sample data and the experiment was simulated by matlab. By comparison with ART and PART, the modified PART had better performances in recall and accuracy of concept clustering.

Clustering-based topical Web crawling using CFu-tree guided by link-context

Topical Web crawling is an established technique for domain-specific information retrieval. However, almost all the conventional topical Web crawlers focus on building crawlers using different classifiers, which needs a lot of labeled training data that is very difficult to labelmanually. This paper presents a novel approach called clustering-based topical Web crawling which is utilized to retrieve information on a specific domain based on link-context and does not require any labeled training data. In order to collect domain-specific content units, a novel hierarchical clustering method called bottom-up approach is used to illustrate the process of clustering where a new data structure, a linked list in combination with CFu-tree, is implemented to store cluster label, feature vector and content unit. During clustering, four metrics are presented. First, comparison variation (CV) is defined to judge whether the closest pair of clusters can be merged. Second, cluster impurity (CIP) evaluates the cluster error. Then, the precision and recall of clustering are also presented to evaluate the accuracy and comprehensive degree of the whole clustering process. Link-context extraction technique is used to expand the feature vector of anchor text which improves the clustering accuracy greatly. Experimental results show that the performance of our proposed method overcomes conventional focused Web crawlers both in Harvest rate and Target recall.

Context-assisted face clustering framework with human-in-the-loop

Automatic face clustering, which aims to group faces referring to the same people together, is a key component for face tagging and image management. Standard face clustering approaches that are based on analyzing facial features can already achieve high-precision results. However, they often suffer from low recall due to the large variation of faces in pose, expression, illumination, occlusion, etc. To improve the clustering recall without reducing the high precision, we leverage the heterogeneous context information to iteratively merge the clusters referring to same entities. We first investigate the appropriate methods to utilize the context information at the cluster level, including using of “common scene”, people co-occurrence, human attributes, and clothing. We then propose a unified framework that employs bootstrapping to automatically learn adaptive rules to integrate this heterogeneous contextual information, along with facial features, together. Finally, we discuss a novel methodology for integrating human-in-the-loop feedback mechanisms that leverage human interaction to achieve the high-quality clustering results. Experimental results on two personal photo collections and one real-world surveillance dataset demonstrate the effectiveness of the proposed approach in improving recall while maintaining very high precision of face clustering.

Exploiting temporal characteristics of features for effectively discovering event episodes from news corpora

An organization performing environmental scanning generally monitors or tracks various events concerning its external environment. One of the major resources for environmental scanning is online news documents, which are readily accessible on news websites or infomediaries. However, the proliferation of the World Wide Web, which increases information sources and improves information circulation, has vastly expanded the amount of information to be scanned. Thus, it is essential to develop an effective event episode discovery mechanism to organize news documents pertaining to an event of interest. In this study, we propose two new metrics, Term Frequency × Inverse Document FrequencyTempo (TF×IDFTempo) and TF×Enhanced‐IDFTempo, and develop a temporal‐based event episode discovery (TEED) technique that uses the proposed metrics for feature selection and document representation. Using a traditional TF×IDF‐based hierarchical agglomerative clustering technique as a performance benchmark, our empirical evaluation reveals that the proposed TEED technique outperforms its benchmark, as measured by cluster recall and cluster precision. In addition, the use of TF×Enhanced‐IDFTempo significantly improves the effectiveness of event episode discovery when compared with the use of TF×IDFTempo.

Research on Building Methods of Hierarchical Structure in Text Classification

There always exists semantic hierarchical relationship in text classification. Therefore, it's inevitable to organize documents in accordance with the hierarchical structure. Based on confusion matrix, this paper attempted to adopt two different algorithms including hierarchical clustering and confusion category to build hierarchical structure of document category, and finally made use of hierarchical classification to carry on experiment, results of which showed that the confusion category strategy is superior to hierarchical clustering strategy and recall and precision of flat classification are both improved.

The impact of digital mammography on screening a young cohort of women for breast cancer in an urban specialist breast unit

To compare the diagnostic performance of full-field digital mammography (FFDM) with screen-film mammography (SFM) in a corporate screening programme including younger women. Data were available on 14,946 screening episodes, 5010 FFDM and 9936 SFM. Formal analysis was by logistic regression, adjusting for age and calendar year. FFDM is compared with SFM with reference to cancer detection rates, cancers presenting as clustering microcalcifications, recall rates and PPV of recall. Overall detection rates were 6.4 cancers per thousand screens for FFDM and 2.8 per thousand for SFM (p < 0.001). In women aged 50+ cancer detection was significantly higher for FFDM at 8.6 per thousand vs. 4.0 per thousand, (p = 0.002). In women <50, cancer detection was also significantly higher for FFDM at 4.3 per thousand vs. 1.4 per thousand, (p = 0.02). Cancers detected as clustering microcalcifications increased from 0.4 per thousand with SFM to 2.0 per thousand with FFDM. Rates of assessment recall were higher for FFDM (7.3% vs. 5.0%, p < 0.001). FFDM provided a higher PPV for assessment recall, (32 cancers/364 recalls, 8.8%) than SFM, (28 cancers/493 recalls, 5.7%). Cancer detection rates were significantly higher for FFDM than for SFM, especially for women <50, and cancers detected as clustering microcalcifications.

A Latent Semantic Indexing-based approach to multilingual document clustering

The creation and deployment of knowledge repositories for managing, sharing, and reusing tacit knowledge within an organization has emerged as a prevalent approach in current knowledge management practices. A knowledge repository typically contains vast amounts of formal knowledge elements, which generally are available as documents. To facilitate users' navigation of documents within a knowledge repository, knowledge maps, often created by document clustering techniques, represent an appealing and promising approach. Various document clustering techniques have been proposed in the literature, but most deal with monolingual documents (i.e., written in the same language). However, as a result of increased globalization and advances in Internet technology, an organization often maintains documents in different languages in its knowledge repositories, which necessitates multilingual document clustering (MLDC) to create organizational knowledge maps. Motivated by the significance of this demand, this study designs a Latent Semantic Indexing (LSI)-based MLDC technique capable of generating knowledge maps (i.e., document clusters) from multilingual documents. The empirical evaluation results show that the proposed LSI-based MLDC technique achieves satisfactory clustering effectiveness, measured by both cluster recall and cluster precision, and is capable of maintaining a good balance between monolingual and cross-lingual clustering effectiveness when clustering a multilingual document corpus.

Combining preference- and content-based approaches for improving document clustering effectiveness

E-commerce and knowledge management applications generate and consume tremendous amounts of online information that is typically available as textual documents. To facilitate subsequent access of and leverage from these textual documents, the efficient and effective management of the ever-increasing volume of documents is essential to both organizations and individuals. Document management practices suggest the popularity of using categories (e.g., folders) for organizing, archiving, and accessing documents. Document clustering represents an appealing approach to enable organizations or individuals to create and maintain document categories automatically. Existing document clustering techniques usually group together similar documents on the basis of their textual content similarity. However, such content-based approaches operate at the lexical level and suffer greatly from the word mismatch problem. Therefore, this study aims to address this problem by exploiting users' document grouping preferences, as exhibited in those individuals' folder sets, to support document clustering. Specifically, we propose a hybrid document clustering technique that combines preference- and content-based approaches. Using a traditional content-based and a preference/content switching document clustering technique as performance benchmarks, our empirical evaluation results show that the proposed hybrid technique improves the clustering effectiveness measured by both cluster precision and cluster recall.