CCS:
Document representation

Keywords:
document representation

Primary CCS:
Document representation

CCS:
Document representation

Keywords:
document representation

Abstract:
... representing documents, certain information is lost. We argue that better document representation can lead to better agreement with the cluster hypothesis. To ... with the cluster hypothesis. To this end, we investigate different document representations on established benchmarks in the context of low-cost evaluation, showing ... benchmarks in the context of low-cost evaluation, showing that different document representations vary in how well they capture document similarity relative to ...

Primary CCS:
Document representation

Title:
Cluster Hypothesis in Low-Cost IR Evaluation with Different Document Representations

Full Text:
... Meanwhile,in representing documents, certain information is lost. Weargue that better document representation can lead to betteragreement with the cluster hypothesis. To this ... with the cluster hypothesis. To this end, we in-vestigate di?erent document representations ... on establishedbenchmarks in the context of low-cost evaluation, showingthat di?erent document representations vary in how wellthey capture document similarity relative to a ... favor of the low-cost evaluation. To this end,we compare multiple document representations, , includingbag-of-words, latent semantic analysis [4], latent dirichlet al-location [5] ...
... Since the choice of words can in-47Table 1: Comparison of Document Representations on Di?erent BenchmarksTask Benchmark Bow Ebow Lda Lsa Para2vecAdhoc TaskTripleTest ... method to encode the word embedding informationfrom word2vec [7] into document representations, , whereasdocument expansion is used in Ebow.3. EVALUATIONIn this section, ... least one subtopic are used. In Lsa, Ldaand Para2vec, the document representation is computedseparately for each query, given the size of the ... improvements when com-pared against Bow. Intuitively, the comparisons among dif-ferent document representations are in terms of their agree-ment degree to the desirable ...

CCS:
Document representation

Keywords:
XML document representation

CCS:
Document representation

Keywords:
document representation

Abstract:
<p>We deal with the problem of document representation for the task of measuring semantic relatedness between documents. A ...

Primary CCS:
Document representation

Full Text:
... Research, Chinazhuhuij@cn.ibm.comShao Sheng CaoXidian Universityshelsoncao@gmail.comABSTRACTWe deal with the problem of document representation forthe task of measuring semantic relatedness between docu-ments. A document ... a similarity measure betweendocuments in terms of their respective representations.Most document representations map the documents tovectors of a fixed length, aiming to ...
... methods to represent concepts (as opposed towords) as continuous vectors.For document representation we build a graph whose nodesare explicit concepts from a ... aconcept as a continuous vector of 200 dimensions.We evaluate our document representation and similaritymeasures on LP50 [10], a standard benchmark for docu-ment ... systems.To summarize, our contributions are the following:? We propose a document representation as a conceptgraph whose nodes and edges are weighted. In ...
... Techniques are specifiedin detail in section 4.2.The key idea of document representation by a conceptgraph is to link the concepts in the ...

CCS:
Document representation

Keywords:
document representation

Abstract:
... human-curated knowledge base to incorporate other related concepts in the document representation. . But these methods are not desirable when applied to ... keyphrases, going beyond just explicit mentions. Compared with the state-of-art document representation approaches, LAKI fills the gap between bag-of-words and concept-based models ...

Primary CCS:
Document representation

Full Text:
... a human-curated knowledge base to incorporate otherrelated concepts in the document representation. . But thesemethods are not desirable when applied to vertical ... latent documentkeyphrases, going beyond just explicit mentions. Comparedwith the state-of-art document representation approaches,LAKI fills the gap between bag-of-words and concept-basedmodels by using ...
... nodes represent domain keyphrases and content units respectively.forward interpretation for document representation, , whichis critical for model verification and for ensuring that ...
... Analysis (LSA) [7] is a topic modelingtechnique learning word and document representations byapplying Singular Value Decomposition to the words-by-documents co-occurrence matrix? Latent ... Latent Keyphrase Inference (LAKI) is proposed in thiswork to derive document representation via inferring la-tent keyphrases in the text.Table 2 provides more ...
... . .? ?? ?? ?? ?Academia YelpTable 7: Examples of document representation by LAKI with top-ranked document keyphrases in the vector (related-ness ...

CCS:
Document representation

Keywords:
document representation

Primary CCS:
Document representation

Full Text:
... byas much as 18% in standard document ranking tasks.KeywordsText Representation, Document Representation, , KnowledgeBase, Bag-of-Entities1. INTRODUCTIONIn the earliest information retrieval systems, query ...

CCS:
Document representation

Keywords:
document representation

References:
X. He, D. Cai, H. Liu, and W.-Y. Ma. Locality preserving indexing for document representation. In Proc. 2004 Int. Conf. on Research and Development in Information Retrieval (SIGIR'04), pages 96--103, Sheffield, UK, July 2004.

Full Text:
... and Indexing?Indexing methodsGeneral TermsAlgorithms, Performance, TheoryKeywordsProbabilistic Latent Semantic Indexing, Manifold Regularization,Document Representation, , Generative ModelPermission to make digital or hard copies of ...
... D. Cai, H. Liu, and W.-Y. Ma. Locality preservingindexing for document representation. . In Proc. 2004 Int.Conf. on Research and Development in ...

CCS:
Document representation

Keywords:
document representation

Abstract:
... to quantify information in probability distributions and derive a new document representation model for text clustering. By extending Shannon entropy to accommodate ...

Full Text:
... and Retrieval?ClusteringGeneral TermsTheory, Algorithms, Performance, ExperimentationKeywordsterm weighting, information measure, semantic information,document representation, , text clusteringPermission to make digital or hard copies of ... clustering research, TF*IDF has been extensivelyused for term weighting and document representation [21,34]. While term frequency (TF) indicates the degree of adocument?s ...
... we apply the proposed least information theory toterm weighting and document representation. . A text docu-ment can be viewed as a set ...
... system basedon the Weka data mining framework [32]. We implementedvarious document representation methods including the pro-posed term weighting schemes and TF*IDF based ...
... look at the impact of feature selection on theeffectiveness of document representation for clustering. Inthis work, we selected features based on their ... for which the top Nf mostfrequent terms were kept for document representation. . Weperformed this experimental analysis on three collections,namely, WebKB, 20Newsgroup, ...
... point degraded clusteringperformance when there were insufficient features for accu-rate document representation. . As shown in Figures 3 and1494, similar patterns about ...

CCS:
Document representation

Keywords:
document representation and indexing

Primary CCS:
Document representation

References:
X. He, D. Cai, H. Liu, and W.-Y. Ma. Locality preserving indexing for document representation. In Proc. 2004 Int. Conf. on Research and Development in Information Retrieval (SIGIR'04), pages 96--103, Sheffield, UK, July 2004.

Full Text:
... large size.3. REGULARIZED LOCALITYPRESERVING INDEXINGAlthough LPI can learn a compact document representation whichis beneficial for many text analysis tasks such as clustering ...
... D. Cai, H. Liu, and W.-Y. Ma. Locality preservingindexing for document representation. . In Proc. 2004 Int.Conf. on Research and Development in ...

CCS:
Document representation

Keywords:
document representation model

Full Text:
... Similarity Measure1. INTRODUCTIONDocument clustering techniques usually rely on four mod-ules: document representation model, similarity measure,clustering model and the clustering algorithm which gener-ates ... and the clustering algorithm which gener-ates clusters based on the document representation model[5]. Among all these modules, the document representationmodel is very ... and crucial for the clustering re-sults.The most basic model for document representation is theVector Space Document (VSD) model. In this model, thedocument ... suggested in the natural language sentences arestill ignored in these document representation models.In this paper, we propose a more informative documentrepresentation model, ...
... experimentally.The baseline approach utilizes the ?bag of words? (BOW)model for document representation and weights each wordin the feature vector by tf-idf measure. ...

CCS:
Document representation

Keywords:
document representation

CCS:
Document representation

Keywords:
complex document representation

Primary CCS:
Document representation

References:
CROFT, W.B. Document representation in probabilistic models of information retrieval. J. Am. Soc. Inf. Sc~. 32, (1981), 451-457.

Full Text:
... Learning?parameter learningGeneral Terms: Experimentation, TheoryAdditional Key Words and Phrases: Complex document representation, , linear indexing func-tions, linear retrieval functions, probabilistic indexing, probabilistic ...
... representations D and Q of these objects.With the mapping a~, document representations D are derived from theoriginal documents ~. In the same ... @ A ~J G D] of pairs ofrelevance judgments and document representations. . Thus, two queries qland q2 have the same representation ...
... = 1, ift,E q:, and z~i ?? O otherwise. The document representation is not furtherspecified in the BII model, and below we ...
... withmultiple occurrences can be abandoned (some concepts for a moredetailed document representation are described by Fuhr [13]). On theother hand, our approach ...
... structures).Document RepresentationIn order to test the effect of an improved document representation, , weperformed a few experiments with the indexing function e~a. ...
... Sot. Inf Set. 37, 2 (19S6), 71-77.6. CROFT, W. B. Document representation in probabilistic models of information retrieval. J.Am. Sot. In; SCL, ...

CCS:
Document representation

Keywords:
document representation and indexing

Primary CCS:
Document representation

References:
X. He, D. Cai, H. Liu, and W.-Y. Ma. Locality preserving indexing for document representation. In Proceedings of ACM SIGIR, 2004.

Full Text:
... Measurement, Performance, Experimentation, TheoryKeywordsOrthogonal Locality Preserving Indexing, Locality Preserving In-dexing, Document Representation and Indexing, Similarity Mea-sure, Dimensionality Reduction, Vector Space ModelPermission to ... Model (VSM) might be one of themost popular model for document representation. . Each documentis represented as a bag of words. Correspondingly, ...
... ), LSI (correspondingto document set DLSI ) and the original document representation( (corresponding to document set D as baseline algorithm). In gen-eral, ...
... D. Cai, H. Liu, and W.-Y. Ma. Locality preservingindexing for document representation. . In Proceedings ofACM SIGIR, 2004.[12] T. Hofmann. Probabilistic latent ...

CCS:
Document representation

Keywords:
document representation and indexing

Abstract:
Document representation and indexing is a key problem for document analysis and ... an indexing. LSI essentially detects the most representative features for document representation rather than the most discriminative features. Therefore, LSI might not ... space, LPI discovers the local structure and obtains a compact document representation subspace that best detects the essential semantic structure. We compare ...

Primary CCS:
Document representation

Title:
Locality preserving indexing for document representation

Full Text:
Microsoft Word - p250-he_1_.docLocality Preserving Indexing for Document Representation Xiaofei He Computer Science Dept. University of Chicago xiaofei@cs.uchicago.edu Deng ... hfliu@cs.toronto.edu Wei-Ying Ma Microsoft Research Asia Beijing, China wyma@microsoft.com ABSTRACT Document representation and indexing is a key problem for document analysis and ... space, LPI discovers the local structure and obtains a compact document representation subspace that best detects the essential semantic structure. We compare ... Indexing, Similarity Measure, Dimen-sionality Reduction, Vector Space Model 1. INTRODUCTION Document representation and indexing is a fundamental problem for efficient clustering, classification, ... Many dimensionality reduction techniques [1][2][5][7][14] [15] have been applied to document representation and indexing. Among these techniques, Latent Semantic Indexing (LSI) [7] ... the most repre-sentative features rather the most discriminative features for document representation. . Therefore, LSI might not be optimal in discriminating documents ... propose a new approach called Locality Preserv-ing Indexing (LPI) to document representation, , which aims to discover the local geometrical structure of ...
... a semantic sub-space. Section 3 introduces Locality Preserving Indexing for document representation. . Theoretical analysis of LPP and its con-nections to LDA ... Locality Preserving Projections (LPP) [13], the core algorithm used for document representation and indexing in this paper. Different from LSI which assumes ...
... preserving subspace. Based on LPP, we describe our method for document representation and indexing. In the document analysis and processing problems one ...
... apply LPP to learn a low dimensional semantic space for document representation. . 4.3 Discriminant Analysis of LPP Traditionally, document indexing and ...

Keywords:
Document representation, Classification

References:
Nguyen, GS., Gao X. and Andreae P.Vietnamese Document Representation and Classification, AI 2009: ADVANCES IN ARTIFICIAL INTELLIGENCE, 2009, Volume 5866/2009, 577 - 586.

Full Text:
... their favorite sound patterns. We aim to develop a new document representation using the phonemes. The vocabulary of phonemes is much smaller ...
... NY. 1997. [10] Nguyen, GS., Gao X. and Andreae P.Vietnamese Document Representation and Classification, AI 2009: ADVANCES IN ARTIFICIAL INTELLIGENCE, 2009, Volume ...

Title:
Enriching document representation via translation for improved monolingual information retrieval

Keywords:
document representation

Abstract:
... this paper proposes the use of translated words to enrich document representation, , going beyond the words in the original source language ...

Full Text:
mt4ir.dviEnriching Document Representation via Translation forImproved Monolingual Information RetrievalSeung-Hoon NaDepartment of Computer ScienceNational ...
... translated using the pro-posed method of expected frequency estimation, producingbilingual document representations (Section 5). When a newtest query is given, the query ...
... follows:c(w,dclu) =Xd??Cludc(w,d?)We now additionally introduce d?clu to indicate the cluster-enhanced document representation of d, which is theweighted representation of the original source ... c(w,d) + ?clu c(w,dclu)where ?clu is the weight of document representation to clus-ter representation. To estimate the smoothed cluster lan-guage model, ...

Keywords:
document representation

Abstract:
... most common way to solve this problem is to enrich document representation with the background knowledge in an ontology. There are two ... text clustering method to address these two issues by enriching document representation with Wikipedia concept and category information. We develop two approaches, ... LA Times) show that clustering performance improves significantly by enriching document representation

Full Text:
... most common way to solve this problem is to enrich document representation with the background knowledge in an ontology. There are two ... text clustering method to address these two issues by enriching document representation with Wikipedia concept and category information. We develop two approaches, ... LA Times) show that clustering performance improves significantly by enriching document representation with Wikipedia concepts and categories. Categories and Subject Descriptors I.5.3 ... similarity measures. General Terms Algorithms, Experimentation Keywords Text Clustering, Wikipedia, Document Representation 1. INTRODUCTION Traditional clustering algorithms are usually based on the ... forms. One way to resolve this problem is to enrich document representation with the background knowledge represented by an ontology. An ontology ... relationships among concepts. All of them can be used for document representation and clustering. The most common way of applying ontologies for ... possible; and a proper matching method which can enrich the document representation by fully leveraging ontology terms and relations but without introducing ...
... clustering. As for how to integrate ontology concepts into the document representation and clustering process, in this paper, we propose two approaches ... baseline. The results show that, in agglomerative clustering method, enriching document representation with Wikipedia concepts and categories by both exact-match and relatedness ... propose a method to improve text classification performance by enriching document representation with Wikipedia concepts. The mapping between each document and Wikipedia ...
... news. The method in both [1] and [2] only augment document representation with Wikipedia concepts without considering the hierarchical relationship embedded in ... In our method, we also integrate Wikipedia category information into document representation based on the hierarchical structure of Wikipedia. We believe that ...

Keywords:
document representation

Full Text:
... Filtering; G.3 [Probabilityand Statistics]: Statistical ComputingGeneral TermsAlgorithms, ExperimentationKeywordsStatistical Topic Modeling, Document Representation? ?Main contact.Permission to make digital or hard copies of all ... research in topic modeling. This research has2 main streams in document representation: : 1) The explo-ration of super and subtopics as in ...

Keywords:
document representation

Full Text:
... STORAGEANDRETRIEVAL?Digital Libraries;I.7.2 [Computing Methodologies]: DOCUMENT ANDTEXT PROCESSING?Document PreparationGeneral TermsALGORITHMSKeywordsLayout Analysis; Document Representation; ; Document Ren-dering1. INTRODUCTIONThe wide spread of documents in digital ... called DoMInUS. The next Section introduces DoMI-nUS and its internal document representation. . Then, Sec-tion 3 describes the identification and reconstruction of ...
... identifyingthe high-level geometrical structure of the document, andspecifically to the document representation they are basedon. Technical details on how this representation is ...

Full Text:
... the art search [11]. Konishi et al. extended the patent document representation model by integrating TF/IDF of International Patent Classification (IPC) and ... and patent document structure ontology, to introduce a content-based patent document representation schema for patent processing, e.g. patent retrieval, classification and clustering ...