MétaCan
Menu
Back to cohort
Record W2554862987 · doi:10.15273/pnsis.v43i2.3645

A COMPARISON OF METHODS FOR MODIFYING THE PARTIAL SINGULAR VALUE DECOMPOSITION IN LATENT SEMANTIC INDEXING

2006· article· en· W2554862987 on OpenAlexafffundvenue
Jane E. Tougas

Bibliographic record

VenueProceedings of the Nova Scotian Institute of Science · 2006
Typearticle
Languageen
FieldComputer Science
TopicSemantic Web and Ontologies
Canadian institutionsDalhousie University
FundersNatural Sciences and Engineering Research Council of CanadaKillam Trusts
KeywordsSingular value decompositionFolding (DSP implementation)Computer scienceMatrix decompositionSearch engine indexingTerm (time)Matrix (chemical analysis)Latent semantic analysisData miningInformation retrievalAlgorithmPhysicsChemistry

Abstract

fetched live from OpenAlex

The tremendous size of the Internet and modern databases has made efficientsearching and information retrieval (IR) important. Latent semantic indexing (LSI) is an IR method that represents a dataset as a term-document matrix. LSI uses a matrix factorization method known as the partial singular value decomposition (PSVD). Calculating the PSVD of a large term-document matrix is computationally expensive. In a rapidly expanding environment, a term-document matrix is altered often as new documents and terms are added. Recomputing the PSVD of the term-document matrix each time these slight alterations occur can be prohibitively expensive. Folding-in is one method of adding new documents or terms to an LSI database; updating the PSVD of the existing LSI database is another. The folding-in method is computationally inexpensive, but may cause deterioration in the accuracy of the PSVD. The PSVD-updating method is computationally more expensive than the folding-inmethod, but better maintains the accuracy of the PSVD. Folding-up is a new method that combines folding-in and PSVD-updating. Folding-up is faster than either recomputing the PSVD or PSVD-updating, but avoids the degradation in the PSVD that can occur when the folding-in method is used on its own.La taille incroyable d’Internet et des bases de données modernes a fait en sorteque la recherche efficace d’informations est maintenant importante. L’indexation par sémantique latente (ISL) est une méthode de recherche d’informations qui représente un jeu de données comme une matrice document-terme. L’ISL comprend l’utilisation d’une méthode de factorisation matricielle connue sous le nom de décomposition partielle en valeurs singulières (DPVS). Le calcul de la DPVS d’une grande matrice document-terme est coûteux sur le plan des calculs. Dans un environnement en expansion rapide, une matrice document-terme est souvent modifiée à mesure que de nouveaux documents et termes sont ajoutés. Le recalcul de la DPVS de la matrice document-terme chaque fois qu’une légère modification est apportée peut devenir très coûteux. L’intégration (folding-in) est une méthode pour ajouter de nouveaux documents ou termes dans une base de donnée ISL, et la mise à jour de la DPVS de la base de données ISL existante en est une autre. La méthode d’intégration est peu coûteuse sur le plan des calculs, mais elle peut entraîner une perte d’exactitude de la DPVS. La méthode de mise à jour de la DPVS est plus coûteuse sur le plan des calculs, mais elle permet de mieux préserver l’exactitude de la DPVS. La méthode d’intégration et de mise à jour (folding-up) est une nouvelle méthode qui combine l’intégration et la mise à jour de la DPVS. Cette méthode est plus rapide que le recalcul ou la mise à jour de la DPVS, mais elle permet d’éviter la perte d’exactitude de la DPVS qui peut survenir quand seule la méthode d’intégration est utilisée.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.

How this classification was reachedexpand

Full frame machine prediction

Teacher imitation

Not calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.

metaresearch head score (Codex)0.007
metaresearch head score (Gemma)0.019
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Simulation or modeling · Consensus signal: none
GenreCandidate signal: Methods · Consensus signal: Methods
Teacher disagreement score0.008
Threshold uncertainty score0.037

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0070.019
Meta-epidemiology (narrow)0.0020.001
Meta-epidemiology (broad)0.0010.003
Bibliometrics0.0040.005
Science and technology studies0.0010.001
Scholarly communication0.0020.005
Open science0.0020.002
Research integrity0.0010.002
Insufficient payload (model declined to judge)0.0040.002

Machine scores (provisional)

The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.

Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.

Opus teacher head0.067
GPT teacher head0.387
Teacher spread0.321 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designSimulation or modeling
Domainnot available
GenreMethods

How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".

Quick stats

Citations0
Published2006
Admission routes3
Has abstractyes

Explore more

Same venueProceedings of the Nova Scotian Institute of ScienceSame topicSemantic Web and OntologiesFrench-language works237,207