Soft-supervised learning for text classification

Subramanya, Amarnag; Bilmes, Jeff

doi:10.3115/1613715.1613857

Cited by 51 publications

(55 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Furthermore, the optimal value for the other parameters were found to be μ 2 = μ 3 = 1. As in previous work [11], we use Precision-Recall Break Even Point (PRBEP) [9] as the evaluation metric. Same evaluation measure, dataset and the same experimental protocol makes the results reported here directly comparable to those reported previously [11].…”

Section: Text Classificationmentioning

confidence: 99%

“…We build on previous research [3,17,11] and construct an objective that reflects three requirements as follows. First, for the labeled vertices we like the output of the algorithm to be close to the a-priori given labels, that is Y v ≈Ŷ v .…”

Section: New Algorithm: Modified Adsorption (Mad)mentioning

confidence: 99%

“…Most recently, the dataset was used by Subramanya and Bilmes [11], who kindly shared their preprocessed complete WebKB graph with us. There are a total of 4, 204 vertices in the graph, with the nodes labeled with one of four categories: course, faculty, project, student.…”

Section: Text Classificationmentioning

confidence: 99%

“…A K-NN graph is created from this complete graph by retaining only top K neighbors of each node, where the value of K is treated as a hyper-parameter. We follow the experimental protocol in [11]. The dataset was randomly partitioned into four sets.…”

Section: Text Classificationmentioning

confidence: 99%

“…Recently, graph based semi-supervised algorithms have achieved considerable attention [2,7,11,14,17]. Such methods represent instances as vertices in a graph with edges between vertices encoding similarities between them.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

New Regularized Algorithms for Transductive Learning

Talukdar

Crammer

2009

Lecture Notes in Computer Science

130

141

View full text Add to dashboard Cite

Abstract. We propose a new graph-based label propagation algorithm for transductive learning. Each example is associated with a vertex in an undirected graph and a weighted edge between two vertices represents similarity between the two corresponding example. We build on Adsorption, a recently proposed algorithm and analyze its properties. We then state our learning algorithm as a convex optimization problem over multi-label assignments and derive an efficient algorithm to solve this problem. We state the conditions under which our algorithm is guaranteed to converge. We provide experimental evidence on various real-world datasets demonstrating the effectiveness of our algorithm over other algorithms for such problems. We also show that our algorithm can be extended to incorporate additional prior information, and demonstrate it with classifying data where the labels are not mutually exclusive.

show abstract

Section: Text Classificationmentioning

confidence: 99%

Section: New Algorithm: Modified Adsorption (Mad)mentioning

confidence: 99%

Section: Text Classificationmentioning

confidence: 99%

Section: Text Classificationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

New Regularized Algorithms for Transductive Learning

Talukdar

Crammer

2009

Lecture Notes in Computer Science

130

141

View full text Add to dashboard Cite

show abstract

Management of Knowledge Sources Supported by Domain Ontologies: Building and Construction Case Studys

Costa

Lima

2015

Intell. Sys. Acc. Fin. Mgmt.

View full text Add to dashboard Cite

SUMMARYThis paper introduces a novel conceptual framework to support the creation of knowledge representations based on enriched semantic vectors, using the classical vector space model approach extended with ontological support. This work is focused on collaborative engineering projects where knowledge plays a key role in the process. Collaboration is the arena, engineering projects are the target and knowledge is the currency used to provide harmony into the arena since it can potentially support innovation and, hence, a successful collaboration. The test bed for the assessment of the approach comes from the Building and Construction sector, which is challenged with significant problems for exchanging, sharing and integrating information among actors. Semantic gaps or lack of meaning definition at the conceptual and technical levels, for example, are problems fundamentally originated through the employment of representations to map the 'world' into models in an endeavour to anticipate other actors' views, vocabulary and even motivations. One of the primary research challenges addressed in this work relates to the process of formalization and representation of document contents, where most existing approaches are limited and only take into account the explicit, word-based information in the document. The research described in this paper explores how traditional knowledge representations can be enriched through incorporation of implicit information derived from the complex relationships (semantic associations) modelled by domain ontologies with the addition of information presented in documents, by providing a baseline for facilitating knowledge interpretation and sharing between humans and machines. Preliminary results were collected using a clustering algorithm for document classification, which indicates that the proposed approach does improve the precision and recall of classifications. Future work and open issues are also discussed.

show abstract

Graph Classification

Tsuda

Saigo

2010

Advances in Database Systems

View full text Add to dashboard Cite

Soft-supervised learning for text classification

Cited by 51 publications

References 17 publications

New Regularized Algorithms for Transductive Learning

New Regularized Algorithms for Transductive Learning

Management of Knowledge Sources Supported by Domain Ontologies: Building and Construction Case Studys

Graph Classification

Contact Info

Product

Resources

About