Academia.eduAcademia.edu

Outline

Title
Abstract
Key Takeaways
Figures
Introduction
Scope of This Work
Statistical Models of Documents
Formation of Statistical Models
Results
Conclusions
References
All Topics
Mathematics
Statistics

Self Organization of a Massive Document Collection

Profile image of Samuel KaskiSamuel Kaski

2002, Neural Networks, …

https://doi.org/10.1109/72.846729
visibility

…

description

12 pages

link

1 file

Abstract
sparkles

AI

Self-organization of massive document collections is explored, emphasizing the challenges of traditional keyword searches and the potential of organizing documents visually in a 2D space. The study discusses the implementation of self-organizing maps (SOM) for effective data retrieval, especially in exploratory data analysis, showcasing its computational advantages over classical methods like multidimensional scaling. Experimental results comparing various projection methods illustrate the effectiveness of SOM in improving document classification and retrieval accuracy.

Key takeaways
sparkles

AI

  1. The WEBSOM methodology organizes 7 million patent abstracts into a 2D similarity map for enhanced data exploration.
  2. Utilizing self-organizing maps (SOM) reduces computational complexity significantly compared to traditional methods.
  3. Dimensionality reduction techniques, including random projections, maintain classification accuracy while speeding up processing.
  4. Document maps enhance user interaction by visually displaying related documents based on content similarity.
  5. Exploratory data analysis techniques enable users to discover relevant information beyond initial search queries.
Figures (8)
Fig. 1. Overview of the construction and operation of the WEBSOM2 system. phases have been explained, e.g., in Honkela et al. [37]. Certain reasons expounded below, however, recently led us to abandon the semantic maps and to encode the word histograms by the newly developed projection methods. The results reported in the present article are totally based on these new develop- ments, which has given us a reason to call the present system the WEBSOM2. An overview of the WEBSOM2 system is depicted in Fig. 1.
Fig. 1. Overview of the construction and operation of the WEBSOM2 system. phases have been explained, e.g., in Honkela et al. [37]. Certain reasons expounded below, however, recently led us to abandon the semantic maps and to encode the word histograms by the newly developed projection methods. The results reported in the present article are totally based on these new develop- ments, which has given us a reason to call the present system the WEBSOM2. An overview of the WEBSOM2 system is depicted in Fig. 1.
It can be shown that the similarity of pairs of projection vec- tors (x;,x,), measured by the inner products, is the same on the average as the similarity of the pairs of corresponding orig- inal document vectors (n;,n,;), and the error thereby made is inversely proportional to m [41]. It was demonstrated experi- mentally [41] for dimensionalities of m exceeding 100 (cf. also Table I) that the classification accuracy is practically as good as with the vector-space method while, with the decreasing dimen- sionality of the document vectors, the time needed to classify a document is decreased radically.
It can be shown that the similarity of pairs of projection vec- tors (x;,x,), measured by the inner products, is the same on the average as the similarity of the pairs of corresponding orig- inal document vectors (n;,n,;), and the error thereby made is inversely proportional to m [41]. It was demonstrated experi- mentally [41] for dimensionalities of m exceeding 100 (cf. also Table I) that the classification accuracy is practically as good as with the vector-space method while, with the decreasing dimen- sionality of the document vectors, the time needed to classify a document is decreased radically.
collection, each document was mapped onto one of the grid points of each map, and all documents that represented a minority class at any grid point were counted as classification errorss .
collection, each document was mapped onto one of the grid points of each map, and all documents that represented a minority class at any grid point were counted as classification errorss .
Fig. 2. Illustration of the relations of the model vectors in a sparse (s, solid lines) and dense (d, dashed lines) grid. Only partial grids are shown in the figure. Here m,,’‘4 shall be interpolated in terms of the three closest sparse models m,'‘), m,’¢), and m,;’*?, respectively.
Fig. 2. Illustration of the relations of the model vectors in a sparse (s, solid lines) and dense (d, dashed lines) grid. Only partial grids are shown in the figure. Here m,,’‘4 shall be interpolated in terms of the three closest sparse models m,'‘), m,’¢), and m,;’*?, respectively.
Fig. 3. Finding the new winner in the vicinity of the old one, whereby the old winner is directly located by a pointer. The pointer is then updated.
Fig. 3. Finding the new winner in the vicinity of the old one, whereby the old winner is directly located by a pointer. The pointer is then updated.
Fig. 4. Distribution of four sample subsections of the patent classification system on the document map. The gray level indicates the logarithm of the number of patents in each node.
Fig. 4. Distribution of four sample subsections of the patent classification system on the document map. The gray level indicates the logarithm of the number of patents in each node.
Fig. 5. Content addressable search was utilized to find information on laser surgery on the cornea of the eye. The best matching locations are marked with circles. Zooming on the area reveals a small cluster of map units that contains patent abstracts mostly about the cornea of the eye, and of surgical operations on it. Several abstracts concemed with the description of interest, i.e., laser surgery on the cornea, are found in the best matching units.
Fig. 5. Content addressable search was utilized to find information on laser surgery on the cornea of the eye. The best matching locations are marked with circles. Zooming on the area reveals a small cluster of map units that contains patent abstracts mostly about the cornea of the eye, and of surgical operations on it. Several abstracts concemed with the description of interest, i.e., laser surgery on the cornea, are found in the best matching units.
Fig. 6. The keyword search mode was utilized to find information on color displays. 30 best-matching units were marked on the display with circles the size of which indicates the goodness of the match. As seen from the map display, the matches are distributed into several tight clusters found in different regions of the map. From two of these clusters the partial contents of a sample matching unit are shown in the insets. Closer inspection of the units reveals different aspects of color and displays. Unit a features a considerable number of abstracts about color filters used in building LCD displays, whereas in b one finds technology related to displaying colors when printing documents (the Descriptive-words lists were found for each map unit using the automatic keyword selection method introduced in [51]). The user who probably did not have printing in mind when formulating the query can then concentrate on the other clusters.
Fig. 6. The keyword search mode was utilized to find information on color displays. 30 best-matching units were marked on the display with circles the size of which indicates the goodness of the match. As seen from the map display, the matches are distributed into several tight clusters found in different regions of the map. From two of these clusters the partial contents of a sample matching unit are shown in the insets. Closer inspection of the units reveals different aspects of color and displays. Unit a features a considerable number of abstracts about color filters used in building LCD displays, whereas in b one finds technology related to displaying colors when printing documents (the Descriptive-words lists were found for each map unit using the automatic keyword selection method introduced in [51]). The user who probably did not have printing in mind when formulating the query can then concentrate on the other clusters.

Related papers

Self-organization of very large document collections
Jarkko Salojärvi

2000

Text mining systems are developed to aid the users in satisfying their information needs, which may vary from searching answers to well-specified questions to learning more of a scientific discipline. The major tasks of web mining are searching, browsing, and visualization. Searching is best suited for answering specific questions of a well-informed user. Browsing and visualization, on the other hand, are beneficial especially when the information need is more general, or the topic area is new to the user [6]. The SOM, applied to organizing very large document collections, can aid in all the three tasks.

View PDFchevron_right
Document Clustering Using Self-Organizing Maps : A Multi-Features Layered Approach
Umar Ayub

2017

Cluster analysis of textual documents is a common technique for better filtering, navigation, understanding and comprehension of the large document collection. Document clustering is an autonomous method that separate out large heterogeneous document collection into smaller more homogeneous sub-collections called clusters. Self-organizing maps (SOM) is a type of artificial neural network (ANN) that can be used to perform autonomous self-organization of high dimension feature space into low-dimensional projections called maps. It is considered a good method to perform clustering as both requires unsupervised processing. In this paper, we proposed a SOM using multi-layer, multi-feature to cluster documents. The paper implements a SOM using four layers containing lexical terms, phrases and sequences in bottom layers respectively and combining all at the top layers. The documents are processed to extract these features to feed the SOM. The internal weights and interconnections between t...

View PDFchevron_right
Exploration of Full-Text Databases With Self-Organizing Maps
Samuel Kaski

Neural Networks, 1996., …, 2002

View PDFchevron_right
Text Retrieval Using Self-Organized Document Maps
Krista Lagus

Neural Processing Letters - NPL, 2002

A map of text documents arranged using the Self-Organizing Map (SOM) algorithm (1) is organized in a meaningful manner so that items with similar content appear at nearby locations of the 2-dimensional map display, and (2) clusters the data, resulting in an approximate model of the data distribution in the high-dimensional document space. This article describes how a document map that is automatically organized for browsing and visualization can be successfully utilized also in speeding up document retrieval. Furthermore, experiments on the well-known CISI collection [3] show significantly improved performance compared to Salton's vector space model, measured by average precision (AP) when retrieving a small, fixed number of best documents. Regarding comparison with Latent Semantic Indexing the results are inconclusive.

View PDFchevron_right
Document Clustering Using the 1 + 1 Dimensional Self-Organising Map
Nigel Allinson

Lecture Notes in Computer Science, 2002

Automatic clustering of documents is a task that has become increasingly important with the explosion of online information. The Self-Organising Map (SOM) has been used to cluster documents effectively, but efforts to date have used a single or a series of 2-dimensional maps. Ideally, the output of a document-clustering algorithm should be easy for a user to interpret. This paper describes a method of clustering documents using a series of 1dimensional SOM arranged hierarchically to provide an intuitive tree structure representing document clusters. Wordnet 1 is used to find the base forms of words and only cluster on words that can be nouns.

View PDFchevron_right
Using the Self-Organizing Map (SOM) Algorithm, as a Prototype E-Content Retrieval Tool
ATHANASIOS DRIGAS

Lecture Notes in Computer Science, 2008

SOM O.D.I.S.S.E.A.S is an intelligent searching tool using the Self-Organizing Map (SOM) algorithm, as a prototype e-content retrieval tool. The proposed searching tool has the ability to adjust and scale into any e-learning system that requires concept-based queries. In the proposed methodology, maps are used for the automatic replacement of the unstructured, the half structured and the multidimensional data of text, in a way that similar entries in the map are represented near between them. The performance and the functionality of the document organization, and the retrieval tool employing the SOM architecture, are also presented. Furthermore, experiments were performed to test the time performance of a learning algorithm used for the direct creation of teams of terms and texts enabling efficient searching and retrieval of the documents.

View PDFchevron_right
A quickly trainable hybrid SOM-based document organization system
Renato Correa

Neurocomputing, 2008

The large volume of nowadays document collections has increased the need of fast trainable document organization systems. This paper presents and evaluates a hybrid system to self-organization of massive document collections based on self-organizing map (SOM). The hybrid system uses prototypes generated by a clustering algorithm to train the document maps, thus reducing the training time of large maps. We test the system with k-means and modified leader clustering algorithms. The experiments are carried out with the Reuters-21758 v1.0 and 20 Newsgroup collections. The performance of the system is measured in terms of text categorization effectiveness on test set and training time. Experimental results show that the proposed system generates effective document maps in less time than SOM. However, the hybrid system using k-means generates better document maps than the one using modified leader at the cost of more long training time.

View PDFchevron_right
Supervised learning for automatic classification of documents using self-organizing maps
dror lev

… Seeking, Searching and Querying in Digital …, 2000

S Su up pe er rv vi is se ed d L Le ea ar rn ni in ng g f fo or r A Au ut to om ma at ti ic c C Cl la as ss si if fi ic ca at ti io on n o of f D Do oc cu um me en nt ts s u us si in ng g S Se el lf f--O Or rg ga an ni iz zi in ng g M Ma ap ps s Abstract Automatic Document Classification that corresponds with user-predefined classes is a challenging and widely researched area. Self-Organizing Maps (SOM) are unsupervised Artificial Neural Networks (ANN) which are mathematically characterized by transforming high-dimensional data into two-dimension representation, enabling automatic clustering of the input, while preserving higher order topology. A closely related algorithm is the Learning Vector Quantization (LVQ), which uses supervised learning to maximize correct data classification. This study presents the application of SOM and LVQ to automatic document classification, based on predefined set of clusters. A set of documents, manually clustered by domain expert was used. Experimental results show considerable success of automatic document clustering that matches manual clustering, with a slight preference for the LVQ.

View PDFchevron_right
On Document Classification with Self-Organising Maps
Jorma Laurikkala, Jyri Saarikoski

Lecture Notes in Computer Science, 2009

This research deals with the use of self-organising maps for the classification of text documents. The aim was to classify documents to separate classes according to their topics. We therefore constructed self-organising maps that were effective for this task and tested them with German newspaper documents. We compared the results gained to those of k nearest neighbour searching and k-means clustering. For five and ten classes, the self-organising maps were better yielding as high average classification accuracies as 88-89%, whereas nearest neighbour searching gave 74-83% and k-means clustering 72-79% as their highest accuracies.

View PDFchevron_right
Browsing Digital Libraries with the Aid of Self-Organizing Maps
Krista Lagus

1999

Powerful methods for exploring and searching collections of free-form textual documents are needed to control the flood of digital information emerging from various sources. In this article we present a method, WEBSOM, for automatic organization of document collections based on full-text analysis using the Self-Organizing Map. The document collection is ordered on the map in such a way that similar documents lie near each other. The WWW-based user interface provides the basic functionalities necessary for intuitive exploration of the document space visualized with a map: moving on the map, zooming, and examining individual documents. We apply the method to the Usenet newsgroup comp.ai.neural-nets.

View PDFchevron_right

References (52)

  1. J. W. Tukey, Exploratory Data Analysis. Reading, MA: Ad- dison-Wesley, 1977.
  2. G. Young and A. S. Householder, "Discussion of a set of points in terms of their mutual distances," Psychometrica, vol. 3, pp. 19-22, 1938.
  3. W. S. Torgerson, "Multidimensional scaling: I. Theory and method," Psychometrica, vol. 17, pp. 401-419, 1952.
  4. J. B. Kruskal and M. Wish, "Multidimensional scaling," Sage Univ. Paper Series on Quantitative Applications in the Social Sciences, New- bury Park, CA, Tech. Rep. 07-011,, 1978.
  5. J. de Leeuw and W. Heiser, "Theory of multidimensional scaling," in Handbook of Statistics, P. R. Krishnaiah and L. N. Kanal, Eds. Amsterdam, The Netherlands: North-Holland, 1982, vol. 2, pp. 285-316.
  6. M. Wish and J. D. Carrol, "Multidimensional scaling and its applica- tions," in Handbook of Statistics, P. R. Krishnaiah and L. N. Kanal, Eds. Amsterdam, The Netherlands: North-Holland, 1982, vol. 2, pp. 317-345.
  7. F. W. Young, "Multidimensional scaling," in Encyclopedia of Statistical Sciences, S. Kotz, N. L. Johnson, and C. B. Reads, Eds. New York: Wiley, 1985, vol. 5, pp. 649-659.
  8. J. W. Sammon, Jr., "A nonlinear mapping for data structure analysis," IEEE Trans. Comput., vol. C-18, pp. 401-409, 1969.
  9. T. Kohonen, "Self-organized formation of topologically correct feature maps," Biol. Cybern., vol. 43, no. 1, pp. 59-69, 1982.
  10. "Clustering, taxonomy, and topological maps of patterns," in Proc.
  11. Sixth Int. Conf. Pattern Recognition, Munich, Germany, Oct. 19-22, 1982, pp. 114-128.
  12. Self-Organizing Maps, 2nd ed. Berlin, Germany: Springer, 1997.
  13. X. Lin, D. Soergel, and G. Marchionini, "A self-organizing semantic map for information retrieval," in Proc. 14th Annu. Int. ACM/SIGIR Conf. Research and Development in Information Retrieval, 1991, pp. 262-269.
  14. J. C. Scholtes, "Unsupervised learning and the information retrieval problem," in Proc. IJCNN '91, Int. Joint Conf. Neural Networks, vol. I, Singapore, 1991, pp. 95-100.
  15. D. Merkl and A. M. Tjoa, "The representation of semantic similarity between documents by using maps: Application of an artificial neural network to organize software libraries," in Proc. FID '94, General Assembly Conf. Congress Int. Federation Information Documentation, 1994.
  16. T. Honkela, S. Kaski, K. Lagus, and T. Kohonen, "Newsgroup explo- ration with WEBSOM method and browsing interface," Helsinki Univ. Technol., Lab. Comput. Inform. Sci., Espoo, Finland, Tech. Rep. A32, 1996.
  17. S. Kaski, T. Honkela, K. Lagus, and T. Kohonen, "Creating an order in digital libraries with self-organizing maps," in Proc. WCNN '96, World Congr. Neural Network , San Diego, CA, Sept. 15-18, 1996, pp. 814-817.
  18. T. Kohonen, S. Kaski, K. Lagus, and T. Honkela, "Very large two-level SOM for the browsing of newsgroups," in Proc. ICANN '96, Int. Conf. Artificial Neural Networks, Bochum, Germany, July 16-19, 1996, pp. 269-274.
  19. K. Lagus, T. Honkela, S. Kaski, and T. Kohonen, "Self-organizing maps of document collections: A new approach to interactive exploration," in Proceedings of the Second International Conference on Knowledge Discovery and Data Mining, E. Simoudis, J. Han, and U. Fayyad, Eds. Menlo Park, CA: AAAI, 1996, pp. 238-243.
  20. T. Kohonen, "Exploration of very large databases by self-organizing maps," in Proc. ICNN '97, Int. Conf. Neural Networks, Houston, TX, 1997, pp. PL1-PL6.
  21. S. Kaski, T. Honkela, K. Lagus, and T. Kohonen, "WEBSOM-Self- organizing maps of document collections," Neurocomputing, vol. 21, pp. 101-117, 1998.
  22. K. Lagus, T. Honkela, S. Kaski, and T. Kohonen, "WEBSOM for textual data mining," Artif. Intell. Rev., vol. 13, pp. 345-364, 1999.
  23. T. Kohonen, S. Kaski, K. Lagus, J. Salojärvi, J. Honkela, V. Paatero, and A. Saarela, "Self organization of a massive text document collection," in Kohonen Maps, E. Oja and S. Kaski, Eds. Amsterdam, The Nether- lands: Elsevier, 1999, pp. 171-182.
  24. T. Kohonen, "Comparison of SOM point densities based on different criteria," Neural Comput., vol. 11, no. 8, pp. 2171-2185, 1999.
  25. "New developments of Learning vector Quantization and the self- organizing map," in Symposium on Neural Networks; Alliances and Per- spectives in Senri. Osaka, Japan: Senri Int. Information Institute, 1992.
  26. Y. Cheng, "Convergence and ordering of Kohonen's batch map," Neural Comput., vol. 9, no. 8, pp. 1667-1676, 1997.
  27. A. Gersho, "Asymptotically optimal block quantization," IEEE Trans. Inform. Theory, vol. 25, pp. 373-380, July, 1979.
  28. R. M. Gray, "Vector quantization," IEEE ASSP Mag., pp. 4-29, Apr. 1984.
  29. J. Makhoul, S. Roucos, and H. Gish, "Vector quantization in speech coding," Proc. IEEE, vol. 73, pp. 1551-1588, Nov. 1985.
  30. T. Kohonen, J. Hynninen, J. Kangas, and J. Laaksonen, "SOM_PAK: The self-organizing map program package," Helsinki Univ. Technol., Lab. Computer Information Sci., Rep. A31, Jan., 1996.
  31. D. Koller and M. Sahami, "Toward optimal feature selection," in Ma- chine Learning: Proc. Thirteenth Int. Conf. (ICML '96), L. Saitta, Ed., 1996, pp. 284-292.
  32. H. Chen, C. Schuffels, and R. Orwig, "Internet categorization and search: A self-organizing approach," J. Vis. Commun. Image Represent., vol. 7, no. 1, pp. 88-102, 1996.
  33. S. Lesteven, Ponçot, and F. Murtagh, "Neural networks and infor- mation extraction in astronomical information retrieval," Vistas Astron., vol. 40, p. 395, 1996.
  34. X. Lin, "Map displays for information retrieval," J. Amer. Soc. Inform. Sci., vol. 48, pp. 40-54, 1997.
  35. D. Merkl, "Text classification with self-organizing maps: Some lessons learned," Neurocomputing, vol. 21, pp. 61-77, 1998.
  36. H. Chen, J. Nunamaker, Jr., R. Orwig, and O. Titkova, "Information vi- sualization for collaborative computing," IEEE Computer, pp. 75-82, Aug. 1998.
  37. H. Ritter and T. Kohonen, "Self-organizing semantic maps," Biol. Cy- bern., vol. 61, no. 4, pp. 241-254, 1989.
  38. T. Honkela, S. Kaski, K. Lagus, and T. Kohonen, "WEBSOM-Self-or- ganizing maps of document collections," in Proc. WSOM '97, Workshop Self-Organizing Maps, Espoo, Finland, June 4-6, 1997, pp. 310-315.
  39. G. Salton and M. J. McGill, Introduction to Modern Information Re- trieval. New York: McGraw-Hill, 1983.
  40. S. Deerwester, S. T. Dumais, G. W. Furnas, and T. K. Landauer, "In- dexing by latent semantic analysis," J. Amer. Soc. Inform. Sci., vol. 41, pp. 391-407, 1990.
  41. S. Kaski, "Dimensionality reduction by random mapping: Fast similarity computation for clustering," D.Sc. thesis, Helsinki Univ. Technol., Fin- land, Mar. 1997.
  42. Proc. IJCNN '98 Int. Joint Conf. Neural Networks, vol. 1, 1998, pp. 413-418.
  43. P. Drineas, A. Frieze, R. Kannan, S. Vempala, and V. Vinay, "Clustering in large graphs and matrices," in Proc. 10th ACM-SIAM Symp. Discrete Algorithms, San Francisco, CA, 1999, pp. 291-299.
  44. C. H. Papadimitriou, P. Raghavan, H. Tamaki, and S. Vempala, "La- tent semantic indexing: A probabilistic analysis," in Proc. Seventeenth ACM SIGACT-SIGMOID-SIGART Symp. Principles of Database Sys- tems, Seattle, WA, June 1-4, 1998, pp. 159-168.
  45. T. Kohonen, "Self-organization of very large document collections: State of the art," in Proc. ICANN98, 8th Int. Conf. Artificial Neural Networks, vol. 1, L. Niklasson, M. Bodén, and T. Ziemke, Eds., 1998, pp. 65-74.
  46. D. Roussinov and H. Chen, "A scalable self-organizing map algorithm for textual classification: A neural network approach to thesaurus gener- ation," CC-AI-Commun., Cogn. Artif. Intell., vol. 15, pp. 81-111, 1998.
  47. J. S. Rodrigues and L. B. Almeida, "Improving the learning speed in topological maps of patterns," in Proc. INNC '90, Int. Neural Networks Conf. , 1990, pp. 813-816.
  48. P. Koikkalainen, "Progress with the tree-structured self-organizing map," in Proc. ECAI'94, 11th Eur. Conf. Artificial Intelligence, A. G. Cohn, Ed., 1994, pp. 211-215.
  49. "Fast deterministic self-organizing maps," in Proc. ICANN'95, Int. Conf. Artificial Neural Networks, vol. II, F. Fogelman-Soulié and P. Gallinari, Eds., Nanterre, France, 1995, pp. 63-68.
  50. T. Kohonen, "Things you haven't heard about the Self-Organizing Map," in Proc. ICNN'93, Int. Conf. Neural Networks, 1993, pp. 1147-1156.
  51. K. Koskenniemi, "Two-level morphology: A general computational model for word-form recognition and production," Ph.D. thesis, Univ. Helsinki, Dept. General Linguistics, 1983.
  52. K. Lagus and S. Kaski, "Keyword selection method for characterizing text document maps," in Proc. ICANN99, Ninth Int. Conf. Artificial Neural Networks, vol. 1, 1999, pp. 371-376.

Related papers

Chapter 10 Self-organization of very large document collections
Krista Lagus

Text mining systems are developed to aid the users in satisfying their information needs, which may vary from searching answers to well-specified questions to learning more of a scientific discipline. The major tasks of web mining are searching, browsing, and visualization. Searching is best suited for answering specific questions of a well-informed user. Browsing and visualization, on the other hand, are beneficial especially when the information need is more general, or the topic area is new to the user [6]. The SOM, applied to organizing very large document collections, can aid in all the three tasks. Statistical models of documents In the vast majority of SOM applications, the input data constitute high-dimensional real feature vectors. In the SOMs that form similarity graphs of text documents, the feature sets that describe collections of words in the documents should be expressible as real vectors, too. The feature sets can simply be weighted histograms of the words, but usual...

View PDFchevron_right
Document Clustering using Self-Organizing Maps
Umar Ayub

MENDEL

Cluster analysis of textual documents is a common technique for better ltering, navigation, under-standing and comprehension of the large document collection. Document clustering is an autonomous methodthat separate out large heterogeneous document collection into smaller more homogeneous sub-collections calledclusters. Self-organizing maps (SOM) is a type of arti cial neural network (ANN) that can be used to performautonomous self-organization of high dimension feature space into low-dimensional projections called maps. Itis considered a good method to perform clustering as both requires unsupervised processing. In this paper, weproposed a SOM using multi-layer, multi-feature to cluster documents. The paper implements a SOM usingfour layers containing lexical terms, phrases and sequences in bottom layers respectively and combining all atthe top layers. The documents are processed to extract these features to feed the SOM. The internal weightsand interconnections between these layer...

View PDFchevron_right
WEBSOM – Self-organizing maps of document collections
Krista Lagus

Neurocomputing, 1998

With the WEBSOM method a textual document collection may be organized onto a graphical map display that provides an overview of the collection and facilitates interactive browsing. Interesting documents can be located on the map using a content-directed search. Each document is encoded as a histogram of word categories which are formed by the self-organizing map (SOM) algorithm based on the similarities in the contexts of the words. The encoded documents are organized on another self-organizing map, a document map, on which nearby locations contain similar documents. Special consideration is given to the computation of very large document maps which is possible with general-purpose computers if the dimensionality of the word category histograms is first reduced with a random mapping method and if computationally efficient algorithms are used in computing the SOMs.

View PDFchevron_right
LSISOM—A Latent Semantic Indexing Approach to Self-Organizing Maps of Document Collections
Nicholas Ampazis

The Self Organizing Map (SOM) algorithm has been utilized, with much success, in a variety of applications for the automatic organization of full-text document collections. A great advantage of the SOM method is that document collections can be ordered in such a way so that documents with similar content are positioned at nearby locations of the 2-dimensional SOM lattice. The resulting ordered map thus presents a general view of the document collection which helps the exploration of information contained in the whole document space. The most notable example of such an application is the WEBSOM method where the document collection is ordered onto a map by utilizing word category histograms for representing the documents data vectors. In this paper, we introduce the LSISOM method which resembles WEBSOM in the sense that the document maps are generated from word category histograms rather than simple histograms of the words. However, a major difference between the two methods is that in WEBSOM the word category histograms are formed using statistical information of short word contexts whereas in LSISOM these histograms are obtained from the SOM clustering of the Latent Semantic Indexing representation of document terms.

View PDFchevron_right
WEBSOM-Self-Organizing Maps of Document Collections1
Samuel Kaski

Neurocomputing, 1998

View PDFchevron_right

Related topics

  • High Dimensional Data
  • Self Organized Map

    • Cited by

    The Topic Analysis of Hospice Care Research Using Co-word Analysis and GHSOM Yu-Hsiang Yang, Huimin Bhikshu, and Rua-Huan Tsaih
    Huimin bhikshu

    Intelligent Computing and Information Science, Communications in Computer and Information Science, 2011

    The purpose of this study was to propose a multi-layer topic map analysis of palliative care research using co-word analysis of informetrics with Growing Hierarchical Self-Organizing Map (GHSOM). The topic map illustrated the delicate intertwining of subject areas and provided a more explicit illustration of the concepts within each subject area. We applied GHSOM, a text-mining Neural Networks tool, to obtain a hierarchical topic map. The result of the topic map may indicate that the subject area of health care science and service played an importance role in multidiscipline within the research related to palliative care.

    View PDFchevron_right
    INSYDER: a content-based visual-information-seeking system for the Web
    Harald Reiterer

    International Journal on Digital Libraries, 2005

    This paper presents INSYDER, a content-based visual-information-seeking system for the Web. The Web can be seen as one huge digital library offering a variety of very useful information for business analysts. INSYDER addresses these possibilities and offers powerful retrieval and visualisation functionalities. The main focus during the development was on the usability of the system. Therefore, a variety of well-established visualisation components were employed to support the user during the information-seeking process (e.g. visual query, result table, bar graph, segment view with tile bars, and scatterplot). Also, the retrieval aspects were developed with the goal of increasing the usability of the system (e.g. natural language search, content-based classification, relevance feedback). Extensive evaluations of the retrieval performance and the usability of the visualisation were conducted. The results of these evaluations offered many helpful insights into developing a new visual-information-seeking system called VisMeB.

    View PDFchevron_right
    Demonstrating conceptual dynamics in an evolving text collection
    Sándor Darányi

    Journal of the American Society for Information Science and Technology, 2013

    Based on real world user demands, we demonstrate how animated visualisation of evolving text corpora displays the underlying dynamics of semantic content. To interpret the results, one needs a dynamic theory of word meaning. We suggest that conceptual dynamics as the interaction between kinds of intellectual, emotional etc. content, and language, is key for such a theory. We demonstrate our methodology by two-way seriation which is a popular technique to analyse groups of similar instances and their features, as well as the connections between the groups themselves. The two-way seriated data may be visualised as a two-dimensional heat map or as a three-dimensional landscape where colour codes or height correspond to the values in the matrix. In this paper we focus on two-way seriation of sparse data in the Reuters-21568 test collection. To achieve a meaningful visualisation thereof we introduce a compactly supported convolution kernel similar to filter kernels used in image reconstruction and geostatistics. This filter populates the high-dimensional sparse space with values that interpolate nearby elements, and provides insight into the clustering structure. We also extend two-way seriation to deal with online updates of both the row and column spaces, and, combined with the convolution kernel, demonstrate a three-dimensional visualisation of dynamics.

    View PDFchevron_right
    Neurocognitive approach to clustering of PubMed query results
    Paul Zender, Włodzisław Duch

    Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2009

    Internet literature queries return a long lists of citations, ordered according to their relevance or date. Query results may also be represented using Visual Language that takes as input a small set of semantically related concepts present in the citations. First experiments with such visualization have been done using PubMed neuronal plasticity citations with manually created semantic graphs. Here neurocognitive inspirations are used to create similar semantic graphs in an automated fashion. This way a long list of citations is changed to small semantic graphs that allow semi-automated query refinement and literature based discovery.

    View PDFchevron_right
    Analysis of Web Proxy Logs
    H. Venter

    IFIP Advances in Information and Communication Technology, 2006

    Network forensics involves capturing, recording and analysing network audit trails. A crucial part of network forensics is to gather evidence at the server level, proxy level and from other sources. A web proxy relays URL requests from clients to a server. Analysing web proxy logs can give unobtrusive insights to the browsing behavior of computer users and provide an overview of the Internet usage in an organization. More importantly, in terms of network forensics, it can aid in detecting anomalous browsing behavior . This paper demonstrates the use of a selforganising map (SOM) , a powerful data mining technique, in network forensics. In particular, it focuses on how a SOM can be used to analyse ' data gathered at the web proxy level.

    View PDFchevron_right
    580 California St., Suite 400
    San Francisco, CA, 94104
    © 2025 Academia. All rights reserved