**3.2 Mapping experiments among knowledge trees**

In order to prove mapping among the knowledge trees that represent the ontologies which are taken into account in each node of the system, two experiments have been carried out.

(a) OOSS KC nodes topics mapping (b) FLAT KC nodes topics mapping

Fig. 10. Mapping between topics of KnowCat nodes

In the first experiment of this series, again, we start up with KC node on OOSS, although in this case two new nodes have been created from it with knowledge trees equal to the original. The documents of the initial node have been divided up into the recently created

number of topics is higher and the number of documents per topic is lower, groupings

(a) OOSS KC node documents similarity (b) CS KC node documents similarity

In order to prove mapping among the knowledge trees that represent the ontologies which are taken into account in each node of the system, two experiments have been carried out.

(a) OOSS KC nodes topics mapping (b) FLAT KC nodes topics mapping

In the first experiment of this series, again, we start up with KC node on OOSS, although in this case two new nodes have been created from it with knowledge trees equal to the original. The documents of the initial node have been divided up into the recently created

Fig. 9. Automatic grouping of documents by topics of knowledge area

**3.2 Mapping experiments among knowledge trees** 

Fig. 10. Mapping between topics of KnowCat nodes

appear like smaller light blocks.

ones, so that every couple of new homologous topics has a similar number of different documents, but relevantly similar. Later the WWV of the topics of the new nodes have been calculated and have been compared with each other. A graph (see Fig. 10 left) has been produced with the values of similarity obtained, where the lines of colour blocks correspond to the topics of one node and the columns to the other. The topics have been organised into both dimensions in order that the homologous topics are in the same position in the corresponding entries of the table. Like in previous graphs, the highest values of similarity are represented by the lighter colours.

As can be seen in the image of this first experiment, the highest grades of similarity -blocks of light colour- are over the diagonal in almost every case. With the proposed approach, this means that it is possible to identify the branches of the knowledge trees that contain documents dealing with the same topics.

For the second experiment two KC nodes on FLAT that have different trees to organise the knowledge have been used. Again the WWV of the topics have been calculated from the documents included within them and the grades of similarity have been calculated from the topics of different nodes comparing their corresponding vectors. The result is shown in a graph (see Fig. 10 right) where the topics of one node are in the axis of abscissa and the other in the organised axis. As on other occasions, the grade levels of similarity are shown in colour blocks, where again, the higher the value of coefficient, the lighter the colours. In this case, the pair of topics that are considered linked to each other through their contents by means of a manual analysis by an expert on the subject have been marked with a cross.

As a result of this second experiment, it can be seen that most of the associations made by an expert fit in over light colour blocks and that every light block is found in topic pairs associated by the expert. Therefore, it is possible to identify the proposed procedure and the topics that deal with related issues in different knowledge trees automatically.

#### **3.3 Automatic association experiment among knowledge nodes**

Starting from the documents included in five KC nodes, the one belonging to CS used in the first group of experiments, the two OOSS prepared for the previous group and the two FLAT used in the same group, a WWV has been established for each of them. In every case, the documents included in the nodes are different. By comparing these WWV a graph (see Fig. 11) has been obtained, in which each line of blocks, vertical and horizontal,

Fig. 11. Grouping of KC nodes per topics

Digestion of Knowledge in a KM System to Reveal Implicit Knowledge 115

Spanish National Plan of R+D, project code TIN2008-02081/TIN, and by the CAM

Adomavicius, G. & Tuzhilin, A. (2005). *Toward the next generation of recommender systems:* 

Alamán, X. & Cobos, R. (1999). *KnowCat: a Web Application for Knowledge Organization*, LNCS

Brants T. (2004). *Natural Language Processing in Information Retrieval*. In proceedings of CLIN

Carreras, X., Chao, I., Padró L. & Padró M. (2004). *FreeLing: An Open-Source Suite of Language* 

Chang, G., Healey, M., McHugh, J., Wang, J. (2001). *Mining the World Wide Web: An* 

Cobos, R. (2003). *Mechanisms for the Crystallisation of Knowledge, a proposal using a collaborative* 

Cobos, R., Pifarré, M., (2008). *Collaborative knowledge construction in the web supported by the* 

Gansner, E., North, S. (2000). *An open graph visualization system and its applications to software* 

Gruber, T. R. (1993). A *Translation Approach to Portable Ontology Specifications*. Knowledge

Kiryakov, A., Popov, B., Terziev, I., Manov, and D., Ognyanoff, D. (2004). *Semantic* 

Moreno-Llorena, J. (2008). *Collaborative Knowledge Management By Means Semantic Information*. Doctoral dissertation. Universidad Autónoma de Madrid. Moreno-Llorena, J., Alamán, X. (2005). *A Proposal of Design for a Collaborative Knowledge* 

Moreno-Llorena, J., Alamán, X., Cobos, R. (2009a). *Modeling of User Interest Based on Its* 

Moreno-Llorena, J., Alamán, X., Cobos, R.(2009b). *Establishment and Maintenance of a* 

*Annotation, Indexing, and Retrieval*. Journal of Web Sematics 2, Issue 1, Elsevier 49-

*Management System by means of Semantic Information*. In: Navarro-Prieto, R. et al., HCI related papers of Interacción 2004, Springer, Dordrecht, The Netherlands,. 307-

*Interaction with a Collaborative Knowledge Management System*. Lecture Notes in Computer Science (ISSN: 0302-9743, impreso y 1611-3349, en línea; ISBN: 978-3-642-

*Knowledge Network by Means of Agents and Implicit Data* . Data Mining and Multiagent Integration (ISBN: 978-1-4419-0523-9, impreso y 978-1-4419-0522-2, en

*KnowCat system*. Computers & Education, Vol.50, No. 3, (April 2009), pp. 962-978,

on Knowledge and Data Engineering No. 17, pp. 734-749, 2005

Baeza, R., Ribeiro, B. (1999). *Modern Information Retrieval*. Addison Wesley, 1999.

*Analyzers*. In proceedings of LREC 2004 Lisbon, Portugal. 2004.

*system*. Doctoral dissertation. Universidad Autónoma de Madrid.

*engineering*. Software – Practice and Experience, 30:1203–1233, 2000.

Geroimenko, V., Chen, C. (2002). *Visualizing the Semantic Web*. Springer, 2002.

02579-2), 330-339, Springer Berlin / Heidelberg. 2009.

*a survey of the state-of-the-art and possible extensions*, In : IEEE Transactions

(Autonomous Community of Madrid), project code S2009/TIC-1650.

1727. Eds. Chen, P.P. et al. Springer, 348-359. 1999.

Berners-Lee, T. (2000). *Semantic Web - XML2000*. Available from http://www.w3.org/2000/Talks/1206-xml2k-tbl

2004 Antwerp, Belgium, 1–13. 2004.

Acquisition, 5(2), pp. 199-220, 1993.

línea),155-166, Springer US. 2009.

ISSN 0360-1315.

79, 2004.

319, 2005.

*introduction search approach*. Kluwer, 2001.

**6. References** 

corresponds to a node. Like in previous diagrams, the lighter colours represent a greater similarity.

In the graph we can see that the level of similarity among the WWV of the nodes that deal with the same topics are high compared with the ones obtained where comparing the node vectors on different topics. This means -using this technique- that it is possible to identify nodes that deal with similar contents and to distinguish them from others on different subjects.
