Desarrollo de algoritmos para el descubrimiento de patrones secuenciales maximales
Information of a document is described by words in a sequential way. From this point of view, the knowledge transmitted in a text is sequentially structured by its author. Thus, text it is not only useful for expressing the author’s ideas, but also it is useful for discovering new knowledge from the...
| Autor: | |
|---|---|
| Tipo de recurso: | tesis doctoral |
| Estado: | Versión aceptada para publicación |
| Fecha de publicación: | 2007 |
| País: | México |
| Institución: | Instituto Nacional de Astrofísica, Óptica y Electrónica |
| Repositorio: | Repositorio Institucional del INAOE |
| Idioma: | español |
| OAI Identifier: | oai:inaoe.repositorioinstitucional.mx:1009/600 |
| Acceso en línea: | http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/600 |
| Access Level: | acceso abierto |
| Palabra clave: | info:eu-repo/classification/Secuencia/Sequence info:eu-repo/classification/Reconocimiento de patrones/Pattern recognition info:eu-repo/classification/Extracción secuencial de patrones/Sequential pattern mining info:eu-repo/classification/cti/7 info:eu-repo/classification/cti/33 info:eu-repo/classification/cti/3304 info:eu-repo/classification/cti/330405 |
| Sumario: | Information of a document is described by words in a sequential way. From this point of view, the knowledge transmitted in a text is sequentially structured by its author. Thus, text it is not only useful for expressing the author’s ideas, but also it is useful for discovering new knowledge from the frequent sequential order of the words in the text. The latter aspect has been part of our motivation, since there are a big amount of electronic documents that could be useful for discovering sequential patterns, which often cannot be seen at first sight and could be helpful for text analysis; achieving in this manner a text mining process. Since the number of frequent sequences can be huge, it is possible considering only those which are not contained in another frequent sequence, it means, the Maximal Frequent Sequences (MFS’s) can be considered as a compact representation of all frequent sequences. When a MFS is found, the words, length and frequency of such MFS are determined by the text. The last characteristic is very important because the frequency allows having support for that MFS. Other important feature of the MFS’s is that they can be extracted independently of the language. Frequent sequences preserve, in some way, the natural sequential order of the text. Moreover, by its legibility, MFS’s are easy to understand by humans. The above mentioned characteristics make MFS’s suitable to be applied in specific text mining tasks like document clustering and classification; or in tasks like information retrieval, question answering, text summarization, etc. However, the MFS discovering problem has received special attention due to the big amount of combinations that have to be reviewed for discovering such patterns. For example, if a frequent sequence has 100 elements then 2 100 − 1 ≈ 10 30 combinations have to be reviewed for extracting such pattern. This problem has been classified as NP-hard. This dissertation deals with the problem of discovering MFS’s in text. It is important to remark, that the main objective of this dissertation was to propose algorithms for improving the search of MFS’s from textual information. However, the proposed algorithms can work over any other kind of sequential information like DNA sequences, WEB logs, etc.; it is, with objects that describe a sequential behavior through symbols. |
|---|