Article Preview
Top1. Introduction
Data warehousing systems and OLAP technologies are efficient to analyze numerical data. However, a great part of enterprise data is presented in textual data, such as: Reports, e-mails, etc. These data are faintly exploited by the current business intelligence systems. To allow efficient Text OLAP analysis, it is essential to define new techniques for integrating semantic of textual data in the OLAP process. One of the most important challenges in this context is the aggregation of textual data. Indeed, the aggregation of numerical data is performed using standard aggregation functions such as: Sum, Average, Min, Max, etc. However, these functions are not suitable to analyze textual data due to their unstructured nature. Therefore, it is important to improve the traditional data cube model and to define new aggregation techniques appropriate to textual data.
OLAP systems allow navigating through multidimensional cubes from one view to another in an interactive way. In addition, they proved their effectiveness to analyze large volumes of data. OLAP data analysis allows expressing complex queries and viewing aggregated results relevant to the decision-making process. However, to analyze textual data, it is often necessary to resort to different areas interested in text processing, such as information retrieval (IR) and text mining (TM). Among the proposals for extending the classical data cube to support textual data, we find the works of (Zhang, Zhai & Han, 2009, 2013) and (Mothe, Chrisment, Dousset & Alaux, 2003), that use techniques from data mining to define their models Topic Cube, MitexCube and DocCube, respectively. The works of (Lin, Ding, Han, Zhu & Zhao, 2008), (Yu et al., 2009) and (Bringay et al., 2011) use classical IR techniques to define textual analysis measures. Pérez, Berlanga, Aramburu and Pedersen. (2007) propose a contextualized cube called R-Cube which is also based on classical IR measures. Finally, some studies present new models with new concepts, as in (Boukraa, Boussaid, Bentayeb & Zegour, 2013) where the authors propose models based on the object paradigm. However, most of these works propose approaches based on the classical IR measure Tf-Idf, which does not take into account the text semantics in Text OLAP.
Furthermore, Text Mining is defined as a knowledge-intensive process in which a user interacts with a document collection over time by using a suite of analysis tools (Felman & Sanger, 2007). Automatic document classification is one of TM techniques that allows the creation of groups of documents assembled according to similarity relations between them.
We think that these techniques can be exploited to define more sophisticated OLAP techniques that are appropriate to textual data. Text OLAP should allow to go beyond a simple extraction of knowledge from text documents: it should allow navigational analysis through text cubes. In this context, we draw on IR and TM techniques to define new Text OLAP operators, especially aggregation operators, adapted to textual data.