Citation Context

...hough it is just slightly more than 1.1% of raw HTML tables. Previous work on HTML tables focused on the problem of recognizing good tables or extracting additional information from individual tables =-=[26, 29, 30]-=-. In this paper we consider a corpus of tables that is five orders of magnitude larger than the largest one considered to date [26], and address two fundamental questions: (1) what are effective metho...

Citation Context

...Web research [8]. Additionally, there are strong links to knowledge extraction from table structures. A general overview of work on recognizing tables and drawing inferences from them can be found in =-=[27]-=-. [21] is an approach for automatic generation of F-Logic frames out of tables, which subsequently supports the automatic population of ontologies from table-like structures. Different approaches for ...

Citation Context

...is topic in the survey to contain its scope and volume. On the topic of table extraction there is an extensive research literature spanning many different communities, including the document analysis =-=[84, 109, 134, 222]-=-, information retrieval [164], web [62, 96], database [36, 165], and machine learning [164, 216] communities. A survey can be found in [84].1.4 Types of Unstructured Sources 1.4 Types of Unstructured...

Citation Context

...orts the automatic table metadata extraction and table search. Researchers in the automatic table extraction field largely focus on analyzing the table structure in a specific document media. Zanibbi =-=[15]-=- provides a survey with detailed description of each method. All the methods can be divided into three categories: pre-defined layout based [10], heuristics based [7][9][11][16], and statistical based...

Citation Context

...tured data elements on the Web. A number of authors have studied the problem of information extraction from a single table, some of which serve a role similar to that of the WebTables relation filter =-=[3, 6, 10, 14, 15]-=-. As discussed in Section 3.1, Wang and Hu detected “true” tables with a classifier and features that involved both content and layout [12]. This last paper processed the most tables, taking as input ...

Citation Context

...of the 21 tables, such that in general we conclude that our results are certainly very promising. 4 Related Work A very recent systematic overview of related work on table recognition can be found in =-=[18]-=-. Several conclusions can be drawn from this survey. Firstly, only few table models have been described explicitly. Apart from the table model of Hurst which we have applied in our approach [7, 8] the...

Citation Context

...ng the appropriate array model for a table with multi-line cells without complete rulings has proved difficult. In the last 15 years, over 200 research papers have been published on table recognition =-=[1, 2, 3, 4, 5]-=-. Most published algorithms for cell alignment treat the table as a 2-D array of cells, and attempt to identify the coordinates and contents of each cell. Methods vary depending on whether the table i...

Citation Context

...opulation computed as the sum of the populations from the states in the region. Automated “table understanding” has been the subject of research in the document analysis community for several decades =-=[10, 13]-=-. Most of these efforts end, however, after only identifying table labels and table instance data. Some researchers have described a semantic-enrichment step in the table-understanding process, but as...