1. Incremental Information Extraction Using RelationalDatabasesABSTRACT: Information extraction systems are traditionally implemented as a pipeline ofspecial-purpose processing modules targeting he extraction of a particular kind ofinformation. A major drawback of such an approach is that whenever a newextraction goal emerges or a module is improved, extraction has to be reappliedfrom scratch to the entire text corpus even though only a small part of the corpusmight be affected. In this paper, we describe a novel approach for informationextraction in which extraction needs are expressed in the form of database queries,which are evaluated and optimized by database systems. Using database queriesfor information extraction enables generic extraction and minimizes reprocessingof data by performing incremental extraction to identify which part of the data isaffected by the change of components or goals. Furthermore, our approachprovides automated query generation components so that casual users do not haveto learn the query language in order t perform extraction. To demonstratethe feasibility of our incremental extraction approach, we performed experimentsto highlight two important aspects of an information extraction system: efficiencyand quality of extraction results. Our experiments show that in the event ofdeployment of a new module ,our incremental extraction approach reduces theprocessing time by 89.64 percent as compared to a traditional pipeline approach. Bapplying our methods to a corpus of 17 million biomedical abstracts, ourexperiments show that the query performance is efficient forreal-time applications. Our experiments also revealed that our approach achieveshigh quality extraction results.
2. EXISTING SYSTEMInformation extraction has been an active research are over the years. The mainfocus has been on improving the accuracy of the extraction systems, and IE hasbeen seen as an one-time execution process. Such paradigm is inadequate for real-world applications when IE is seen as long running processes. An example of areal-world application of IE is the extraction from evolving text [4], [5], such asthe frequent update of the content of web documents. Hence, there is a need tominimize reprocessing of the text corpora. In our case, we assume the text corporato be static. While new documents can be added to our text collection, the contentof the existing documents are assumed not to be changed, which is the case forMedline abstracts. Our focus is on managing the processed data so that in the eventof the deployment of an improved component or a new extraction goal, theaffected subset of the text corpus can be easily identified. .PROPOSED SYSTEM:Our proposed framework also follows traditional IE approaches in terms of firstpreprocessing the corpus and then performing extraction. However, our frameworkalso manages the intermediate processing output such as the parse trees andsemantic information using RDBMS. In the event of a deployment of an improvedcomponent or a change of extraction goals, our approach only requires the newmodule to be applied to the text collection. The intermediate processing data arethen inserted into the parse tree database so that both the new and existingprocessing data can be utilized for extraction.To address the high computational cost associated with extraction, documentfiltering is a common approach in which only the promising documents areconsidered for extraction These promising documents are documents that arerelevant for extraction. Such an approach can potentially miss out documents thatshould have been used for extraction. In our filtering approach, sentences areselected solely based on the lexical clues that are provided in a PTQL query. Thisfiltering process utilizes the efficiency of IR engines so that a complete scan of theparse tree database is not needed without sacrificing any sentences that shouldhave been used for extraction