International Business Machines Corporation

Sebastian Goeser

Roland Seiffert

Peter Gerstl

Jochen F Doerre

Roy Jefferson Byrd

Adrian Mueller

A method and system determines discrete policy target groups for information objects stored in an enterprise IT system. The method and system provide cleansed information about information objects stored on the enterprise IT system. Criteria for sorting the information objects is determined. Initial sorting of the information objects is carried out, resulting in an initial set of clusters. The information objects are clustered into discrete policy target groups based on the information about the information objects and the initial set of clusters, and human-understandable names and definite descriptions for policy target groups are computed.

Mining of policy data source description based on file, storage and application meta-data

A method, computer system, and a computer program product for processing a set of data objects of one or more users stored in a computer system is provided. The present invention may include providing data object properties and user properties of the set of data objects. The present invention may include determining a combined property of the data object properties and user properties. The present invention may include determining time dependent properties and at least two metrics, wherein a first metric indicates the deletability of a cluster and a second metric indicates the quality of the cluster, and wherein the first metric is determined using time dependent properties of the cluster. The present invention may include inputting the set of combined properties to a clustering algorithm and using the clustering algorithm to cluster the data objects to identify a best deletable cluster to determine if a data object is deletable.

Method to estimate the deletability of data objects

A method for obtaining structured information from a natural language text by aggregation. The aggregation includes steps of reading in a natural language text, recognizing sentences and sentence constituents for the text read in in accordance with an extractor, which allocates an information-bearing structure to each sentence or sentence constituent, allocating a substitution symbol to one or more sentences or sentence constituents, introducing the results into a further information-bearing structure which comprises functors and argument terms and is described by a grammar.

Method for obtaining structured information exists in special data format from a natural language text by aggregation

This mechanism relates to a method within the area of information mining within a multitude of documents stored on computer systems. More particularly, this mechanism relates to a computerized method of generating a content taxonomy of a multitude of electronic documents. The technique proposed by the current invention is able to improve at the same time the scalability and the coherence and selectivity of taxonomy generation. The fundamental approach of the current invention comprises a subset selection step, wherein a subset of a multitude of documents is being selected. In a taxonomy generation step a taxonomy is generated for that selected subset of documents, the taxonomy being a tree structured taxonomy hierarchy. Moreover this method comprises a routing selection step assigning each unprocessed document to the taxonomy hierarchy based on largest similarity.

Taxonomy generation for document collections

Natural language processing is used in a method and apparatus for the computer-backed control of business processes and process sequences. The method involves the automatic checking of at least one of the conditions of an activity of the business process using a method of natural language processing. Control of the process conditions of activities which currently checked by humans can be checked automatically.

Method and apparatus for process control by using natural language

A method is described for a computerized search for words in an electronic database with a large number of documents stored in memory. With this method, a Boolean retrieval method is used to determine in which of a large number of documents an initial word meets a Boolean condition. A probabilistic retrieval method is then used to determine in which of the documents fulfilling the Boolean condition, the relevance of appearance of a second word exceeds a specified value. The two retrieval methods use different indexes for this. The disadvantages normally found with this are avoided by the two different indexes having a common element that can be processed by both retrieval methods.

Growing community of inventors

Stuttgart, Germany

Sebastian Goeser