International Business Machines Corporation

Thomas A Hampp-Bahnmueller

Michael Baessler

Yannick Saillet

Jojo Joseph

Pavlo Petrenko

Peng Hui Jiang

Sachindra Joshi

Yan Xu

Albert Maier

Thomas Schaeck

Raghuram Krishnapuram

Kenney Ng

Danish Contractor

Daniel Pittner

Sebastian Nelke

Oliver Suhre

Peter Gerstl

David E Johnson

Pi Jun Jiang

Michael Bässler

Amir Jaibaji

Dana W Morris

Dieter Schieber

Andrea Elias

Thilo Goetz

Brent Benton

Several aspects for optimizing unstructured document analysis comprise operating a document system, where the document system comprises a plurality of documents comprising unstructured content and a full-text index; receiving a request to identify documents comprising a type of data elements; selecting a sample out of the plurality of documents; determining data elements of the type in the sample of documents; determining an indicator context expression for the type of data elements out of the determined data elements of the type; determining a query for searching, using a search engine, the full-text index using the indicator context expression; and determining the documents in the document system being compliant to the determined query.

Smart identification of indicator text with full-text search or optimized document analysis

The present disclosure relates to a method of metadata enrichment using an enrichment comprising multiple steps. The method comprises: determining for an input data asset a metadata value descriptive of the input data asset. Characteristics of the metadata value of the input data asset may be determined. At least one informativeness score of the metadata value of the input data asset may be computed using the determined characteristics. An execution of the enrichment step may be skipped in case an input characteristic of the enrichment step is not part of the determined characteristics. In case the input characteristic of the enrichment step is part of the determined characteristics, the enrichment step may be adapted and executed or the enrichment step may be executed without adaptation. Labels resulting from the executed enrichment steps may be combined for providing one or more labels of the data asset.

Optimizing metadata enrichment of data assets

A method for providing one or more random sample documents from a corpus of documents using a search engine is provided. The providing of each of the random sample documents comprises selecting randomly a time window from a set of time windows. A search query is sent to the search engine defining a search for documents of the corpus with time-stamps within the time window defined by the randomly selected time window. In response to the sending of the search query, a search result is receiving from the search engine. The search result comprises a set of the documents of the corpus with time-stamps within the time window. One of the documents comprised by the received set of documents is then selected randomly.

Random sampling from a search engine

Techniques are described relating to unstructured document processing. An associated computer-implemented method includes identifying a plurality of deduplicated data blocks associated with a collection of unstructured documents. The method further includes sorting the plurality of deduplicated data blocks in descending order based upon at least one block frequency metric, selecting a highest sorted unprocessed deduplicated data block, applying text analytics to the selected deduplicated data block, and applying at least one result of the text analytics to any document among the collection of unstructured documents including the selected deduplicated data block. The method is terminated responsive to satisfaction of at least one stopping condition.

Analyzing deduplicated data blocks associated with unstructured documents

A method for providing one or more random sample documents from a corpus of documents using a search engine is provided. The providing of each of the random sample documents comprises selecting randomly a time window from a set of time windows. A search query is sent to the search engine defining a search for documents of the corpus with time-stamps within the time window defined by the randomly selected time window. In response to the sending of the search query, a search result is received from the search engine. The search result comprises a set of the documents of the corpus with time-stamps within the time window. One of the documents comprised by the received set of documents is then selected randomly.

A method for processing electronic documents comprises an iteration including: (i) applying, by a computer device, a first statistical test process to a first subset of the documents, the first statistical test process estimating whether or not content of the documents of the first subset comply with a predefined criterion; (ii) in response to a result of the first statistical test process, estimating, by the computer device, that the documents of the first subset do not comply with the criterion, selecting, by the computer device, a part of the documents of the first subset, and moving, by the computer device, the part of the documents to a second subset of the documents; and (iii) applying, by the computer device, a second statistical test process to the second subset of the documents, the second statistical test process calculating at least one statistical metric related to the documents of the second subset.

Processing electronic documents

A method and a computer-readable medium for method for searching a plurality of documents. Each document is structured into a set of blocks and each block is associated with a block ID. The method includes receiving a search query including a search term having at least one search term attribute; identifying at least one block ID based on a correlation between the at least one search term attribute and the set of blocks; and identifying at least one document based on a correlation between the set of blocks and the documents. Methods for generating a data structure for searching documents are also described.

Method and system for performing search queries using and building a block-level index

Providing access to media data shared by multiple users. A predefined edge weight is assigned to each edge of a linked data structure based on a dependency category of the edge. A first access rating value is assigned to each node. A rating residue value is calculated as the difference between the two first access rating values of the nodes connected by each edge. The data structure is traversed from a seed node, and for each edge traversed, calculating a second access rating value using an edge weight value and the first access rating value. Repeating until the rating residue values meet a predefined convergence criterion. The nodes having access rating values meeting a predefined data removal criterion are selected from the nodes of the linked data structure. The data entities corresponding to the selected nodes are then removed.

Communication method and system for accessing media data

Provided are techniques for creating an inverted index for features of a set of data elements, wherein each of the data elements is represented by a vector of features, wherein the inverted index, when queried with a feature, outputs one or more data elements containing the feature. The features of the set of data elements are ranked. For each feature in the ranked list, the inverted index is queried for data elements having the feature and not having any previously selected feature and a cluster of the data elements is created based on results returned in response to the query.

Clustering a collection using an inverted index of features

An embodiment of a method for enhanced content browsing includes loading a web page in a user interface; detecting entities of a first specified type in the web page by an analysis service; tagging the detected entities in the web page; calling an action service associated with the analysis service when a detected entity is activated; and displaying a result of the action service in the user interface. Embodiments of systems for enhanced content browsing are also provided.

Enhanced content web browsing

A framework for information extraction from natural language documents is application independent and provides a high degree of reusability. The framework integrates different Natural Language/Machine Learning techniques, such as parsing and classification. The architecture of the framework is integrated in an easy to use access layer. The framework performs general information extraction, classification/categorization of natural language documents, automated electronic data transmission (e.g., E-mail and facsimile) processing and routing, and plain parsing. Inside the framework, requests for information extraction are passed to the actual extractors. The framework can handle both pre- and post processing of the application data, control of the extractors, enrich the information extracted by the extractors. The framework can also suggest necessary actions the application should take on the data. To achieve the goal of easy integration and extension, the framework provides an integration (outside) application program interface (API) and an extractor (inside) API.

Growing community of inventors

Stuttgart, Germany

Thomas A Hampp-Bahnmueller