International Business Machines Corporation

Google Inc.

Jan Hendrik Pieper

Daniel Gruhl

Varun Bhagwan

Mark Andrew Smith

Tyrone W A Grandison

Andrew Stephen Tomkins

Marcus Felipe Fontoura

Alfredo Alba

Mihail Corneliu Constantinescu

Nadav Eiron

Daniel Norin Meredith

Subashini Balachandran

Joerg Meyer

Tyrone Wilberforce Andre Grandison

Krishna Tatavarthi

Bjoern Carlin

Hsiang-ling Lin

A method resolves ambiguous spotted entity names in a data corpus by determining an activation level value for each of a plurality of nodes corresponding to a single ambiguous entity name. The activation levels for each of the nodes may be modified by inputting outside domain knowledge corresponding to the nodes to increase the activation value of the nodes, spotting entity names corresponding to the nodes to increase the activation value of the nodes, searching the data corpus to spot newly posted entity names to increase the activation value of the nodes, and searching the data corpus to reduce or deactivate the activation value of the nodes by eliminating false positives. The ambiguous entity name is assigned to the node determined to have the highest activation level and is then outputted to a user.

Holistic disambiguation for entity name spotting

Systems and methods for deleting non-key values from an index distributed over a plurality of computing devices maintains a non-key master list that includes the non-key values that are stored on each of the plurality of computing devices and receives a list of non-key values to delete. The systems and methods further intersect the list of non-key values to delete with the non-key master list, creating a first delete list for a first one of the plurality of computing devices that includes non-key values to be deleted that are stored on the first computing device. The systems and methods further transmit the first delete list to the first computing device and update the non-key master list based on the list of non-key values to delete.

Processing bulk deletions in distributed databases

A system and method of indexing a plurality of entities located in a taxonomy, the entities comprising sets of terms, comprises receiving terms in an index structure; building a posting list for an entity with respect to the locations of the set of terms defining the entity and data associated with the respective terms; and indexing a name of a group comprising the entities within this group at the location of the entities with the data of the group comprising the name of the respective entity at each location. The building of the posting list comprises storing the location of the term and data associated with the term in an entry in the posting list for the term. The method comprises indexing aliases of the name of the group comprising the term, and using an inverted list index to associate data with each occurrence of an index term.

Method and framework to support indexing and searching taxonomies in large scale full text indexes

Embodiments of the invention relate to creating an operating system and file system independent incremental data backup. A first data backup of a source system and second version of the data on the source system is received. A second data backup of the second version of the data is created by determining differences between the first data backup and the second version of the data. Each portion of the second version of the data that is the same as a portion of the first data backup is referenced in the second data backup. Each portion of the second version of the data that is different than all portions of the first data backup is included in the second data backup. The second data backup is appended to the first data backup to create an incremental data backup.

Operating system and file system independent incremental data backup

Embodiments of the invention provide a system and method for determining preferences from information mashups and, in particular, a system and method for constructing a ranked list from multiple sources. In an exemplary embodiment, the system and method tunably combines multiple ranked lists by computing a score for each item within the list, wherein the score is a function of the associated rank of the item within the list. In one exemplary embodiment, the function is equal to 1/(n^(1/p)), where p is a tuning parameter that enables selection between responsiveness in the combined ranking to one candidate ranked highly in one source versus responsiveness in the combined ranking to a candidate with lower but broader support among the various sources ranking the candidates.

System and method for constructing targeted ranking from multiple information sources

Data deduplication compression in a streaming storage application, is provided. The disclosed deduplication process provides a deduplication archive that enables storage of the archive to, and extraction from, a streaming storage medium. One implementation involves compressing fully sequential data stored in a data repository to a sequential streaming storage, by: splitting fully sequential data into data blocks; hashing content of each data block and comparing each hash to an in-memory lookup table for a match, the in-memory lookup table storing all hashes that have been encountered during the compression of the fully sequential data; for each data block without a hash match, adding the data block as a new data block for compression of fully sequential data; and encoding duplicate data blocks using the in-memory lookup table into data segments.

Data deduplication for streaming sequential data storage applications

A method, system, and article for compressing an input stream of uncompressed data. The input stream is divided into one or more data segments. A hash is applied to a first data segment, and an offset and length are associated with this first segment. This hash, together with the offset and length data for the first segment, is stored in a hash table. Thereafter, a subsequent segment within the input stream is evaluated and compared with all other hash entries in the hash table, and a reference is written to a prior hash for an identified duplicate segment. The reference includes a new offset location for the subsequent segment. Similarly, a new hash is applied to an identified non-duplicate segment, with the new hash and its corresponding offset stored in the hash table. A compressed output stream of data is created from the hash table retained on storage media.

Method and apparatus for data compression

A system for transforming domain specific unstructured data into structured data including an intake platform controlled by feed back from a control platform. The intake platform includes an intake acquisition module for acquiring data building baseline data related to a domain and problem of interest, an intake pre-processing module, an intake language module, an intake application descriptors module, and an intake adjudication module. The control platform includes a control data acquisition module, a control data consistency collator, a control auditor, a control event definition and policy repository, an error resolver, and an output that outputs results of the workflow into structured data enabled to be used in data analytics.

System for monitoring global online opinions via semantic extraction

The invention provides a method and apparatus for determining sizing of chunk portions in data de-duplication. The method chunks input data into segments where each segment has a first size, assigns an identifier to each of the data segments, assigns an index to each of the identifiers, creates a suffix structure and a longest common prefix structure from the indexes, detects repeated sequences of indexes and non-repeated indexes from the suffix structure and the longest common prefix structure, determines a second size based on said detected repeated sequences and non-repeated indexes, and chunks the input data into a second plurality of data segments each having the second size.

Growing community of inventors

San Jose, CA, United States of America

Jan Hendrik Pieper