Dramatically compressing the size of a dataset while preserving the ability to decompress a single record or field.

Using co-occurring topics from different topic models to aid in entity disambiguation; extending conventional LDA topic modeling in so that each component is treated as conditionally-independent.

Using topic models to develop relationships between topics, events and entities; and further use of that data to detect and capture events.

Employing a co-occurrence knowledge base to allow a user to submit partial or complete queries, which are processed on-the-fly to detect entities, their misspelled variations, and other conceptual features.

Extending existing search suggestion mechanisms to employ an entity co-occurrence knowledge base, a trends co-occurrence knowledge base, fuzzy matching modules and entity extraction modules.

Extending traditional feature disambiguation systems by providing a feedback mechanism to take human verification of the disambiguated results and update the knowledge base on-the-fly.

Providing fault-tolerance in a distributed system by automatically detecting failures and recovering from them by moving processing modules and their dependencies to other nodes in the system.

A system of multiple network segments with bandwidth and latency tiers, applied to a distributed in-memory data platform.

A search suggestion generation mechanism that employs a co-occurrence knowledge base to suggest new search queries and possible search expansions or suggestions.

Employing an evolving and efficiently linkable feature knowledge base capable of storing secondary features to disambiguate entities.

An alert mechanism based on user-specified disambiguated features, providing a framework to support collaborative knowledge sharing among different users.

Leveraging user-specified disambiguated features to provide better relevance and precision than traditional systems, as well as the ability to build alert specifications around secondary features.

Technology to automate the deployment, installation and configuration of data, metadata and software of the primary data store of a distributed database.

Extending the search capabilities of contentment management systems with additional managed properties obtained through natural language processing of the underlying unstructured text content.

Granting users the ability to specify a query and a detailed scoring/ranking specification as separate aspects in a search request.

Generating facets and corresponding frequency counts as part of the search process by running them as part of the query in an in-memory database.

Associating extracted facts with other features retrieved from the text and employing a fact-template store containing commonly used fact-sentence structures.

Broad system architecture of our in-memory computing platform, FinchDB.

Learning different event types by extracting entities and topic vectors and comparing them to existing entities and topic vectors; developing new knowledge base entries as new entities and topic vectors are discovered.

An on-the-fly, analytics-based search mechanism for live and large data streams.

Computing topic relatedness based on topic models and using those relationships as a supplementary feature in various data and text analytics.

Granting users the ability to find entities of interest via entity co-occurrence, even as the data is updated and new information is discovered.

Ensuring each database query can specify which analytics modules and parameters are to be applied on-the-fly to intermediate query answers without having to first retrieve data out of the database.

A feature co-occurrence knowledge base that exploits the full range of state-of-the-art natural language process tools.

A generalizable framework to train customized and source-specific event templates and models for different input data streams.