Abstract: Churn prediction is performed by monitoring quality of service levels provided to customers. A time in which the customer is due to either churn or renew his agreement with the service provider may be monitored or computed. Machine learning methods may be utilized to determine a probability of churn based on historic data. Based upon the determination an output to retention personnel may be provided and an improved offer may be made to customers that are deemed in risk of churning.

Abstract: A method and system are provided of merging results in distributed information retrieval. A search manager is in communication with a plurality of components, wherein a component is a search engine working on a document collection and returning results in the form of a list of documents to a search query. The search manager submits a query to the plurality of components, receives results from each component in the form of a list of documents; estimates the success of a component in handling the query to generate a merit score for a component per query; applies the merit score to the results for the component; and merges results from the plurality of components by ranking in order of the applied merit score.

Abstract: Methods, apparatus and systems are provided that enable the generation of random regression suites for verification of a hardware or software design to be formulated as optimization problems. Solution of the optimization problems using probabilistic methods provides information on which set of test specifications should be used, and how many tests should be generated from each specification. In one mode of operation regression suites are constructed that use the minimal number of tests required to achieve a specific coverage goal. In another mode of operation regression suites are constructed so as to maximize task coverage when a fixed number of tests are run or within a fixed cost.

Abstract: A method and system for analyzing a document set (202, 420) are provided. The method includes determining a set of terms (312) from the terms of the document set that minimizes a distance measurement (405) from the given set of documents (420). The method includes using a greedy algorithm to build the set of terms incrementally, at each stage finding a single word that is closest to the document set (202, 420). The set of terms is evaluated to assess the ability to find the document set (202, 420). The set of terms are compared with expected terms to evaluate the ability to find the document set (202, 420). A measure of the ability to find a document set (202, 420) is provided by computing a distance measure (403) between a document set and an entire collection.

Abstract: Methods, apparatus and systems are provided that enable the generation of random regression suites for verification of a hardware or software design to be formulated as optimization problems. Solution of the optimization problems using probabilistic methods provides information on which set of test specifications should be used, and how many tests should be generated from each specification. In one mode of operation regression suites are constructed that use the minimal number of tests required to achieve a specific coverage goal. In another mode of operation regression suites are constructed so as to maximize task coverage when a fixed number of tests are run or within a fixed cost.

Abstract: Methods, apparatus and systems are provided that enable the generation of random regression suites for verification of a hardware or software design to be formulated as optimization problems. Solution of the optimization problems using probabilistic methods provides information on which set of test specifications should be used, and how many tests should be generated from each specification. In one mode of operation regression suites are constructed that use the minimal number of tests required to achieve a specific coverage goal. In another mode of operation regression suites are constructed so as to maximize task coverage when a fixed number of tests are run or within a fixed cost.

Abstract: A query difficulty prediction unit includes a query difficulty predictor to determine the extent of overlap between query documents received from a search engine operating on an input query and sub-query documents received from the search engine operating on sub-queries of the input query. The unit generates a query difficulty prediction from the extent of overlap.

Abstract: Test generation is improved by learning the relationship between an initial state vector for a stimuli generator and generation success. A stimuli generator for a design-under-verification is provided with information about the success probabilities of potential assignments to an initial state bit vector. Selection of initial states according to the success probabilities ensures a higher success rate than would be achieved without this knowledge. The approach for obtaining an initial state bit vector employs a CSP solver. A learning system is directed to model the behavior of possible initial state assignments. The learning system develops the structure and parameters of a Bayesian network that describes the relation between the initial state and generation success.

Abstract: A method and system for analyzing a document set (202, 420) are provided. The method includes determining a set of terms (312) from the terms of the document set that minimizes a distance measurement (405) from the given set of documents (420). The method includes using a greedy algorithm to build the set of terms incrementally, at each stage finding a single word that is closest to the document set (202, 420). The set of terms is evaluated to assess the ability to find the document set (202, 420). The set of terms are compared with expected terms to evaluate the ability to find the document set (202, 420). A measure of the ability to find a document set (202, 420) is provided by computing a distance measure (403) between a document set and an entire collection.

Abstract: A coverage-directed test generation technique for functional design verification relies on events that are clustered according to similarities in the way that the events are stimulated in a simulation environment, not necessarily related to the semantics of the events. The set of directives generated by a coverage-directed test generation engine for each event is analyzed and evaluated for similarities with sets of directives for other events. Identified similarities in the sets of directives provide the basis for defining event clusters. Once clusters have been defined, a common set of directives for the coverage-directed test generation engine is generated that attempts to cover all events in a given cluster.

Abstract: A Bayesian network correlating coverage data and input data to a test verification system for coverage directed test generation (CDG) of a device under test. In one embodiment, the Bayesian network is part of a CDG engine which also includes a data analyzer which analyzes coverage data from a current test run of a test verification system and from previous test runs to determine which coverage events from a coverage model have occurred therein, at what frequency and which ones have not yet occurred, a coverage model listing coverage events which define the goal of the test verification system and a task manager coupled to the data analyzer and the Bayesian network which refers to the coverage model and queries the Bayesian network to produce input data to achieve desired coverage events.

Abstract: A method and system for the detection of missing content in a searchable repository is provided. A system includes: a missing content query identifier (401) for identifying queries to a search engine (102) for which no or little relevant content is returned; a missing content detector (110) which clusters missing content queries by topic; and an output provider for providing details of a missing content topic.

Abstract: A method and system are provided of merging results in distributed information retrieval. A search manager (104) is in communication with a plurality of components, wherein a component is a search engine (106-108) working on a document collection and returning results in the form of a list of documents to a search query. The search manager (104) submits a query (202) to the plurality of components, receives results (213) from each component in the form of a list of documents; estimates (208) the success of a component in handling the query to generate a merit score (210) for a component per query; applies (220) the merit score (210) to the results for the component; and merges (222) results from the plurality of components by ranking in order of the applied merit score.

Abstract: Test generation is improved by learning the relationship between an initial state vector for a stimuli generator and generation success. A stimuli generator for a design-under-verification is provided with information about the success probabilities of potential assignments to an initial state bit vector. Selection of initial states according to the success probabilities ensures a higher success rate than would be achieved without this knowledge. The approach for obtaining an initial state bit vector employs a CSP solver. A learning system is directed to model the behavior of possible initial state assignments. The learning system develops the structure and parameters of a Bayesian network that describes the relation between the initial state and generation success.

Abstract: Methods, apparatus and systems are provided that enable the generation of random regression suites for verification of a hardware or software design to be formulated as optimization problems. Solution of the optimization problems using probabilistic methods provides information on which set of test specifications should be used, and how many tests should be generated from each specification. In one mode of operation regression suites are constructed that use the minimal number of tests required to achieve a specific coverage goal. In another mode of operation regression suites are constructed so as to maximize task coverage when a fixed number of tests are run or within a fixed cost.

Abstract: A query difficulty prediction unit includes a query difficulty predictor to determine the extent of overlap between query documents received from a search engine operating on an input query and sub-query documents received from the search engine operating on sub-queries of the input query. The unit generates a query difficulty prediction from the extent of overlap.

Abstract: A coverage-directed test generation technique for functional design verification relies on events that are clustered according to similarities in the way that the events are stimulated in a simulation environment, not necessarily related to the semantics of the events. The set of directives generated by a coverage-directed test generation engine for each event is analyzed and evaluated for similarities with sets of directives for other events. Identified similarities in the sets of directives provide the basis for defining event clusters. Once clusters have been defined, a common set of directives for the coverage-directed test generation engine is generated that attempts to cover all events in a given cluster.

Abstract: A Bayesian network correlating coverage data and input data to a test verification system for coverage directed test generation (CDG) of a device under test. In one embodiment, the Bayesian network is part of a CDG engine which also includes a data analyzer which analyzes coverage data from a current test run of a test verification system and from previous test runs to determine which coverage events from a coverage model have occurred therein, at what frequency and which ones have not yet occurred, a coverage model listing coverage events which define the goal of the test verification system and a task manager coupled to the data analyzer and the Bayesian network which refers to the coverage model and queries the Bayesian network to produce input data to achieve desired coverage events.

Abstract: A discriminative feature selection method for selecting a set of features from a set of training data sequences is described. The training data sequences are generated by at least two data sources, and each data sequence consists of a sequence of data symbols taken from an alphabet. The method is performed by first building a suffix tree from the training data. The suffix tree contains only suffixes of the data sequences having an empirical probability of occurrence greater than a first predetermined threshold, from at least one of the sources. Next the suffix tree is pruned of all suffixes for which there exists in the suffix tree a shorter suffix having equivalent predictive capability, for all of the data sources.