Abstract: Generating textual entailment pair by a natural language processing (NLP) system. The NLP system receives two input texts, such as a question and a candidate answer. The NLP system queries a database and retrieves passages likely to include text that support the candidate answer. The NLP system generates parse trees and performs term matching on the passages and scores them according to the matching. The NLP system detects anchor pairs in the question and in the passage and aligns subgraphs (within the parse trees) of one to the other based on matching. The NLP system identifies aligned terms in the question and the passage that are not in the aligned subgraphs. The NLP system identifies text fragments, for the question and the passage, within the non-aligned segments of their respective parse trees, that connect the aligned term to the aligned portion of the subgraph.

Abstract: Generating textual entailment pair by a natural language processing (NLP) system. The NLP system receives two input texts, such as a question and a candidate answer. The NLP system queries a database and retrieves passages likely to include text that support the candidate answer. The NLP system generates parse trees and performs term matching on the passages and scores them according to the matching. The NLP system detects anchor pairs in the question and in the passage and aligns subgraphs (within the parse trees) of one to the other based on matching. The NLP system identifies aligned terms in the question and the passage that are not in the aligned subgraphs. The NLP system identifies text fragments, for the question and the passage, within the non-aligned segments of their respective parse trees, that connect the aligned term to the aligned portion of the subgraph.

Abstract: A method, system and computer program product for generating answers to questions. In one embodiment, the method comprises receiving a query, conducting a search through one or more data sources to identify candidate answers to the query, and providing each of the candidate answers with a preliminary score. The method further comprises filtering out any of the candidate answers with a preliminary score that does not satisfy a defined condition. The candidate answers having preliminary scores that satisfy this condition form a subset of the candidate answers. Each of the candidate answers in this subset is processed to produce further scores. A ranking function is applied to these further scores to determine a ranking for each of the candidate answers in the subset; and after this ranking function is applied, one or more of the candidate answers are selected as one or more final answers to the query.

Abstract: A method, system and computer program product for generating answers to questions. In one embodiment, the method comprises receiving an input query, identifying a plurality of candidate answers to the query; and for at least one of these candidate answers, identifying at least one proof of the answer. This proof includes a series of premises, and a multitude of documents are identified that include references to the premises. A set of these documents is selected that include references to all of the premises. This set of documents is used to generate one or more scores for the one of the candidate answers. A defined procedure is applied to the candidate answers to determine a ranking for the answers, and this includes using the one or more scores for the at least one of the candidate answers in the defined procedure to determine the ranking for this one candidate answer.

Abstract: A method, computer program product, and computer system, for receiving a first set of ground truth instances from a first source. A second set of ground truth instances may be received from a second source. The first and second sets of ground truth instances may be weighed differently based on a level of trust associated with each of the first and second sources. The weighted first and second sets of ground truth instances may be applied in a machine learning task executed by a computer.

Abstract: According to an aspect, concept identification in a question answering system includes receiving, at a computer processor, a text span as a subject of a search query. An aspect also includes searching a title-oriented document (TOD) corpus for the text span, and matching, by the computer processor, concepts in title-oriented documents TODs of the TOD corpus to the span of text. The matching of the concepts includes decomposing the title-oriented documents into a series of passages, scoring the passages identified as possible matches, and merging and ranking results of the scoring to produce final scores for each concept associated with the title-oriented documents.

Abstract: A method is provided that includes generating answer-passage pairs, each associated with a respective one of multiple feature vectors. The method includes, for each answer in the pairs, merging the feature vectors associated with any of the pairs that include the answer to obtain a merged answer vector. The method includes, for each passage in the pairs, (i) merging the feature vectors associated with any of the pairs that includes the passage to obtain a merged passage vector, (ii) merging the feature vectors from the merged answer vector for each answer that is associated with the passage via at least one of the pairs to obtain a merged passage-answer vector, and (iii) concatenating the merged passage vector and the merged passage-answer vector to obtain a concatenated passage vector. The method includes ranking the concatenated passage vector for each passage to obtain a ranked list of passages with associated confidence scores.

Abstract: A method, system and computer program product for generating answers to questions. In one embodiment, the method comprises receiving an input query, decomposing the input query into a plurality of different subqueries, and conducting a search in one or more data sources to identify at least one candidate answer to each of the subqueries. A ranking function is applied to each of the candidate answers to determine a ranking for each of these candidate answers; and for each of the subqueries, one of the candidate answers to the subquery is selected based on this ranking. A logical synthesis component is applied to synthesize a candidate answer for the input query from the selected the candidate answers to the subqueries. In one embodiment, the procedure applied by the logical synthesis component to synthesize the candidate answer for the input query is determined from the input query.

Abstract: A method, system and computer program product for generating answers to questions. In one embodiment, the method comprises receiving an input query; conducting a search to identify candidate answers to the input query, and producing a plurality of scores for each of the candidate answers. For each of the candidate answers, one, of a plurality of candidate ranking functions, is selected. This selected ranking function is applied to the each of the candidate answers to determine a ranking for the candidate answer based on the scores for that candidate answer. One or more of the candidate answers is selected, based on the rankings for the candidate answers, as one or more answers to the input query. In an embodiment, the ranking function selection is performed using information about the question. In an embodiment, the ranking function selection is performed using information about each answer.

Abstract: Embodiments may provide a computer implemented method in a data processing system comprising a processor and a memory, the memory comprising instructions that are executed by the processor to cause the processor to implement a system for bookmarking, the method comprising: receiving, by the processor, at least one search query from a user; returning, by the processor, a set of search results and a search score for each search result to the user; recording, by the processor, behavior of the user with respect to the search results; selecting, by the processor, at least one search result based on recorded behavior and the search score; determining, by the processor, at least one topic for the at least one selected search result; and generating, by the processor, a bookmark for each selected search result, under the at least one topic.

Abstract: A mechanism is provided in a data processing system for using paraphrase metrics for answering questions. The mechanism receives an input question and generating a candidate answer from a corpus of information. The candidate answer has a supporting passage from the corpus of information. The mechanism divides the input question into a first sequence of tokens and divides the supporting passage into a second sequence of tokens. The mechanism identifies a plurality of subsequences of tokens within the second sequence of tokens and applies a paraphrase metric to compare the first sequence of tokens to each of the plurality of subsequences of tokens to generate a plurality of paraphrase metric scores. The mechanism then determines a confidence score for the candidate answer based on a highest paraphrase metric score within the plurality of paraphrase metric scores.

Abstract: A method of generating secondary questions in a question-answer system. Missing information is identified from a corpus of data using a computerized device. The missing information comprises any information that improves confidence scores for candidate answers to a question. The computerized device automatically generates a plurality of hypotheses concerning the missing information. The computerized device automatically generates at least one secondary question based on each of the plurality of hypotheses. The hypotheses are ranked based on relative utility to determine an order in which the computerized device outputs the at least one secondary question to external sources to obtain responses.

Abstract: A method of generating secondary questions in a question-answer system. Missing information is identified from a corpus of data using a computerized device. The missing information comprises any information that improves confidence scores for candidate answers to a question. The computerized device automatically generates a plurality of hypotheses concerning the missing information. The computerized device automatically generates at least one secondary question based on each of the plurality of hypotheses. The hypotheses are ranked based on relative utility to determine an order in which the computerized device outputs the at least one secondary question to external sources to obtain responses.

Abstract: Exemplary methods and devices herein receive an inquiry and automatically analyze words used in the inquiry, potential answers, and data maintained by evidence sources using the computerized device to determine the sensitivity level associated with the inquiry. The sensitivity level associated with the inquiry represents an emotional and cognitive state of the user. Such methods and devices automatically generate at least one follow-up question based on the sensitivity level associated with the inquiry and receive a follow-up response into the computerized device in response to the follow-up question(s). The methods and devices also automatically produce scores for the potential answers using the computerized device based on the inquiry, the follow-up responses, and ratings of the evidence sources. Following this, these methods and devices automatically generate output answers to the inquiry based on the sensitivity level associated with the inquiry using the computerized device.

Abstract: A system, method and computer program product for conducting questions and answers with deferred type evaluation based on any corpus of data. The method includes processing a query including waiting until a “Type” (i.e. a descriptor) is determined AND a candidate answer is provided. Then, a search is conducted to look (search) for evidence that the candidate answer has the required Lexical Answer Type (e.g., as determined by a matching function that can leverage a parser, a semantic interpreter and/or a simple pattern matcher). Prior to or during candidate answer evaluation, a process is provided for extracting and storing collections of entity-type pairs from semi-structured text documents. During QA processing and candidate answer scoring, a process is implemented to match the query LAT against the lexical type of each provided candidate answer and generate a score judging a degree of match.

Abstract: Embodiments may provide a computer implemented method in a data processing system comprising a processor and a memory, the memory comprising instructions that are executed by the processor to cause the processor to implement a system for bookmarking, the method comprising: receiving, by the processor, at least one search query from a user; returning, by the processor, a set of search results and a search score for each search result to the user; recording, by the processor, behavior of the user with respect to the search results; selecting, by the processor, at least one search result based on recorded behavior and the search score; determining, by the processor, at least one topic for the at least one selected search result; and generating, by the processor, a bookmark for each selected search result, under the at least one topic.

Abstract: According to an aspect, learning parameters in a feed forward probabilistic graphical model includes creating an inference model via a computer processor. The creation of the inference model includes receiving a training set that includes multiple scenarios, each scenario comprised of one or more natural language statements, and each scenario corresponding to a plurality of candidate answers. The creation also includes constructing evidence graphs for each of the multiple scenarios based on the training set, and calculating weights for common features across the evidence graphs that will maximize a probability of the inference model locating correct answers from corresponding candidate answers across all of the multiple scenarios. In response to an inquiry from a user that includes a scenario, the inference model constructs an evidence graph and recursively constructs formulas to express a confidence of each node in the evidence graph in terms of its parents in the evidence graph.

Abstract: According to an aspect, learning parameters in a feed forward probabilistic graphical model includes creating an inference model via a computer processor. The creation of the inference model includes receiving a training set that includes multiple scenarios, each scenario comprised of one or more natural language statements, and each scenario corresponding to a plurality of candidate answers. The creation also includes constructing evidence graphs for each of the multiple scenarios based on the training set, and calculating weights for common features across the evidence graphs that will maximize a probability of the inference model locating correct answers from corresponding candidate answers across all of the multiple scenarios. In response to an inquiry from a user that includes a scenario, the inference model constructs an evidence graph and recursively constructs formulas to express a confidence of each node in the evidence graph in terms of its parents in the evidence graph.

Abstract: A method, system and computer program product for generating answers to questions. In one embodiment, the method comprises receiving an input query, identifying a plurality of candidate answers to the query; and for at least one of these candidate answers, identifying at least one proof of the answer. This proof includes a series of premises, and a multitude of documents are identified that include references to the premises. A set of these documents is selected that include references to all of the premises. This set of documents is used to generate one or more scores for the one of the candidate answers. A defined procedure is applied to the candidate answers to determine a ranking for the answers, and this includes using the one or more scores for the at least one of the candidate answers in the defined procedure to determine the ranking for this one candidate answer.

Abstract: A computer querying an application programming interface with each of multiple synthetic samples, each of the synthetic samples representing a separate sample assigned an original class from among multiple classes classified by a particular machine learning model and distorted to induce the particular machine learning model to misclassify the separate sample as a different class from among the classes. The computer accumulating, by the computer, a score of a number of results returned by the application programming interface that match an expected class label assignment of the different class for each of the synthetic samples. The computer, in response to the score exceeding a threshold, verifying, by the computer, that a service provided by the application programming interface is running the particular machine learning model.