Abstract: A computer-implemented method according to one embodiment includes identifying a plurality of different summaries for a single instance of content, calculating a relevancy score for each of the plurality of different summaries, and selecting one of the plurality of different summaries, based on the relevancy score for each of the plurality of different summaries.

Abstract: A computer-implemented method, system and computer program product for generating a client-specific document quality model, by: analyzing data using existing quality heuristics to identify new, unexpected or problem patterns in the data; forming the quality heuristics into one or more clusters for each container level of the data; exploring each of the clusters to identify sources of the patterns; and developing new quality heuristics based on the sources of the patterns, wherein the new quality heuristics are used to generate the client-specific document quality model.

Abstract: Systems and methods for generating and annotating cell documents include extracting tables from a document using a table extraction engine. Headers are extracted for each of the tables using a header detection engine. Cells are extracted from each of the tables using a cell extraction engine. A cell document is generated for each of the cells which are each correlated to corresponding portions of the headers, each cell document recording the correlation between the cells and the headers. Each cell document is annotated to generate annotated cell documents with a cell recognition model trained to perform natural language processing on the cell documents by classifying each term in each of the cell documents and extracting relationships between the terms of each of the cell documents.

Abstract: A method, system and computer-usable medium are disclosed for associating data cells with headers and tables having one or more embedded header structures. In certain embodiments, a table having rows and columns is received, wherein the table includes a plurality of cells, wherein each cell is populated with at least one of a header name, data value, or no information, the table having at least one embedded header. A determination is made as to whether a cell is a header cell or data cell. If the cell is a header cell, a count of consecutive column headers is maintained. A current list of column headers is dynamically updated based on the count of the consecutive column headers. Upon encountering a data cell, the current list of column headers is assigned to the data cell.

Abstract: Mechanisms are provided to implement an answer change notification system. The mechanisms receive a change operation to change a portion of a user profile data structure associated with a user and identify a first entry in a question and answer (QA) log data structure, corresponding to the user, having an indicator identifying a previous answer of the first entry as being dependent upon information in the user profile data structure. The mechanisms resubmit a question of the first entry to a question and answer (QA) system to generate a new answer to the question in response to identifying the entry. The mechanisms receive the new answer from the QA system and output, to a client device associated with the user, a notification identifying the new answer to the question in response to receiving the new answer.

Abstract: A method for training a question answering system includes providing training questions to a question answering system executing on a computer and to a plurality of subject matter experts. The question answering system generates first answers to each training question. Second answers to the training questions are received from the subject matter experts. Feature scores for each of the first answers and the second answers are generated and compared across the second answers and the first answers. Each of the feature scores is representative of a quality of an answer that is indicative of relevance to a corresponding training question. Based on the comparing, a measure of consistency of the feature scores of the second answers is determined, and a measure of consistency of the feature scores of the second answers to the first answers is determined. The measures of consistency are transmitted to the subject matter experts.

Abstract: Systems and methods for generating and annotating cell documents include extracting tables from a document using a table extraction engine. Headers are extracted for each of the tables using a header detection engine. Cells are extracted from each of the tables using a cell extraction engine. A cell document is generated for each of the cells which are each correlated to corresponding portions of the headers, each cell document recording the correlation between the cells and the the headers. Each cell document is annotated to generate annotated cell documents with a cell recognition model trained to perform natural language processing on the cell documents by classifying each term in each of the cell documents and extracting relationships between the terms of each of the cell documents.

Abstract: Embodiments relate to an intelligent computer platform to provide a contextual analogy response. The aspect of providing a contextual analogy response includes denoting an analogy phrase within a communication. An anaphora within the analogy phrase is detected and a set of sentences are parsed into grammatical components wherein the grammatical type for each parsed component is identified. A sentence is created with the detected anaphora and an action term from the analogy phrase. The set of sentences and the analogy phrase are matched with the identified components that are assigned to a solved analogy association. A related contextual response is attached to the solved analogy association to generate the outputted analogy response.

Abstract: A method, system and computer-usable medium are disclosed for minimizing reevaluation of a ground truth corpus in response to concept drift. Certain embodiments are directed to a computer implemented comprising: generating a knowledge graph using a ground truth corpus, where the knowledge graph includes concept nodes, context definition nodes, and document nodes, where each concept node has one or more edges to a context definition node and to a document node; updating a context definition node in the knowledge graph based on context drift; identifying edges between the updated context definition node and concept nodes affected by the updated context definition; and identifying edges between the affected concept nodes and corresponding document nodes to identify document nodes affected by the context drift; and reevaluating documents in the ground truth corpus corresponding to the affected document nodes pursuant to updating the ground truth corpus to compensate for the context drift.

Abstract: A retraining service accesses conversational logs, each of the conversational logs recording a separate conversation, between a separate user and a conversational service, and at least one outcome identified with the separate conversation. The retraining service assess, from the conversational logs, at least one conversation gap and response with the at least one outcome matching a type of outcome that indicates the response impacted user experience in a negative way from among types of outcomes. The retraining service evaluates one or more recommendations for retraining the response to promote a positive type of outcome from among the types of outcomes. The retraining service outputs the one or more recommendations to the conversational service for directing retraining of the response by the conversational service.

Abstract: Embodiments relate to an intelligent computer platform to support natural language (NL) processing. The request is analyzed and a lexical answer type (LAT) related to the received request is identified. A knowledge graph (KG) related to the LAT is identified and leveraged to extract a first concept related to the LAT and a second concept related to the first concept. First and second cluster are created, with the first cluster having the LAT and first concept as qualifiers, and the second cluster having the first and second concepts as qualifiers. Each of the formed clusters is populated with one or more documents. An inter-cluster assessment is conducted based on the relevancy of the populated document(s) to the received input. In addition, a machine learning model (MLM) corresponding to the KG is identified and utilized to selectively augment the MLM with the LAT, first and second concepts, and a corresponding relationship to the inter-cluster assessment.

Abstract: Embodiments relate to an intelligent computer platform to receive a request for processing against a corpus. The request is analyzed and a lexical answer type (LAT), a first concept relevant to the received request and a second concept related to the identified first concept, are each identified. The LAT, together with the first and second concepts are utilized to create a first and second cluster. Documents are selectively populated into the clusters based on the respective LAT and concept qualifiers. The clusters are subject to sorting based on relevancy to the received request.

Abstract: Provided is a validation framework for modelling possible failures that might occur when an orchestrated transaction calls external services to ensure that error handling and reporting is robust and well designed. The disclosed techniques ensure that no changes are necessary to either the code making a call or the services that might be called. The techniques are not limited to web servers and REST APIs as they may be used to test and validate any kind of system that employs well defined APIs. The claimed subject matter, or “validation framework” may be added to an existing API or created as a new module that acts as a proxy server in a non-micro service type of system. Although described with respect to a gateway-API service, the claimed subject matter is equally applicable to other systems that process orchestrated transactions.

Abstract: A method for creating content includes making a corpus of images available, adding a plurality of tag data to each of the images, receiving a query, extracting a trigger from the query, identifying a set of the images in the corpus, wherein identified images have tag data matching the extracted trigger, and creating a video comprising the identified images.

Abstract: A method, system and computer-usable medium are disclosed for adjusting fact-based answers provided by a question/answer (QA) system. A user submits a question to the QA system, where it is categorized into a question type. The QA system then processes the question to generate an answer. The QA system then generates an answer adjustment if it is determined that the question type and answer meet a predicted undesirable outcome. The answer adjustment may include a warning, a disclaimer, a recommendation, an alternative fact-based answer, a referral to an assistance service, or any combination thereof.

Abstract: A method, system and computer-usable medium are disclosed for computing file system management. A corpus of content is processed to extract metadata associated with folders and files referenced by a directory structure. Natural Language Processing (NLP) operations are then performed on the corpus to generate concept and entity data associated with each folder and file, followed by performing Natural Language (NL) classification operations to generate intent classification data, which in turn is processed to determine ranked, dominant intents for each folder and file. The corpus content, extracted metadata, concept and entity data, and ranked dominant intents are then processed to generate indexed content and term data. Application context data associated with an interaction is collected and processed to determine a user intent, which is then processed with the indexed content and term data to identify a corresponding folder and file, which in turn are provided to the user.

Abstract: A method identifies and removes bias from a machine learning model. A user/computer inputs a plurality of input training data into a machine learning system to generate an output of labeled output data. The user/computer evaluates the labeled output data according to a consistency metric to associate the labeled output data with a corresponding consistency assessment. The user/computer selects each labeled output data having a consistency assessment indicating a consistency assessment that is greater than a predetermined threshold to form a labeled output data subset, and then creates additional labeling for the labeled output data subset. The user/computer utilizes the additional labeling to distinguish each labeled training data from labeled output data subset as being mislabeled and biased, and then adjusts the learning machine based on the labeled output data subset being mislabeled and biased.

Abstract: A method improves a functionality of a conversational agent that is generated by an artificial intelligence (AI) system. A conversational agent receives a first utterance from a first entity. The AI system identifies an unverified response to the first utterance; sends the unverified response to the first entity; and receives a positive feedback indication about the unverified response from the first entity. The AI system searches a data store in order to identify an entry for a second utterance by a second entity, where the second entity has sent a positive feedback for the unverified response. The AI system sends the second utterance and the unverified response to the first entity, and receives a positive feedback for the unverified response to the second utterance from the first entity in order to mark the unverified response as a verified response, which responds to future receipts of the first utterance.

Abstract: A classifier receives a document and analyzes the document to determine one or more predicted roles of one or more signatories, each predicted role determined based on one or more signature elements in the content of the document executed by the one or more signatories. The classifier evaluates each of the one or more predicted roles in view of a plurality of expected signatory role characteristics of a plurality of categories of documents of a transaction to select a particular category associated with the document from among the plurality of categories. The classifier classifies the document within the transaction as a particular logical type identified by the particular category from among a plurality of logical types for the transaction.

Abstract: Provided is a validation framework for modelling possible failures that might occur when an orchestrated transaction calls external services to ensure that error handling and reporting is robust and well designed. The disclosed techniques ensure that no changes are necessary to either the code making a call or the services that might be called. The techniques are not limited to web servers and REST APIs as they may be used to test and validate any kind of system that employs well defined APIs. The claimed subject matter, or “validation framework” may be added to an existing API or created as a new module that acts as a proxy server in a non-micro service type of system. Although described with respect to a gateway-API service, the claimed subject matter is equally applicable to other systems that process orchestrated transactions.