Abstract: One embodiment provides a method, including: receiving information corresponding to a dialog flow between a conversational agent and a user; identifying, from the received information, dialog states, each dialog state corresponding to a possible response by the conversational agent; classifying a set of the identified dialog states as grounded dialog states, wherein the classifying comprises (i) receiving scoring input representing a desirability for each of the grounded dialog states and (ii) generating a score vector; assigning a score vector to each of the identified dialog states other than the grounded dialog states, wherein the assigning a score vector comprises extrapolating the score vector of a grounded dialog state to dialog states that the grounded dialog state is dependent upon; and computing an aggregate score vector for the dialog flow, wherein the aggregate vector score identifies an overall desirability of the dialog flow.

Abstract: Methods, systems, and computer program products for unsupervised learning of interpretable conversation models from conversation logs are provided herein. A computer-implemented method includes obtaining human-to-human conversation logs; training a deep learning model by (i) learning, in an unsupervised manner, semantic labels for dialog contexts in the multiple human-to-human conversation logs, (ii) mapping the learned semantic labels to query responses across the multiple human-to-human conversation logs, and (iii) inferring one or more entities from the multiple conversation logs based at least in part on the mapping; constructing a human-interpretable conversation model based at least in part on patterns determined via the trained deep learning model; and outputting the human-interpretable conversation model to at least one user.

Abstract: A computer generates labels for machine learning algorithms by retrieving, from a data storage circuit, multiple label sets that contain labels that each classify data points in a corpus of data. A graph is generated that includes a plurality of edges, each edge between two respective labels from different label sets of the multiple label sets. Weights are determined for the plurality of edges based upon a consistency between data points classified by two labels connected by the edges. An algorithm is applied that groups labels from the multiple label sets based upon the weights for the plurality of edges. Data points are identified from the corpus of data that represent conflicts within the grouped labels. An electronic message is transmitted in order to present the identified data points to entities for further classification. A new label set is generated using the further classification received from the entities.

Abstract: A method for simplifying user interactions with decision tree dialog managers is provided. The method may include receiving from a client computer, by a server computer, a user input. The server computer may identify one or more candidate nodes of a decision tree corresponding to the received user input. An entropy value may be calculated by the server computer for each of the identified candidate nodes. The server computer may then select a current node from among the candidate nodes, whereby the selected current node has a lowest calculated entropy value. A prompt associated with the selected current node may be transmitted to the user by the server computer.

Abstract: One embodiment provides a method, including: obtaining a plurality of chatlogs involving a conversational agent, wherein the plurality of chatlogs represent interactions between users and the conversational agent; identifying dialog states of the conversational agent, wherein each dialog state corresponds to a possible response by the conversational agent; generating a dialog tree including one or more dialog flows of the conversational agent, wherein a dialog flow represents an interaction that occurred between the conversational agent and a user; and providing a conversational reflow user interface for a conversational reflow system allowing execution of the conversational agent to mimic interactions between users and the conversational agent, wherein the conversational reflow system advances through dialog flows based upon input provided by a user at the conversational reflow user interface and the conversational reflow user interface displays output based upon the dialog states of the dialog tree.

Abstract: Methods, systems, and computer program products in connection with a system for bidding on multi-modal sub-spaces for e-commerce searches is provided herein. A computer-implemented method includes representing, in an embedding space affiliated with at least one electronic commerce website: (i) images related to one or more bids and (ii) items of text related to the one or more bids, the images and the items of text both coming from one or more bidding entities. Such a method also includes mapping a query to the embedding space based on a comparison of features of the query to (i) the images and (ii) the items of text. Further, such a method additionally includes selecting at least one of the one or more bidding entities based on the mapping, and outputting, to at least one user in response to the query, search results associated with the at least one selected bidding entity.

Abstract: One embodiment provides a method, including: receiving information corresponding to a dialog flow between a conversational agent and a user; identifying, from the received information, dialog states, each dialog state corresponding to a possible response by the conversational agent; classifying a set of the identified dialog states as grounded dialog states, wherein the classifying comprises (i) receiving scoring input representing a desirability for each of the grounded dialog states and (ii) generating a score vector; assigning a score vector to each of the identified dialog states other than the grounded dialog states, wherein the assigning a score vector comprises extrapolating the score vector of a grounded dialog state to dialog states that the grounded dialog state is dependent upon; and computing an aggregate score vector for the dialog flow, wherein the aggregate vector score identifies an overall desirability of the dialog flow.

Abstract: Provided herein is a system, method, and computer program product for modifying a language conversation model of the language learning system. Modifying the language conversation model includes receiving, using a conversational sub-system, voice inputs. The conversational sub-system converts the voice inputs to voice input data and processes the voice input data. The conversational sub-system detects an error in processing the voice input data and, based at least in part on the error, stores additional data comprising additional voice input data in a memory. The conversational sub-system applies machine learning to the additional data to derive a function that is not enabled within the language conversation model. The conversational sub-system develops an update that enables the language conversation model to implement the function. The update is applied to the language conversation model.

Abstract: Methods, systems, and computer program products for generating probabilistic annotations for entities and relations using reasoning and corpus-level evidence are provided herein. A method includes determining candidate entities and candidate relationships between entities from text contained within textual documents associated with a given domain; generating questions to support or refute legitimacy of a given one of the candidate entities and/or a given one of the candidate relationships; automatically determining an answer to each of the questions based on analysis of one or more additional items of knowledge pertaining to the given domain; computing a confidence score for each answer; and assigning a confidence value in support of the legitimacy of the given one of the candidate entities and/or the given one of the candidate relationships, wherein said confidence value is based on an aggregation of the confidence scores for the answers to the one or more questions.

Abstract: A method and system. Target clusterability is calculated as an average of a respective clusterability of at least one target data item comprised by a target domain. Target-side matchability is calculated as an average of a respective matchability of each target centroid of the target domain to source centroids of a source domain, wherein the source domain comprises at least one source data item. Source-side matchability is calculated as an average of a respective matchability of each source centroid of said source centroids to the target centroids. Source-target pair matchability is calculated as an average of the target-side matchability and the source-side matchability. Cross-domain clusterability between the target domain and the source domain is calculated as a linear combination of the calculated target clusterability and the calculated source-target pair matchability. The cross-domain clusterability is transferred to a device.

Abstract: A method and system. Target clusterability is calculated as an average of a respective clusterability of at least one target data item comprised by a target domain. Target-side matchability is calculated as an average of a respective matchability of each target centroid of the target domain to source centroids of a source domain, wherein the source domain comprises at least one source data item. Source-side matchability is calculated as an average of a respective matchability of each source centroid of said source centroids to the target centroids. Source-target pair matchability is calculated as an average of the target-side matchability and the source-side matchability. Cross-domain clusterability between the target domain and the source domain is calculated as a linear combination of the calculated target clusterability and the calculated source-target pair matchability. The cross-domain clusterability is transferred to a device.

Abstract: A computer receives a conversational input. The computer detects, based on the conversational input, relevant entities and relevant entity values, wherein the relevant entities and relevant entity values correspond to entities and entity values extracted from a catalog. The computer identifies, based on the relevant entities and relevant entity values, one or more matching products or services. The computer displays the one or more matching products or services. The computer assesses one or more attributes of the matching products or services for refinement. The computer provides one or more attribute refinement options based on the assessed one or more attributes. The computer receives user refinement in response to providing the one or more attribute refinement options. The computer receives a product or service selection.

Abstract: One embodiment provides a method, including: utilizing at least one processor to execute computer code that performs the steps of: obtaining at least one image of a user of a social medium from one or more posts on the social medium that are shared by the user; identifying a characteristic of the user by comparing characteristics of the at least one image of the user to other persons, wherein the other persons are clustered into characteristic groups based upon one or more images of each of the other persons; determining attributes of apparel included in the images of each of the other persons by parsing the one or more images and any text associated with the images of each of the other persons; and generating apparel style rules for a particular characteristic by associating the determined attributes of the apparel with the identified characteristic. Other aspects are described and claimed.

Abstract: A computer-implemented method for knowledge based ontology editing, is provided. The method receives a language instance to update a knowledge base, using a computer. The method semantically parses the language instance to detect an ontology for editing. The method maps one or more nodes for the ontology for editing based on an ontology database and the knowledge base. The method determines whether the mapped nodes are defined or undefined within the knowledge base. The method calculates a first confidence score based on a number of the defined and undefined mapped nodes. Furthermore, the method updates the knowledge base when the first confidence score meets a pre-defined threshold.

Abstract: A computer-implemented method includes receiving a question. The question is directed to a question answering system. The method includes extracting two or more queries from the question. The method includes generating two or more question sets, each of the two or more question sets addressing all of the two or more queries. The method includes selecting one of the two or more question sets to yield a selected question set, based on analyzing each of the two or more question sets according to a pre-trained model. The method includes presenting the selected question set to the question answering system.

Abstract: An approach is provided in which a system provides a question and a set of options to a user. The question corresponds to a first node in a decision tree and at least a portion of the options correspond to nodes that are directly connected to the first node. The system determines that the user's response corresponds to a second node that is different than one of the directly connected nodes and, in turn, displays a second question to the user corresponding to the second node.

Abstract: Methods, systems, and computer program products for resolving textual numerical queries using natural language processing techniques are provided herein.

Abstract: Provided are techniques for creating an inverted index for features of a set of data elements, wherein each of the data elements is represented by a vector of features, wherein the inverted index, when queried with a feature, outputs one or more data elements containing the feature. The features of the set of data elements are ranked. For each feature in the ranked list, the inverted index is queried for data elements having the feature and not having any previously selected feature and a cluster of the data elements is created based on results returned in response to the query.

Abstract: Two or more variables in a source code are determined. One or more variable groupings for the two or more variables are determined. A potential error for a variable of the two or more variables is determined based on the determined variable groupings.

Abstract: Generating a natural language dialog by finding missing semantic information in a user question by comparing it to the closest question available in a question corpus. Incrementally improved question precision is targeted during each round of the natural language dialog by generating follow-up questions that clarify semantic and syntactic characteristics of the user question. The follow-up questions are derived from analysis of the user question to identify areas of improvement on the user question.