International Business Machines Corporation

Inbal Ronen

Arnon Yogev

Leonid Bolshinsky

Eitan Shapiro

David Boaz

Ofer Lavi

David Amid

Ella Rabinovich

Segev Shlomov

Ateret Anaby-Tavor

Lior Leiba

Roy Levin

Jonathan Bnayahu

Ido Guy

Yishai Abraham Feldman

Avi Yaeli

Asaf Adi

Sivan Yogev

Naama Tepper

Sara Porat

Ateret Anaby Tavor

Shiri Kremer-Davidson

Ksenya Kveler

Naama Zwerdling

Maya Barnea

Erel Uziel

Nurit Dor

Netta Aizenbud-Reshef

Alon Jacovi

Ori Bar El

Ateret Anaby - Tavor

An example system includes a processor to receive a bot design and escalation logs associated with a chat bot configured based on the bot design. The processor can compute a similarity score between each of a number of bot response nodes in the bot design and the escalation logs. The processor can generate a sorted list of the bot response nodes in the bot design based on the similarity scores.

Adjusting chat bot design based on matched escalation logs

A method, computer system, and a computer program product for automated agent intent detection enhancement are provided. A first message from a first user is received. The first message is generated during a first conversation between the first user and a first automated agent. A computer produces a second message that includes a same request as the first message but a different language modality than the first message. The second message and the first message are combined to form a combined message. The combined message is input into the first automated agent such that the first automated agent produces an intent classification for the first message.

Message mapping and combination for intent classification

Automatic measurement of semantic textual similarity of conversations, by: receiving two conversation texts, each comprising a sequence of utterances; encoding each of the sequences of utterances into a corresponding sequence of semantic representations; computing a minimal edit distance between the sequences of semantic representations; and, based on the computation of the minimal edit distance, performing at least one of: quantifying a semantic similarity between the two conversation texts, and outputting an alignment of the two sequences of utterances with each other.

Automatic measurement of semantic similarity of conversations

The computer receives a group of conversation data associated with the escalation node, identifies agent responses in the conversation data, and clusters them into agent response types. The computer identifies dialog state feature value sets for the conversations. The computer identifies feature value set associations with response types, and generates, Boolean expressions representing the feature value sets associated with each of the response types. The computer makes a recommendation to add to at least one child node for the escalation node, with the child node corresponding to one of the response types. The child node has, as an entry condition, the Boolean expression for the response type to which the child node corresponds. The child node has as an action, which according to some aspects, provides a response representative of the cluster of agent responses for the response type to which the child node corresponds.

Computerized dialog system improvements based on conversation data

A document-oriented search can be executed to generate a set of document results, at least one of the documents associated with at least one potential expert. The document results can be analyzed to produce a list of potential experts. An expertise score for at least one of the potential experts can be calculated based on a content score and a metadata score for the at least one of the potential experts. A confidence score for the potential expert can be calculated based on a diversity-constrained content score and a diversity-constrained metadata score for the at least one of the potential experts, the diversity-constrained content score and the diversity-constrained metadata score calculated using an evidence diversity score for the at least one of the potential experts. A list of experts with associated confidence scores that are above a confidence score threshold can be sent to a client device.

Calculating expertise confidence based on content and social proximity

A computerized method comprising using at least one hardware processor for receiving a plurality of digital responses in response to a digital broadcast, each digital response associated with at least one of a plurality of computerized devices. The method comprises an action of calculating a plurality of priority scores, one each for some of the plurality of digital responses, and an action of selecting an ordered subset of the plurality of digital responses based on the plurality of priority scores. The method comprises an action of presenting to a user the ordered subset, and an action of receiving at least one digital counter-response for at least one of the plurality of digital response. The method comprises an action of sending the at least one digital counter-response to a respective one of the plurality of computerized devices associated with the respective digital response.

Broadcast response prioritization and engagements

A method, computer system, and a computer program product for generating a chat summary personalized to a user is provided. The present invention may include receiving a plurality of input interactions associated with the user. The present invention may include determining a user profile based on the received plurality of input interactions, whereby the determined user profile includes a plurality of topics of interest. The present invention may include receiving a plurality of missed messages. The present invention may include determining a plurality of message clusters from the plurality of missed messages, whereby a topic is associated with each message cluster. The present invention may include ranking the determined plurality of message clusters based on comparing the topic associated with each message cluster to the plurality of topics of interest. The present invention may include presenting the ranked plurality of message clusters to the user.

Context-based personalized summarization of missed messages

A method includes executing, via a processor, a document-oriented search based on a query in an index of documents to generate a set of document results, each document associated with at least one potential expert. The method includes analyzing the document results to produce a list of potential experts. The method includes calculating an expertise score for each potential expert based on a calculated content score and metadata score for each potential expert. The method includes calculating an evidence diversity score for each potential expert. The method includes calculating a confidence score for each potential expert based on a diversity-constrained content score and a diversity-constrained metadata score for each potential expert. The method includes displaying a list of potential experts with associated confidence scores.

Software that uses data collected from queries performed on a graph database to dynamically improve graph partitioning. The software performs the following operations: (i) identifying a partitioned graph database including a set of edges and a set of vertices, wherein each vertex of the set of vertices is associated with one or more edges of the set of edges; (ii) determining an edge traversal value for one or more edges of the set of edges, wherein the edge traversal value for a respective edge relates to a number of times that the respective edge is traversed in response to one or more queries of the graph database; and (iii) calculating a first vertex score for a first vertex on a first graph partition, based, at least in part, on the edge traversal value(s) for one or more edges associated with the first vertex.

Dynamic partitioning of graph databases based on edge sampling

A method and system for highlighting email recipients according to a user's social network are provided. The method includes receiving an email message at a user's email client, the email message having multiple recipients, one of the recipients being the user. The method further includes obtaining a social network list for the user and comparing the social network list for the user with the email message recipients and highlighting recipients in the email message who are also in the social network list. The social network list for a user may be obtained by different methods including using the user's contact resources, or using an aggregating social network system including weighting relationships between contacts.

Method and system for highlighting email recipients

A method and system for providing relationships in search results are provided. The method includes indexing an entity in a search index as an entity index entry, the entity index entry including facets providing information on the entity type and related entities. Search results are obtained by a search engine in the form of ranked result entities, wherein the result entities include multiple types of entities. The method then includes retrieving index entries to determine relationships between search result entities and providing the relationships in search results. The method further includes, for each result entity, retrieving its entity index entry and cross-checking the facets of the retrieved entity index entry for other result entities.

Method and system for providing relationships in search results

A method for managing a software development process, the method includes: receiving information representative of instances of simple software development process events; detecting, by a complex event processing (CEP) engine and in response to the received information, at least one complex situation; and invoking at least one software management process reaction responsive to a detection of the at least one complex situation.

Device, method and computer program product for managing a software development process

A novel and useful mechanism for optimizing the detection of dependencies in a software system via feedback between static and dynamic engines. The present invention enables dynamic analysis to be more effective by using the results of static analysis to identify sections that need to be traced during dynamic analysis. The method also increases the accuracy of static analysis by using dynamic values as inputs to static analysis. Dynamic values used as inputs for static analysis are a result of dynamic analysis. The mechanism first executes a static analysis, then executes a dynamic analysis targeting the sections identified by the static analysis and finally executes a static analysis using the dynamic values identified during the dynamic analysis in order to detect dependencies. The present invention also provides a mechanism to analyze a specific section of a software system by first executing a dynamic analysis targeting the specific section and then executing a static analysis using the dynamic values identified during the dynamic analysis.

Growing community of inventors

Haifa, Israel

Inbal Ronen