Abstract: The technology disclosed relates to de-duplicating contextually similar news feed items. In particular, it relates to assembling a set of news feed items from a plurality of electronic sources and preprocessing the set to generate normalized news feed items that share common company-name mentions and token occurrences. The normalized news feed items are used to calculate one or more resemblance measures based on a sequence alignment score and/or a hyperlink score. The sequence alignment score determines contextual similarity between news feed item pairs, arranged as sequences, based on a number of matching elements in the news feed item sequences and a number of edit operations, such as insertion, deletion, and substitution, required to match the news feed item sequences. The hyperlink score determines contextual similarity between news feed item pairs by comparing the respective search results retrieved in response to supplying the news feed item pairs to a search engine.

Abstract: The technology disclosed presents a novel spatial attention model that uses current hidden state information of a decoder long short-term memory (LSTM) to guide attention and to extract spatial image features for use in image captioning. The technology disclosed also presents a novel adaptive attention model for image captioning that mixes visual information from a convolutional neural network (CNN) and linguistic information from an LSTM. At each timestep, the adaptive attention model automatically decides how heavily to rely on the image, as opposed to the linguistic model, to emit the next caption word. The technology disclosed further adds a new auxiliary sentinel gate to an LSTM architecture and produces a sentinel LSTM (Sn-LSTM). The sentinel gate produces a visual sentinel at each timestep, which is an additional representation, derived from the LSTM's memory, of long and short term visual and linguistic information.

Abstract: The technology disclosed presents a novel spatial attention model that uses current hidden state information of a decoder long short-term memory (LSTM) to guide attention and to extract spatial image features for use in image captioning. The technology disclosed also presents a novel adaptive attention model for image captioning that mixes visual information from a convolutional neural network (CNN) and linguistic information from an LSTM. At each timestep, the adaptive attention model automatically decides how heavily to rely on the image, as opposed to the linguistic model, to emit the next caption word. The technology disclosed further adds a new auxiliary sentinel gate to an LSTM architecture and produces a sentinel LSTM (Sn-LSTM). The sentinel gate produces a visual sentinel at each timestep, which is an additional representation, derived from the LSTM's memory, of long and short term visual and linguistic information.

Abstract: Systems and methods are provided for coordinating an emergency response at a facility. The system includes employee's user devices (UDs), an employee database that stores employee metadata, a module that stores facilities metadata, and a server system that hosts an application that interfaces with the UDs and monitors their locations within the facility. When a trigger event occurs (e.g., request for help or evacuation order), the application can determine the location of at least a first UD of an employee and identification information for the employee based on employee metadata. Based on facilities metadata, the application can generate a floorplan that includes an icon that represents the location of the first UD within the facility and the identification information. Some of the employees are members of an emergency response team (ERT). The floorplan can be displayed at the ERT member's UDs along with other information that helps ERT members coordinate the emergency response.

Abstract: The technology disclosed presents a novel spatial attention model that uses current hidden state information of a decoder long short-term memory (LSTM) to guide attention and to extract spatial image features for use in image captioning. The technology disclosed also presents a novel adaptive attention model for image captioning that mixes visual information from a convolutional neural network (CNN) and linguistic information from an LSTM. At each timestep, the adaptive attention model automatically decides how heavily to rely on the image, as opposed to the linguistic model, to emit the next caption word. The technology disclosed further adds a new auxiliary sentinel gate to an LSTM architecture and produces a sentinel LSTM (Sn-LSTM). The sentinel gate produces a visual sentinel at each timestep, which is an additional representation, derived from the LSTM's memory, of long and short term visual and linguistic information.

Abstract: Disclosed are methods, apparatus, systems, and computer-readable storage media for recommending an event to a user. In some implementations, one or more servers receive information identifying a plurality of events. The one or more servers store data of the plurality of events in a first one or more data tables having an action field, an intent field, and a user field, and analyze the data of the first one or more data tables to generate one or more pairs, each pair including information identifying a set of events and a target event. The one or more servers may calculate a similarity score for each of the one or more pairs and store the respective similarity score in a second one or more data tables having a set field, a target event field, and a similarity score field.

Abstract: A database system operates a feed system and communicates with different remote computing devices associated with different users. The database system may display feed items from the feed system on the remote computing devices and enable users to author new posts to the feed system. An authoring tool supports creation of new posts, and provides transparency as to which users will have access to the post. The authoring tool supports managing intended recipients, including users, groups, topics, files, and other entities. The authoring tool creates, and shows to the author, an automatic summary of recipients before the post is added to the feed system.

Abstract: Systems and methods are provided for semi-synchronously exporting data. A web server creates an export record in response to receiving a data request from a client computer. The web server sends a token to the client computer, wherein receipt of the token enables the client computer to be used for other purposes while waiting for receipt of requested data. At least one of multiple batch processing servers processes the data request in response to accessing the export record. At least one of the multiple batch processing servers updates the export record to reflect the processing of the data request. The web server exports the processed data request to the client computer in response to identifying an update of the export record.

Abstract: Systems and methods are provided for coordinating an emergency response at a facility. The system includes employee's user devices (UDs), an employee database that stores employee metadata, a module that stores facilities metadata, and a server system that hosts an application that interfaces with the UDs and monitors their locations within the facility. When a trigger event occurs (e.g., request for help or evacuation order), the application can determine the location of at least a first UD of an employee and identification information for the employee based on employee metadata. Based on facilities metadata, the application can generate a floorplan that includes an icon that represents the location of the first UD within the facility and the identification information. Some of the employees are members of an emergency response team (ERT). The floorplan can be displayed at the ERT member's UDs along with other information that helps ERT members coordinate the emergency response.

Abstract: A combined directed graph is created having a corresponding node for each node in a first directed graph lacking a corresponding node in a second directed graph, each node in the second graph lacking a corresponding node in the first graph, and each node in the first graph having a corresponding node in the second graph. A corresponding directed arc is created in the combined directed graph for each arc in the first graph lacking a corresponding arc in the second directed graph, each arc in the second graph lacking a corresponding arc in the first graph, and each arc in the first graph having a corresponding arc in the second graph. A recommendation is output for a user to interact with a recommended object based on an object interaction and a conditional probability, in the combined graph, which corresponds to the recommended object and the object interaction.

Abstract: Methods and systems are provided for managing files in an on-demand system. One exemplary method involves a server obtaining an input value from a data input field of an electronic file that was generated based on a template that maintains an association between the data input field and a field of an object in a database. The method continues with the server storing the input value in the field of the object in the database based on the association. In one or more embodiments, an input value indicated by a graphical user interface element within an electronic file is mapped to a column of an object table in the database that corresponds to the field of the object associated with the data input field that corresponds to the graphical user interface element.

Abstract: A database system that supports associating a custom metadata type with loosely coupled triggers may be implemented to include instructions to receive a definition of a custom metadata type, the definition comprising at least one relationship field associated with a triggerable database object. The database system may also include instructions to associate the definition of the custom metadata type with a loosely-coupled trigger, and to cause the loosely-coupled trigger to execute based on generating a custom metadata record of the custom metadata type. The custom metadata record including a triggerable database object value for the relationship field.

Abstract: A remote debug session for a server group is provided. A server group including multiple servers that perform workload sharing receives a request to debug code executed at the server group. The code is executed on behalf of a client of a database associated with the server group. At least one of the servers of the group initiates a debugging session and establishes a communication connection with the client. The server group maintains the connection open with the client for the duration of the debugging session. Subsequent requests related to the debug session can be handled in a number of ways by the server group, and all communication to the client about processing the requests is through the connection.

Abstract: A system, method and apparatus for executing scripts to provide offline prompts in a social networking application are provided. A social networking application operating on a mobile device in an offline mode receives an input comprising a sequence of characters from a user of the mobile device. The social networking application executes a script that was received by the mobile device while being operated in an online mode. The script analyzes each character of the user input as each character is received by the mobile device, and generates a prioritized list of social network contact identifiers, wherein the prioritized list is a prediction of social network contact identifiers associated with the content of the user input and data associated with the script. The social networking application further causes the prioritized list of social network contact identifiers to be provided to the user in response to the user input.

Abstract: In one embodiment, a computer-implemented method executable by a server system to store data in a data cache and refresh the data based on a dynamic schedule is provided. The method includes: receiving, by a processor, data from a first resource; storing, by the processor, the data in a data cache; determining, by the processor, a type of the data, and an access frequency of the data; determining, by the processor, a dynamic schedule based on the type of the data, and the access frequency of the data; and refreshing the data cache with new data from the first resource based on the dynamic schedule.

Abstract: Methods and systems are provided for integrating social media channels into a customer service management (CRM) portal. The method includes receiving data for a consumer post, parsing the data into a social post object and a social persona object, creating a case from the objects, and displaying the case in the CRM portal.

Abstract: Methods and systems are provided for displaying calendared events on a mobile computing device. The method includes: receiving, by the mobile computing device, a meeting invitation including meeting information and an email address; transmitting the email address to a customer relationship management (CRM) database; receiving, by the mobile computing device, contextual information associated with the email address from the CRM database; and displaying, by the mobile computing device, the meeting information and the contextual information.

Abstract: Disclosed are some examples of methods, apparatus, systems and computer program products related to a user interface providing a call center service. In some implementations, the user interface presents session tabs and application tabs. A session tab can have a parent-child relationship with an application tab. Customer data can be retrieved from a database, where the customer data includes account data and child data. The customer data can be provided to cause display in the user interface of the account data in association with a first session tab and the child data in association with a first application tab. Edits to the child data can cause the first application tab to be designated as unsaved.

Abstract: The technology disclosed relates to an end-to-end neural network for question answering, referred to herein as “dynamic coattention network (DCN)”. Roughly described, the DCN includes an encoder neural network and a coattentive encoder that capture the interactions between a question and a document in a so-called “coattention encoding”. The DCN also includes a decoder neural network and highway maxout networks that process the coattention encoding to estimate start and end positions of a phrase in the document that responds to the question.

Abstract: The technology disclosed provides a quasi-recurrent neural network (QRNN) that alternates convolutional layers, which apply in parallel across timesteps, and minimalist recurrent pooling layers that apply in parallel across feature dimensions.