A template built by a user may be converted by a Server Script Generation Engine (SSGE) into script code. In converting, the SSGE may load and parse a framework file containing static script syntax to locate insertion points, each associated with an iteration number, and may iteratively parse the template, utilizing the iteration number to resolve, in order, tags and sub-tags contained in the template. If a tag is set to respond to the iteration number, a function of the tag is invoked to process any related sub-tags and return a script associated therewith at the appropriate insertion point. The framework file (with the appropriate script code inserted) is compiled and stored in a compiled script object which can be run multiple times to perform all of the output functions expected by the user in lieu of the need to reconvert the template.

In accordance with an example embodiment, large scale category classification based on sequence semantic embedding and parallel learning is described. In one example, one or more closest matches are identified by comparison between (i) a publication semantic vector that corresponds to at least part of the publication, the publication semantic vector based on a first machine-learned model that projects the at least part of the publication into a semantic vector space, and (ii) a plurality of category vectors corresponding to respective categories from a plurality of categories.

In accordance with an example embodiment, large scale category classification based on sequence semantic embedding and parallel learning is described. In one example, one or more closest matches are identified by comparison between (i) a publication semantic vector that corresponds to at least part of the publication, the publication semantic vector based on a first machine-learned model that projects the at least part of the publication into a semantic vector space, and (ii) a plurality of category vectors corresponding to respective categories from a plurality of categories.

A machine accesses a preexisting set of natural language text documents in multiple natural languages. Each natural language text document in at least a portion of the preexisting set is associated with an event. The machine trains, using the preexisting set of natural language text documents and the associated events, an event encoder to learn associations between texts and event annotations. The event encoder leverages a parser in each of the two or more natural languages. The machine generates, using the event encoder, new event annotations for texts. The machine trains, using the preexisting set of natural language text documents and the new event annotations for the texts generated by the event encoder, an event extraction engine to extract events from natural language texts in the two or more natural languages. The event extraction engine leverages the parser in each of the two or more natural languages.

An entity transaction attribute determination method for determining an attribute state or an attribute category of a to-be-predicted entity on a preset transaction is provided. The method comprises obtaining a plurality of historical relational networks sequentially arranged under a temporal order; determining, for each of the historical relational networks and through vector fusion of neighbor nodes, a plurality of description vectors of the to-be-predicted entity; processing, through a pre-trained time-series neural network, the description vectors to obtain an output result; and determining, according to the output result, the attribute state or the attribute category of the preset transaction attribute for the to-be-predicted entity. The method improves the accuracy of predicting a preset transaction attribute of an entity through the analysis of the description vectors.

An aspect provides a method, including: receiving, at an input component of an information handling device, user input comprising one or more words; identifying, using a processor of the information handling device, an emotion associated with the one or more words; creating, using the processor, an emotion tag including the emotion associated with the one or more words; storing the emotion tag in a memory; analyzing one or more emotion tags; and modifying an operation of an application based on the analyzing. Other embodiments are described and claimed.

A system for inputting data to a web page that is selectively accessed and displayed through a web browser executing on a computing device includes an electronic scratch pad on the computing device that is configured to identify data entry fields on an accessed web page and to retrieve a label corresponding to any identified data entry field of the web page. The electronic scratch pad records the label and any user input that is input using a user input device to the corresponding data entry field.

The technology disclosed provides a so-called “joint many-task neural network model” to solve a variety of increasingly complex natural language processing (NLP) tasks using growing depth of layers in a single end-to-end model. The model is successively trained by considering linguistic hierarchies, directly connecting word representations to all model layers, explicitly using predictions in lower tasks, and applying a so-called “successive regularization” technique to prevent catastrophic forgetting. Three examples of lower level model layers are part-of-speech (POS) tagging layer, chunking layer, and dependency parsing layer. Two examples of higher level model layers are semantic relatedness layer and textual entailment layer. The model achieves the state-of-the-art results on chunking, dependency parsing, semantic relatedness and textual entailment.

The technology disclosed provides a so-called “joint many-task neural network model” to solve a variety of increasingly complex natural language processing (NLP) tasks using growing depth of layers in a single end-to-end model. The model is successively trained by considering linguistic hierarchies, directly connecting word representations to all model layers, explicitly using predictions in lower tasks, and applying a so-called “successive regularization” technique to prevent catastrophic forgetting. Three examples of lower level model layers are part-of-speech (POS) tagging layer, chunking layer, and dependency parsing layer. Two examples of higher level model layers are semantic relatedness layer and textual entailment layer. The model achieves the state-of-the-art results on chunking, dependency parsing, semantic relatedness and textual entailment.

A system and method may allow for improved accuracy for address matching. The system may receive an address input and preprocess the address input. The address input may be standardized to create a standardized address input. The standardized address input may be compared to a stored address. The system may calculate a first address matching score based on the comparison. The system may reinvestigate the standardization of the address input and calculate a second address matching score based on a second comparison. The system may compare the first address matching score to the second address matching score to improve accuracy in address matching.