In an approach to AI explaining for natural language processing, responsive to receiving an input text for a machine learning model, an output is generated from the machine learning model. A plurality of alteration techniques are applied to the input text to generate one or more alternate outputs, where each alternate output corresponds to an alteration technique. A variation rate of the alternate output is calculated for each alteration technique. A preferred technique of generating neighboring data of the input text is generated based on a comparison of the variation rate of the alternate output for each alteration technique.

A platform receives transaction information for an entity, where the transaction information identifies a plurality of transactions associated with the entity, and receives entity information associated with the entity. The platform identifies, using a first model, a selected set of transactions based on the transaction information and the entity information, where the first model outputs information identifying the selected set of transactions based on the selected set of transactions being associated with an event, a theme, or a transaction parameter. The platform determines, using a second model, potential modifications to a document based on the selected set of transactions, where the second model receives the information identifying the selected set of transactions or information identifying the event, the theme, or the transaction parameter, and where the second model outputs information identifying the potential modifications. The platform provides the information identifying the potential modifications.

A platform receives transaction information for an entity, where the transaction information identifies a plurality of transactions associated with the entity, and receives entity information associated with the entity. The platform identifies, using a first model, a selected set of transactions based on the transaction information and the entity information, where the first model outputs information identifying the selected set of transactions based on the selected set of transactions being associated with an event, a theme, or a transaction parameter. The platform determines, using a second model, potential modifications to a document based on the selected set of transactions, where the second model receives the information identifying the selected set of transactions or information identifying the event, the theme, or the transaction parameter, and where the second model outputs information identifying the potential modifications. The platform provides the information identifying the potential modifications.

An AI platform may receive a request for information on text. The text is processed through a text mining pipeline for dynamic attribute extraction. An engine determines entities in the text and utilizes the entities to determine a relationship pattern. The engine identifies a trigger by matching one of the entities with a predefined entity in a utility authority file, locates an entity in close proximity to the trigger, identifies a value or regular expression in close proximity to the trigger in the text, and creates a triplet containing the entity, the trigger, and the value or regular expression, the triplet representing the relationship pattern. The engine applies an action to the triplet, wherein the action comprises obtaining the value from the text or translating the regular expression. The engine attaches the value or a result from the translating to the entity as a dynamic attribute of the entity.

An AI platform may receive a request for information on text. The text is processed through a text mining pipeline for dynamic attribute extraction. An engine determines entities in the text and utilizes the entities to determine a relationship pattern. The engine identifies a trigger by matching one of the entities with a predefined entity in a utility authority file, locates an entity in close proximity to the trigger, identifies a value or regular expression in close proximity to the trigger in the text, and creates a triplet containing the entity, the trigger, and the value or regular expression, the triplet representing the relationship pattern. The engine applies an action to the triplet, wherein the action comprises obtaining the value from the text or translating the regular expression. The engine attaches the value or a result from the translating to the entity as a dynamic attribute of the entity.

Automated conversion of incompatible data files into compatible benefit packages for a pharmacy benefit management (PBM) platform. In an embodiment, client files, comprising representations of benefit plans, are received. Each representation comprises a plurality of field values. For each representation, a layout and a set of translation rules are received, the plurality of field values in the representation are mapped to a plurality of attribute values based on the layout, and a benefit package, comprising a plurality of components, is generated by applying the set of translation rules to the plurality of attribute values, such that the benefit package represents the benefit plan.

Automated conversion of incompatible data files into compatible benefit packages for a pharmacy benefit management (PBM) platform. In an embodiment, client files, comprising representations of benefit plans, are received. Each representation comprises a plurality of field values. For each representation, a layout and a set of translation rules are received, the plurality of field values in the representation are mapped to a plurality of attribute values based on the layout, and a benefit package, comprising a plurality of components, is generated by applying the set of translation rules to the plurality of attribute values, such that the benefit package represents the benefit plan.

The embodiments herein achieve a picture based communication system. The system allows users option to select one or more pictures, and any associated attributes. The selection of one or more pictures, and any associated attributes is taken as input. The selected words and attributes are converted to a graph representation, and subsequently the graph representation is converted to a sentence in target language. The method further involves predicting new relations, words, and attributes for further selection by user.

A sign language translation system may capture infrared images of the formation of a sign language sign or sequence of signs. The captured infrared images may be used to produce skeletal joints data that includes a temporal sequence of 3D coordinates of skeletal joints of hands and forearms that produced the sign language sign(s). A hierarchical bidirectional recurrent neural network may be used to translate the skeletal joints data into a word or sentence of a spoken language. End-to-end sentence translation may be performed using a probabilistic connectionist temporal classification based approach that may not require pre-segmentation of the sequence of signs or post-processing of the translated sentence.

A sign language translation system may capture infrared images of the formation of a sign language sign or sequence of signs. The captured infrared images may be used to produce skeletal joints data that includes a temporal sequence of 3D coordinates of skeletal joints of hands and forearms that produced the sign language sign(s). A hierarchical bidirectional recurrent neural network may be used to translate the skeletal joints data into a word or sentence of a spoken language. End-to-end sentence translation may be performed using a probabilistic connectionist temporal classification based approach that may not require pre-segmentation of the sequence of signs or post-processing of the translated sentence.