Abstract: Techniques for predicting marketing outcomes using contrastive learning are disclosed, including: obtaining historical marketing messages; obtaining historical open rates associated respectively with the historical marketing messages; based on the historical marketing messages, generating latent space representations associated respectively with the historical marketing messages; based on the latent space representations and respective contents of the historical marketing messages, training a first machine learning model to map contents of marketing messages to corresponding latent space representations of the marketing messages; based at least on the latent space representations and the historical open rates, training a second machine learning model to map latent space representations of marketing messages to predicted open rates of the marketing messages.

Abstract: Techniques are disclosed for transforming feature vectors to a compact form that improves the efficiency and speed of computing analytical operations. A system may generate feature vectors that correspond to a set of entities and select a subset of the feature vectors to act as reference vectors. The system executes a first stage of data condensation by representing each feature vector in terms of a distance value to each of the reference vectors thereby reducing a number of attributes in each vector to the number of selected reference vectors. The system may further condense the data by characterizing the distance values for the remaining vectors relative to each particular reference vector in terms of statistical characteristics of the distance value distribution.

Abstract: Techniques are disclosed for determining individual contributions of corresponding individual entities that cooperatively contribute to accomplishing a goal or outcome. One technique determines an individual contribution of a target entity to an overall outcome value by identifying a first sequence of entities that includes the target entity and a corresponding sequence outcome value. Other sequences and their corresponding outcome values may be identified that partially match the first sequence while excluding the target entity. The outcome values for the other sequence(s) may be removed from the outcome value of the first sequence, thereby isolating the individual contribution of the target entity.

Abstract: A marketing campaign management system that includes a campaign platform that categorizes intended recipient profiles corresponding to intended recipients of marketing campaign messages based on their levels of fatigue. Recipient profiles categorized as “saturated” are then sub-segmented to determine their relative levels of fatigue or saturation. The intended recipient profiles that are most fatigued are removed from a recipients list such that they will not be sent any campaigns until after a rest state has been completed. A transitional safe state ensures that the intended recipient profiles inserted back into the list are not immediately removed once again.

Abstract: A marketing campaign management system that includes a campaign platform that categorizes intended recipient profiles corresponding to intended recipients of marketing campaign messages based on their levels of fatigue. Recipient profiles categorized as “saturated” are then sub-segmented to determine their relative levels of fatigue or saturation. The intended recipient profiles that are most fatigued are removed from a recipients list such that they will not be sent any campaigns until after a rest state has been completed. A transitional safe state ensures that the intended recipient profiles inserted back into the list are not immediately removed once again.

Abstract: Data units within a de-duplication device are stored in single instances within a database of data units and then referenced as many times as necessary within a referencing dictionary, potentially achieving substantial compaction of the referencing dictionary and thus the storage requirement of the de-duplication device as a whole.

Abstract: A write-data hash value corresponding to a block of write data is generated within a deduplicating data storage system. A block of lookup table entries is retrieved from a location in a lookup table stored within a block-based storage medium, the lookup table location being indicated by a first portion of the write-data hash value and each lookup table entry including a pointer to a respective stored data volume, a portion of a hash value that corresponds to the stored data volume, and a reference count indicating a quantity of references to the stored data volume. A second portion of the write-data hash value is compared to the portions of the hash values within the block of lookup table entries, and the reference count is incremented within one of the lookup table entries for which the portion of the hash value is determined to match the second portion of the write-data hash value.

Abstract: Redundant data is removed from a volume of data by partitioning the volume of data into fixed-length input segments and determining, for each of the input segments, whether a selected portion of the input segment matches a portion of a segment within a de-duplication dictionary. If the portion of the input segment matches a portion of the segment within the dictionary, the segment within the de-duplication dictionary is compared with the input segment and a token representative of the segment within the dictionary is substituted for at least part of the input segment determined to match the segment within the dictionary.

Abstract: Redundant data is removed from a volume of data by partitioning the volume of data into fixed-length input segments and, for each of the input segments, traversing nodes of a search tree in accordance with the value of a fixed-size portion of the input segment to determine if the search tree contains a pointer to a matching fixed-sized portion of a segment within a dictionary. If the search tree contains the pointer, the input segment is compared with the segment within the dictionary pointed to by the pointer, and a token representative of the segment within the dictionary is substituted for at least part of the input segment determined to match the segment within the dictionary.

Abstract: Redundant data is removed from a volume of data by partitioning the volume of data into fixed-length input segments and, for each of the input segments, traversing nodes of a search tree in accordance with the value of a fixed-size portion of the input segment to determine if the search tree contains a pointer to a matching fixed-sized portion of a segment within a dictionary. If the search tree contains the pointer, the input segment is compared with the segment within the dictionary pointed to by the pointer, and a token representative of the segment within the dictionary is substituted for at least part of the input segment determined to match the segment within the dictionary.

Abstract: An apparatus comprises a computer accessible medium encoded with a plurality of instructions which, when executed, compile a stylesheet into one or more data structures, and a hardware circuit coupled to receive the data structures and a document. The hardware circuit is configured to perform at least a portion of transforming the document as specified in the stylesheet using the data structures.

Abstract: An apparatus comprises a stylesheet compiler and a document processor. The stylesheet compiler is configured to identify expressions in a stylesheet and is configured to generate one or more expression trees representing the expressions. Expressions having one or more common nodes are represented as children of a subtree that represents the common nodes. Coupled to receive a document and the expression tree, the document processor is configured to evaluate the expressions represented in the one or more expression trees against the document.

Abstract: In various embodiments, an appliance may be used to implement various functions. In some embodiments, an appliance may be used with a DBMS server. The appliance may generate a request, and the DBMS server may generate a result. The appliance may transform the result to a different format for return to a requestor corresponding to the request. In still other embodiments, the appliance may be used with a computer system configured to execute other backend applications such as content management systems, filesystems, etc. In other embodiments, the appliance may receive first output data from a first server of a plurality of servers that is to be supplied as first input data to a second server of the plurality of servers. The appliance may be configured to transform the first output data to a first markup language format and to transform the first markup language format to the second format.

Abstract: Numerous embodiments of structured data transformation are disclosed. Some embodiments include characterizing at least a portion of an expression in a stylesheet as evaluatable at a phase wherein the phase is one of a compile phase, a parse phase, or a transformation phase of applying the stylesheet to a document to transform the document in a document transformation appliance, applying at least a portion of the expression to the document in the phase determined in the characterizing, combining results from applying the at least a portion of the expression and from applying other expressions to the document to generate a transformed document, and outputting the transformed document or saving the transformed document to memory.