Adobe, Inc.

Michael J Welch

Walter Chang

Larry Melvin Masinter

This specification describes technologies relating to multi core processing for parallel speech-to-text processing. In some implementations, a computer-implemented method is provided that includes the actions of receiving an audio file; analyzing the audio file to identify portions of the audio file as corresponding to one or more audio types; generating a time-ordered classification of the identified portions, the time-ordered classification indicating the one or more audio types and position within the audio file of each portion; generating a queue using the time-ordered classification, the queue including a plurality of jobs where each job includes one or more identifiers of a portion of the audio file classified as belonging to the one or more speech types; distributing the jobs in the queue to a plurality of processors; performing speech-to-text processing on each portion to generate a corresponding text file; and merging the corresponding text files to generate a transcription file.

Multi-core processing for parallel speech-to-text processing

Methods and apparatus for capturing, analyzing, and converting documents are provided. Provided in some embodiments is a method that includes obtaining a script document comprising script elements, assessing a layout of the script document, wherein the layout comprises a physical location of the script elements within the script document, determining an identification of one or more of the script elements based at least in part on the physical location of the one or more script elements within the script document, and assessing the one of the script elements using a finite state machine, wherein the finite state machine assesses one or more of the script elements preceding or following the one of the script elements to determine an accuracy of the identification of the one or more of the script elements determined.

Method and apparatus for capturing, analyzing, and converting scripts

A term analyzer receives an ordered collection of text-based terms. The term analyzer analyzes groupings of consecutive text-based terms in the ordered collection to identify occurrences of different combinations of text-based terms. In addition, the term analyzer maintains frequency information representing the occurrences of the different combinations of text-based terms in the collection. The frequency information can then be used to determine relatively significant keywords and/or keyword phrases in the document. In an example configuration, the term analyzer creates a tree in which a first term in a given grouping of the groupings is defined as a parent node in the tree and a second term in the given grouping is defined as a child node of the parent node in the tree. The method of the analyzer generalizes to create a tree of multi-word terms in which the terms can be efficiently ranked by occurrence.

Document analysis and multi-word term detector

This specification describes technologies relating to generating custom language models for audio content. In some implementations, a computer-implemented method is provided that includes the actions of receiving a collection of source texts; identifying a type from a collection of types for each source text, each source text being associated with a particular type; generating, for each identified type, a type-specific language model using the source texts associated with the respective type; and storing the language models.

Custom language models for audio content

Speech-to-text processing based on a time-ordered classification of audio file segments

A lexical mapper enables efficient text mining via use of lexical identifiers such as numerical identifier values. For example, the lexical mapper receives a collection of text-based strings associated with a document. Based on values of the strings, the lexical mapper maps the text-based strings to a set of lexical identifiers. In lieu of using the character strings, text mining applications process the numerical identifier values to analyze the document.

Generation and processing of numerical identifiers

A term analyzer receives an ordered collection of text-based terms. The ordered collection can contain terms from a document that have been filtered to remove 'noise' such as stopwords. The term analyzer analyzes groupings of consecutive text-based terms in the ordered collection to identify occurrences of different combinations of text-based terms in the ordered collection. In addition, the term analyzer maintains frequency information representing the occurrences of the different combinations of text-based terms in the collection. The frequency information can then be used to determine relatively significant keywords and/or keyword phrases in the document. In an example configuration, the term analyzer creates a tree in which a first term in a given grouping of the groupings is defined as a parent node in the tree and a second term in the given grouping is defined as a child node of the parent node in the tree. The method of the analyzer generalizes to create a tree of multi-word terms in which the terms can be efficiently ranked by occurrence.

A document analyzer receives a collection of text-based terms associated with a document. The document analyzer performs a statistical analysis on the text-based terms to identify a distribution of where the text-based terms appear in the document and relative frequency indicating how often the text-based terms appear in the document. The document analyzer utilizes the distribution and relative frequency information derived from the statistical analysis to rank multiple themes associated with the document. For example, a received listing of multiple themes may not be presented in any useful order, although it can be assumed that the themes in the listing are present in the document. Based on application of distribution and relative frequency information derived from the analysis, the document analyzer can identify which themes are most relevant to the document as a whole and/or which of themes correspond to different portions (e.g., pages or sections) of the document.

Document analyzer and metadata generation

A method and a system is provided to process data. For example, the method and system may be used to store (e.g., archive) documents. In an embodiment, the method comprises receiving a quantum of data and creating n data pieces of size s from the quantum of data. The method may comprise generating k random numbers of size s, wherein k defines a minimum number of processed data pieces required to reconstruct the quantum of data, and performing polynomial arithmetic modulo prime on the n data pieces. The polynomial arithmetic may utilize polynomial of order k and the prime may be selected based on a bit processing capability of a processor used to process the data. The prime may be 2+1 when the bit processing capability of a processor is 16 bits and 2+1 when the bit processing capability of a processor is 32 bits.

Method and system to perform secret sharing

Document analyzer and metadata generation and use

A method and a system is provided of to process data. For example, the method and system may be used to archive documents. In an embodiment, the method includes receiving original data for archiving and converting the original data into a plurality of shares using secret sharing. The plurality of shares may be distributed (e.g., to a plurality of remote repositories that each store a share. The plurality of shares may be distributed via a wide area network to a plurality of Internet Service Providers, for example, via an email message which is stored thereby to archive the share. A header of the email message may be used as a time stamp for each share. The repositories may be independent and geographically dispersed and a monitor may be provided to monitor a share stored at each repositories.

Growing community of inventors

Los Angeles, CA, United States of America

Michael J Welch