Back to results
US 7,548,863 B2Granted
Adaptive context sensitive analysis
AssigneeApple Inc.
InventorsYasuo Kida, Michi Iwasaki, Keisuke Hara, Takumi Takano
Life of this patent
Estimating the term…
Abstract
A method, apparatus, system, and signal-bearing medium that converts input data, such as speech or phonetic characters, to text by finding documents that are similar to the input data and using the similar documents to create a customized dictionary including a weighted list of words found in the similar documents. The words that are weighted higher may have a higher probability of being used in the input data, so words in the converted text are chosen based on the weighted list. A vector space model may be used to search for the similar documents.
Claims47
A method comprising:
- receiving input data, the input data being data for conversion from a phonetic representation or speech to written text characters of one or more words of a language, the input data having a plurality of data pieces;
- calculating, by a processor of a computer via an iterative process, a vector for the input data in a textual format of the language, the iterative process including starting with a first data piece of the input data converted to a selected textual format to form a current vector of the input data and iteratively updating the current vector with a next data piece of the input data using elements of the current vector until all the data pieces are converted into the textual format;
- selecting a subset of a plurality of documents based on a closeness of the vector for the input data to a plurality of vectors for text of the plurality of documents, the text of the plurality of documents being written in the language;
- determining a frequency of the text in the subset of the plurality of documents;
- and
- converting the input data to a representation of one or more written text characters of the language based on the frequency of the text in the subset of the plurality of documents.
3 more elements
An apparatus comprising:
- means for receiving input data, the input data being data for conversion from a phonetic representation or speech to written text characters of one or more words of a language, the input data having a plurality of data pieces;
- means for calculating, via an iterative process, a vector for the input data in a textual format of the language, the iterative process including starting with a first data piece of the input data converted to a selected textual format to form a current vector of the input data and iteratively updating the current vector with a next data piece of the input data using elements of the current vector until all the data pieces are converted into the textual format;
- means for dividing a plurality of documents into a first subset similar to the input data and a second subset dissimilar to the input data based on the vector for the input data, the plurality of documents having text being written in the language;
- means for determining a frequency text in the first subset;
- and
- means for converting the input data to a representation of one or more written text characters of the language based on the frequency of the text in the first subset.
3 more elements
A machine-readable storage medium having executable instructions stored thereon, which when executed, causes a machine to perform operations comprising:
- receiving input data, the input data being data for conversion from a phonetic representation or speech;
- to written text characters of one or more words of a language, the input data having a plurality of data pieces;
- calculating by a processor of the machine, via an iterative process, an input vector for the input data in a textual format of the language, the iterative process including starting with a first data piece of the input data converted to a selected textual format to form a current vector of the input data and iteratively updating the current vector with a next data piece of the input data using elements of the current vector until all the data pieces are converted into the textual format;
- finding a subset of a plurality of documents based on the input vector, the plurality of documents having text being written in the language;
- determining a frequency of the text in the subset;
- and
- converting the input data to a representation of one or more written text characters of the language based on the frequency.
4 more elements
A method comprising:
- finding, by a processor of a computer;
- at least one document having text that is similar to input data, the input, data being data for conversion from a phonetic representation or speech;
- to written text characters of one or more words of a language and having a plurality of data pieces, the finding based on the text of the at least one document and the input data converted to a textual format from an iterative process, the iterative process including starting with a first data piece of the input data converted to a selected textual format to form a current format of the input data and iteratively updating the current format with a next data piece of the input data using elements of the current format until all the data pieces are converted into the textual format, the text of the at least one document being written in the language;
- creating by the processor a customized dictionary comprising a weighted list of the text that occur in the at least one document;
- and
- converting the input data to a representation of one or more written text characters of the language based on the weighted list.
3 more elements
An apparatus comprising:
- means for finding at least one document having text that is similar to input data, the input data being data for conversion from a phonetic representation or speech to written text characters of one or more words of a language and having a plurality of data pieces, the finding based on the text of the at least one document and the input data converted to a textual format from an iterative process, the iterative process including starting with a first data piece of the input data converted to a selected textual format to form a current format of the input data and iteratively updating the current format with a next data piece of the input data using elements of the current format until all the data pieces are converted into the textual format, the text of the at least one document being written in the language;
- means for creating a customized dictionary comprising a weighted list of the text that occur in the at least one document;
- and
- means for converting the input data to a representation of one or more written text characters of the language based on the weighted list.
1 more element
A machine-readable storage medium having executable instructions stored thereon, which when executed, causes a machine to perform operations comprising:
- finding by a processor of the machine, at least one document having text that is similar to input data, the input data being data for conversion from a phonetic representation or speech to written text characters of one or more words of a language and having a plurality of data pieces, the finding based on the text of the at least one document and the input data converted to a textual format from an iterative process, the iterative process including starting with a first data piece of the input data converted to a selected textual format to form a current format of the input data and iteratively updating the current format with a next data piece of the input data using elements of the current format until all the data pieces are converted into the textual format, the text of the at least one document being written in the language;
- creating a customized dictionary comprising a weighted list of the text that occur in the at least one document;
- and
- converting the input data to a representation of one or more written text characters of the language based on the weighted list.
1 more element
A system comprising:
- a processor;
- and
- a storage device coupled to the processor, wherein the storage device comprises instructions, which when executed on the processor cause the processor to perform operations that comprise:
- finding at least one document having text that is similar to input data, the input data being data for conversion from a phonetic representation or speech to written text characters of one or more words of a language and having a plurality of data pieces, the finding based on the text of the at least one document and the input data converted to a textual format from an iterative process, the iterative process including starting with a first data piece of the input data converted to a selected textual format to form a current format of the input data and iteratively updating the current format with a next data piece of the input data using elements of the current format until all the data pieces are converted into the textual format, the text of the at least one document being written in the language;
- creating a customized dictionary comprising a weighted list of the text that occur in the at least one document;
- and
- converting the input data to a representation of one or more written text characters of the language based on the weighted list.
4 more elements
A method of converting input data into text, the method comprising the acts of:
- determining a phonetic representation or speech possible text for representations of the input data, the input data being data for conversion, by a processor of a computer to written text characters of one or more words of a language and having a plurality of data pieces, the determining based on the input data converted to a textual format from an iterative process, the iterative process including starting with a first data piece of the input data converted to a selected textual format to form a current format of the input data and iteratively updating the current format with a next data piece of the input data using elements of the current format until all the data pieces are converted into the possible text;
- comparing the possible text against text of a set of one or more pre-existing documents, the text of the set of one or more pre-existing documents being written in the language;
- selecting at least one of the set of pre-existing documents based on similarity between the possible text and the text of the set of pre-existing documents;
- associating by the processor weights with each of the possible text based on the text of the at least one selected pre-existing document;
- and
- for those of the representations having plural possible text, selecting one of the plural possible text based on the associated weights:
- and
- providing the selected text as a representation of one or more written text characters of the language.
5 more elements
A method of converting input data into text, the method comprising the acts of:
- using a dictionary to determine, by a processor of a computer, text for representations of input data, the input data being data for conversion from a phonetic representation or speech to written text characters of one or more words of a language, a first of the representations being associated with multiple text of the language, the determined text based on the input data converted to a textual format from an iterative process the iterative process including starting with a first data piece of the input data converted to a selected textual format to form a current format of the input data and iteratively updating the current format with a next data piece of the input data using elements of the current format until all the data pieces are converted into the determined text;
- selecting a set of one or more pre-existing documents having text similar to the determined text for the representations of the input data, the text of the set of one or more pre-existing documents being written in the language;
- assigning by the processor, weights to the multiple text associated with the first representation based, at least in part, on occurrence of the multiple text in the selected set of one or more pre-existing documents;
- and
- converting, in accordance with the assigned weights, the first representation to a first of the multiple text.
2 more elements
Dependencies are inferred from the claim text.