Construction of trainable semantic vectors and clustering, classification, and searching using trainable semantic vectors

• US 7,444,356 B2
• Filed: 04/14/2004
• Issued: 10/28/2008
• Est. Priority Date: 01/27/2000
• Status: Expired due to Term

• Current Assignee: Manning & Napier Information Services LLC
• Original Assignee: Manning & Napier Information Services LLC
• Inventors: Yuan, Bo, Snyder, David L., Calistri-Yeh, Randall J., Osborne, George B.
• Primary Examiner(s): PANNALA, SATHYANARAYA R
• Application Number:
  US10/823,617
• Publication Number:
  US 20040199546A1
• Time in Patent Office: 1,658 Days
• Field of Search: 707/102, 707/104.1, 707/3, 707/5
• US Class Current: 1/1
• CPC Class Codes: G06F 16/36 Creation of semantic tools, e.g. ontology or thesauri
• CPC Class Codes: Y10S 707/99931 Database or file accessing
• CPC Class Codes: Y10S 707/99932 Access augmentation or optimizing
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching
• CPC Class Codes: Y10S 707/99934 Query formulation, input preparation, or translation
• CPC Class Codes: Y10S 707/99935 Query augmenting and refining, e.g. inexact access
• CPC Class Codes: Y10S 707/99936 Pattern matching access
• CPC Class Codes: Y10S 707/99939 Privileged access
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing
• CPC Class Codes: Y10S 707/99948 Application of database or data structure, e.g. distributed, multimedia, or image

Abstract

An apparatus and method are disclosed for producing a semantic representation of information in a semantic space. The information is first represented in a table that stores values which indicate a relationship with predetermined categories. The categories correspond to dimensions in the semantic space. The significance of the information with respect to the predetermined categories is then determined. A trainable semantic vector (TSV) is constructed to provide a semantic representation of the information. The TSV has dimensions equal to the number of predetermined categories and represents the significance of the information relative to each of the predetermined categories. Various types of manipulation and analysis, such as searching, classification, and clustering, can subsequently be performed on a semantic level.

Claims

1. A method of classifying new datasets within a predetermined number of categories based on assignment of a plurality of sample datasets to each category, the method comprising the computer-executed steps:

constructing a trainable semantic vector for each sample dataset relative to the predetermined categories in a multi-dimensional semantic space;

constructing a trainable semantic vector for each category based on the trainable semantic vectors for the sample datasets;

receiving a new dataset;

constructing a trainable semantic vector for the new dataset;

determining a distance between the trainable semantic vector for the new dataset and the trainable semantic vector of each category; and

classifying the new dataset within the category whose trainable semantic vector has the shortest distance to the trainable semantic vector of the new dataset;

wherein;

the new data set or each of the sample data sets includes at least one data point;

each data point corresponds to at least one of a word, a phrase, a sentence, a color, a typography, a punctuation, a picture, and a character string; and

the trainable semantic vector for each sample data set or the new dataset is constructed by performing the steps of;

for each data point, identifying a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

determining the significance of each data point with respect to the predetermined categories;

constructing a trainable semantic vector for each data point, wherein each trainable semantic vector has dimensions equal to the number of predetermined categories and represents the relative strength of its corresponding data point with respect to each of the predetermined categories; and

combining the trainable semantic vector for each of the at least one data point to form the semantic vector of the sample dataset or the new dataset.

2. The method of claim 1, wherein the datasets correspond to documents.

3. The method of claim 1, wherein the datasets correspond to email messages and the categories correspond to frequently asked questions with substantially static responses.

4. The method of claim 1, further comprising the steps:

detecting when a prescribed number of new datasets has been classified; and

updating the trainable semantic vectors for each of the categories.

5. The method of claim 4, wherein the step of updating comprises the step of re-constructing trainable semantic vectors for each category based on the trainable semantic vectors for the sample datasets and the trainable semantic vectors for the new datasets added to each category.

6. A method of classifying new datasets within a predetermined number of categories based on assignment of a plurality of sample datasets to each category, the method comprising the computer-executed steps:

constructing a trainable semantic vector for each sample dataset relative to the predetermined categories in a multi-dimensional semantic space;

receiving a new dataset;

constructing a trainable semantic vector for the new dataset;

identifying a select number of sample datasets whose trainable semantic vectors are closest in distance to the trainable semantic vector for the new dataset; and

classifying the new dataset in the category containing the greatest number of the select sample datasets;

wherein;

the new data set or each of the sample data sets includes at least one data point;

each data point corresponds to at least one of a word, a phrase, a sentence, a color, a typography, a punctuation, a picture, and a character string; and

the trainable semantic vector for each sample data set or the new dataset is constructed by performing the steps of;

for each data point, identifying a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

determining the significance of each data point with respect to the predetermined categories;

constructing a trainable semantic vector for each data point, wherein each trainable semantic vector has dimensions equal to the number of predetermined categories and represents the relative strength of its corresponding data point with respect to each of the predetermined categories; and

combining the trainable semantic vector for each of the at least one data point to form the semantic vector of the sample dataset or the new dataset.

7. The method of claim 6, wherein the datasets correspond to documents.

8. The method. of claim 6 wherein the datasets correspond to email messages and the categories correspond to frequently asked questions with substantially static responses.

9. The method of claim 6, further comprising the steps:

detecting when a prescribed number of new datasets has been classified; and

adding the new datasets to the set of sample datasets.

10. A method of classifying new datasets within a predetermined number of categories, the method comprising the computer-executed steps:

receiving a new dataset;

constructing a trainable semantic vector for the new dataset, where the dimensions of the trainable semantic vector correspond to the predetermined number of categories;

classifying the dataset in the category whose corresponding dimension in the trainable semantic vector has the largest value;

wherein;

the new data set includes one or more data point;

each data point corresponds to at least one of a word, a phrase, a sentence, a color, a typography, a punctuation, a picture, and a character string; and

the trainable semantic vector for the new dataset is constructed by performing the steps of;

for each data point within the new dataset, identifying a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

determining the significance of each data point with respect to the predetermined categories;

constructing a trainable semantic vector for each data point, wherein each trainable semantic vector has dimensions equal to the number of predetermined categories and represents the relative strength of its corresponding data point with respect to each of the predetermined categories; and

combining the trainable semantic vector for each data point to form the semantic vector of the new dataset.

11. The method of claim 10, wherein the datasets correspond to documents.

12. The method of claim 10, wherein the datasets correspond to email messages and the categories correspond to frequently asked questions with substantially static responses.

13. A system for classifying new datasets within a predetermined number of categories based on assignment of a plurality of sample datasets to each category, the system comprising:

a computer including a data processor and a data storage device carrying computer-readable instructions which, upon execution by the data processor, control the computer to;

construct trainable semantic vector for each sample dataset relative to the predetermined categories in a multi-dimensional semantic space;

construct a trainable semantic vector for each category based on the trainable semantic vectors for the sample datasets;

receive a new dataset;

construct a trainable semantic vector for the new dataset;

determine a distance between the trainable semantic vector for the new dataset and the trainable semantic vector of each category; and

classify the new dataset within the category whose trainable semantic vector has the shortest distance to the trainable semantic vector of the new dataset;

wherein;

the new data set or each of the sample data sets includes at least one data point;

each data point corresponds to at least one of a word, a phrase, a sentence, a color, a typography, a punctuation, a picture, and a character string; and

the trainable semantic vector for each sample data set or the new dataset is constructed by performing the steps of;

for each data point, identifying a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

determining the significance of each data point with respect to the predetermined categories;

constructing a trainable semantic vector for each data point, wherein each trainable semantic vector has dimensions equal to the number of predetermined categories and represents the relative strength of its corresponding data point with respect to each of the predetermined categories; and

combining the trainable semantic vector for each of the at least one data point to form the semantic vector of the sample dataset or the new dataset.

14. A system for classifying new datasets within a predetermined number of categories based on assignment of a plurality of sample datasets to each category, the system comprising:

a computer including a data processor and a data storage device carrying computer-readable instructions which, upon execution by the data processor, control the computer to;

construct a trainable semantic vector for each sample dataset relative to the predetermined categories in a multi-dimensional semantic space;

receive a new dataset;

construct a trainable semantic vector for the new dataset;

identify a select number of sample datasets whose trainable semantic vectors are closest in distance to the trainable semantic vector for the new dataset; and

classify the new dataset in the category containing the greatest number of the select sample datasets;

wherein;

the new data set or each of the sample data sets includes at least one data point;

each data point corresponds to at least one of a word, a phase, a sentence, a color a typography, a punctuation, a picture, and a character string; and

the trainable semantic vector for each sample data set or the new dataset is constructed by performing the steps of;

for each data point, identifying a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

determining the significance of each data point with respect to the predetermined categories; and

constructing a trainable semantic vector for each data point, wherein each trainable semantic vector has dimensions equal to the number of predetermined categories and represents the relative strength of its corresponding data point with respect to each of the predetermined categories; and

combining the trainable semantic vector for each of the at least one data point to form the semantic vector of the sample dataset or the new dataset.

15. A computer-readable storage medium carrying one or more sequences of instructions for classifying new datasets within a predetermined number of categories based on assignment of a plurality of sample datasets to each category, wherein execution of the one or more sequences of instructions by one or more processors causes the one or more processors to perform the computer-executed steps of:

constructing a trainable semantic vector for each sample dataset relative to the predetermined categories in a multi-dimensional semantic space;

constructing a trainable semantic vector for each category based on the trainable semantic vectors for the sample datasets;

receiving a new dataset;

constructing a trainable semantic vector for the new dataset;

determining a distance between the trainable semantic vector for the new dataset and the trainable semantic vector of each category; and

classifying the new dataset within the category whose trainable semantic vector has the shortest distance to the trainable semantic vector of the new dataset;

wherein;

the new data set or each of the sample data sets includes at least one data point;

each data point corresponds to at least one of a word, a phrase, a sentence, a color a typography, a punctuation, a picture, and a character string; and

the trainable semantic vector for each sample data set or the new dataset is constructed by performing the steps of;

for each data point, identifying a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

determining the significance of each data point with respect to the predetermined categories;

constructing a trainable semantic vector for each data point, wherein each trainable semantic vector has dimensions equal to the number of predetermined categories and represents the relative strength of its corresponding data point with respect to each of the predetermined categories; and

combining the trainable semantic vector for each of the at least one data point to form the semantic vector of the sample dataset or the new dataset.

16. A computer-readable storage medium carrying one or more sequences of instructions for classifying new datasets within a predetermined number of categories based on assignment of a plurality of sample datasets to each category, wherein execution of the one or more sequences of instructions by one or more processors causes the one or more processors to perform the steps of:

constructing a trainable semantic vector for each sample dataset relative to the predetermined categories in a multi-dimensional semantic space;

receiving a new dataset;

constructing a trainable semantic vector for the new dataset;

identifying a select number of select datasets whose trainable semantic vectors are closest in distance to the trainable semantic vector for the new dataset; and

classifying the new dataset in the category containing the greatest number of the select dataset;

wherein;

the new data set or each of the sample data sets includes at least one data point;

each data point corresponds to at least one of a word, a phrase, a sentence, a color, a typography, a punctuation, a picture, and a character string; and

the trainable semantic vector for each sample data set or the new dataset is constructed by performing the steps of;

for each data point, identifying a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

determining the significance of each data point with respect to the predetermined categories;

constructing a trainable semantic vector for each data point, wherein each trainable semantic vector has dimensions equal to the number of predetermined categories and represents the relative strength of its corresponding data point with respect to each of the predetermined categories; and

combining the trainable semantic vector for each of the at least one data point to form the semantic vector of the sample dataset or the new dataset.