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

• US 8,024,331 B2
• Filed: 06/02/2008
• Issued: 09/20/2011
• Est. Priority Date: 01/27/2000
• Status: Expired due to Fees

• Current Assignee: Manning & Napier Information Services LLC
• Original Assignee: Manning & Napier Information Services LLC
• Inventors: Calistri-Yeh, Randall J., Osborne, George B., Yuan, Bo, Snyder, David L.
• Primary Examiner(s): Vo; Tim T.
• Assistant Examiner(s): Fan; Shiow-Jy
• Application Number:
  US12/131,878
• Publication Number:
  US 20080281814A1
• Time in Patent Office: 1,205 Days
• Field of Search: 707/999.005, 707/E17.098, 707/728, 707/758, 704/9, 706/12
• US Class Current: 707/728
• 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 for a data processing system to efficiently identify at least one data set from a collection of datasets according to a query containing information indicative of desired datasets, the method comprising the machine-executed steps:

constructing a semantic vector for each dataset;

receiving the query containing information indicative of desired datasets;

constructing a semantic vector for the query;

comparing the semantic vector for the query to the semantic vector of each dataset;

selecting datasets whose semantic vectors are closest to the semantic vector for the query; and

generating a result including information of the selected datasets according to a result of the selecting step;

wherein;

the query or each of the datasets includes at least one data point; and

the semantic vector for the query or each of the datasets is constructed by the steps of;

for each data point, constructing a table for storing information indicative of a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

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

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

combining the semantic vector for each of the at least one data point to form the semantic vector of the query or each of the datasets.

2. The method of claim 1, wherein the datasets correspond to documents and the query is a natural language query.

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

performing a second search for datasets within the collection of datasets, wherein the second search using a method other than semantic vectors;

combining the two search results to obtain a combined weighted score for each dataset in either of the two search results;

selecting datasets whose combined weighted score is largest.

4. The method of claim 1, further comprising a step of clustering the selected datasets in real time.

5. A method for efficiently identifying data points in a semantic lexicon related to a dataset, the method comprising the machine-executed steps:

constructing a semantic vector for the dataset;

comparing the semantic vector for the dataset to semantic vector of each of the data points in the semantic lexicon;

selecting data points whose semantic vectors are closest to the semantic vector for the dataset; and

adding said selected data points to said dataset;

wherein;

the dataset includes at least one data point; and

the semantic vector for the dataset is constructed by the steps of;

for each data point, constructing a table for storing information indicative of a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

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

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

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

6. The method of claim 5, wherein the dataset is a document and the data points are words.

7. The method of claim 5, wherein the dataset is a natural language query in a search system and the data points are words.

8. A system for identifying at least one data set from a collection of datasets according to a query containing information indicative of desired datasets, the system comprising:

a computer configured to;

construct a semantic vector for each dataset;

receive the query containing information indicative of desired datasets;

construct a semantic vector for the query;

compare the semantic vector for the query to the semantic vector of each dataset;

select datasets whose semantic vectors are closest to the semantic vector for the query; and

generate a result including information of the selected datasets according to a result of the selecting step;

wherein;

the query or each of the datasets includes at least one data point; and

the semantic vector for the query or each of the datasets is constructed by the machine-executed steps of;

for each data point, constructing a table for storing information indicative of a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

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

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

combining the semantic vector for each of the at least one data point to form the semantic vector of the query or each of the datasets.

9. A non-transitory computer-readable medium carrying one or more sequences of instructions for efficiently identifying at least one data set from a collection of datasets according to an inquiry containing information indicative of desired datasets, 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 semantic vector for each dataset;

receiving the query containing information indicative of desired datasets;

constructing a semantic vector for the query;

comparing the semantic vector for the query to the semantic vector of each dataset;

selecting datasets whose semantic vectors are closest to the semantic vector for the query; and

generating a result including information of the selected datasets according to a result of the selecting step;

wherein;

the query or each of the datasets includes at least one data point; and

the semantic vector for the query or each of the datasets is constructed by the steps of;

for each data point, constructing a table for storing information indicative of a relationship between each data point and predetermined categories corresponding to dimensions in the semantic space;

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

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

combining the semantic vector for each of the at least one data point to form the semantic vector of the query or each of the datasets.