Method and device for classifying internet objects and objects stored on computer-readable media

• US 20020059221A1
• Filed: 10/17/2001
• Published: 05/16/2002
• Est. Priority Date: 10/19/2000
• Status: Active Grant

• Current Assignee: Anthony David Whitehead, Patrick Ryan Morin
• Original Assignee: Anthony David Whitehead, Patrick Ryan Morin
• Inventors: Morin, Patrick Ryan, Whitehead, Anthony David
• Granted Patent: US 7,383,282 B2
• Time in Patent Office: Days
• US Class Current: 707/5
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06V 10/56 relating to colour
• CPC Class Codes: G06V 20/10 Terrestrial scenes scenes under surveillance with static cameras G06V20/52; scenes perceived from the exterior of a vehicle G06V20/56; scenes perceived from the interior of a vehicle G06V20/59
• CPC Class Codes: Y10S 707/955 Object-oriented
• CPC Class Codes: Y10S 707/959 Network
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing

Abstract

The present invention discloses a method and device for classifying Internet objects, and/or objects stored on computer-readable media. The method is an automated method that uses a device to gather information from as many sources as possible within an object, and to analyze and combine this information in order to classify the object.

An automated method and device for identifying Internet objects containing adult content by analyzing descriptor data, metadata, text data, image data, video data, audio data, plugin information and relations between objects, is also disclosed. An automated method and device for identifying objects stored on computer-readable media containing adult content by analyzing text data, image data, video data, audio data is also disclosed.

Claims

1. A method of classifying an Internet object comprising the steps of:

computing (i) a descriptor coefficient, (ii) a name coefficient, (iii) a text coefficient, (iv) an image coefficient, (v) an audio coefficient, (vi) a video coefficient, (vii) a plug-in coefficient, and (viii) a relational coefficient.

2. A method of classifying an Internet object as recited in claim 1, including combining and weighting all of said coefficients except for said relational coefficient, and generating a first weighted sum.

3. A method as recited in claim 2, including providing a threshold value;

determining if said first weighted sum is greater than or less than said threshold value;

classifying said Internet object whereby, if said first weighted sum is greater than said threshold value, said object is classified as adult content, and if said first weighted sum is less than said threshold value, said object is not classified as adult content.

4. A method as recited in claim 3, including computing said relational coefficient if said object is not classified as adult content in said classifying step;

and

Combining and weighting all of said coefficients; and

generating a second weighted sum.

5. A method as recited in claim 4, including providing a threshold value;

determining if said second weighted sum is greater than or less than said threshold value;

classifying said Internet object whereby, if said second weighted sum is greater than said threshold value, said object is classified as adult content, and if said second weighted sum is less than said threshold value, said object is not classified as adult content.

6. A method of classifying an Internet object as recited in claim 3, including blocking or filtering access to Internet objects that exceed the tolerated threshold for adult content by storing the URLS of sites which contain said Internet object in a database, if said first weighted sum is greater than said threshold value whereby said sites can be blocked or filtered by referencing said database when the Internet is accessed.

7. A method of classifying an Internet object as recited in claim 4, including blocking or filtering access to Internet objects that exceed the tolerated threshold for adult content by storing the URLS of sites which contain said Internet object in a database, if said second weighted sum is greater than said threshold value whereby said sites can be blocked or filtered by referencing said database when the Internet is accessed.

8. A method of classifying an Internet object as recited in claim 1 wherein said step of computing the descriptor coefficient comprises the steps of:

(a) Identifying the PICS code in a metadata of an Internet object;

(b) Parsing labels of said PICS code; and

(c) Calculating a descriptor coefficient by computing the weighted average of the ratings of said labels.

9. A method of classifying an Internet object as recited in claim 1 wherein said step of computing the name coefficient comprises the steps of:

(a) Compiling a keyword list of words that are likely to indicate adult content;

(b) Generating a weighted keyword list by assigning each word or phrase in the keyword list a weight, such that a higher weight is given to words or phrases that are more likely to indicate adult content;

(c) Providing an automated robot with the keyword list;

(d) Computing substrings in the name of the Internet object and comparing each substring to the weighted keyword list; and

(e) Adding weights of all weighted keywords found in all substrings to compute the name coefficient.

10. A method of classifying an Internet object as recited in claim 1 wherein said step of computing the text coefficient comprises the steps of:

(a) Compiling a keyword and phrases list of words and phrases that are likely to indicate adult content;

(b) Assigning each word or phrase in the keyword and phrases list a weight, such that a higher weight is given to words that are more likely to indicate adult content;

(c) Providing an automated robot with the keyword and phrases list;

(d) Eliminating any formatting from the text of the Internet object to obtain an unformatted text;

(e) Reducing each individual word in the unformatted text to a stem, using a stemming algorithm;

(f) Comparing each stemmed word to the weighted keyword and phrases list;

(g) Calculating the text coefficient by first adding weights of all weighted keywords and phrases found in all stemmed words, and then multiplying that total by a scaling or weighting constant, which is a mathematical function of the total number of words in the Internet object; and

(h) Dividing the resulting number by the number of words in an Internet object.

11. A method of classifying an Internet object as recited in claim 1 wherein said step of computing the image coefficient comprises the steps of:

(a) Establishing a range of Hue-Saturation-Value (“

HSV”

) colour space values for skin tones;

(b) Converting the colour of the Internet object into the HSV colour space;

(c) Comparing every pixel of the Internet object image to identify which pixels fall within the colour space values;

(d) Dividing the number of skin tone pixels in the Internet object image by the total number of pixels in the image;

(e) Disregarding images that have less than a threshold proportion of skin pixels;

(f) Assigning each image in an Internet object that meets or exceeds the threshold proportion of skin pixels a nudity coefficient equal to the percentage of skin pixels in the image;

(g) Weighting the nudity coefficient of each image in an Internet object using specified weighting;

(h) Calculating the image coefficient by first adding together all of the weighted nudity coefficients generated by an Internet object; and

(i) Multiplying the total of all weighted nudity coefficients by a scaling or weighting constant, which is a mathematical function of the total number of images in the Internet object.

12. A method of classifying an Internet object as recited in claim 1 wherein said step of computing the audio coefficient comprises the steps of:

(a) Converting any audio components in the Internet object into unformatted text using a speech recognition algorithm;

(b) Concatenating the text data; and

(c) Evaluating the resultant text data using the method set out in claim 6 above (i.e. text coefficient method), and assigning a text coefficient using the method set out in claim 6 above.

13. A method of classifying an Internet object as recited in claim 1 wherein said step of computing the video coefficient comprises the steps of:

(a) Dividing the video data into image and audio data as required by the media type;

(b) Evaluating the resultant image data using the method set out in claim 11 (i.e. image coefficient method), and assigning an image coefficient for each frame of video data using the method set out in claim 11;

(c) Evaluating the resultant audio data using the method set out in claim 12 above (i.e. audio coefficient method), and assigning an audio coefficient for the audio data using the method set out in claim 12 above; and

(d) Calculating the video coefficient by adding a weighted sum of the image and audio coefficients, with weighting being determined automatically or can be set by the user.

14. A method of classifying an Internet object as recited in claim 1 wherein said step of computing the relational coefficient comprises the steps of:

(a) Determining whether the CoAC of an Internet object, using the calculations in the following individual coefficients (i) the descriptor coefficient, (ii) the name coefficient, (iii) the text coefficient, (iv) the image coefficient, (v) the audio coefficient, (vi) the video coefficient, and (vii) the plug-in coefficient, results in a CoAC greater than a tolerated threshold;

(b) If not, engaging in calculation of the relational coefficient by compiling a list of all Internet objects that the Internet object under consideration links to (the “

linked-to objects”

), as well as a list of all Internet objects that the Internet object under consideration is part of (the “

part of objects”

);

(c) Checking the database referred to in claim 6 to determine which of the “

linked-to”

objects and “

part of”

objects contain adult content that exceeds the tolerated threshold;

(d) Calculating an average CoAC of both the “

linked to”

objects and the “

part of”

objects that exceed the tolerated thresholds for adult content; and

(e) Calculating the relational coefficient by calculating a weighted average of the two averages determined in part (d) above, in which weighting can be determined automatically or can be set by the user.

16. A method of classifying an object stored on computer-readable media as recited in claim 15, including combining and weighting all of said coefficients, and generating a weighted sum.

17. A method of classifying an object stored on computer-readable media as recited in claim 15, including providing a threshold value;

Determining if said weighted sum is greater than or less than said threshold value;

Classifying said object stored on computer-readable media whereby, if said weighted sum is greater than said threshold value, said object is classified as adult content, and if said weighted sum is less than said threshold value, said object is not classified as adult content.

18. A method of classifying an object stored on computer-readable media as recited in claim 15, including blocking or filtering access to objects stored on computer-readable media that exceed the tolerated threshold for adult content by storing identifying information about such objects in a database, if said first weighted sum is greater than said threshold value, whereby said objects can be blocked or filtered by referencing said database when such objects are accessed.

19. A method of classifying an object stored on computer-readable media as recited in claim 15, wherein said step of computing the text coefficient comprises the steps of:

(a) Compiling a keyword and phrases list of words and phrases that are likely to indicate adult content;

(b) Assigning each word or phrase in the keyword and phrases list a weight, such that a higher weight is given to words that are more likely to indicate adult content;

(c) Providing an automated robot with the keyword and phrases list;

(d) Eliminating, where necessary, any formatting from the text of the object to obtain an unformatted text;

(e) Reducing each individual word in the unformatted text to a stem, using a stemming algorithm;

(f) Comparing each stemmed word to the weighted keyword and phrases list;

(g) Calculating the text coefficient by first adding weights of all weighted keywords and phrases found in all stemmed words, and then multiplying that total by a scaling or weighting constant, which is a mathematical function of the total number of words in the object stored on computer-readable media; and

(h) Dividing the resulting number by the number of words in an object stored on computer-readable media.

20. A method of classifying an object stored on computer-readable media as recited in claim 15, wherein said step of computing the image coefficient comprises the steps of:

(a) Establishing a range of Hue-Saturation-Value (“

HSV”

) colour space values for skin tones;

(b) Converting the colour of the object into the HSV colour space;

(c) Comparing every pixel of the object image to identify which pixels fall within the colour space values;

(d) Dividing the number of skin tone pixels in the object image by the total number of pixels in the image;

(e) Disregarding images that have less than a threshold proportion of skin pixels;

(f) Assigning each image in an Internet object that meets or exceeds the threshold proportion of skin pixels a nudity coefficient equal to the percentage of skin pixels in the image;

(g) Weighting the nudity coefficient of each image in an Internet object, using specified weighting;

(h) Calculating the image coefficient by first adding together all of the weighted nudity coefficients generated by an object stored on computer-readable media; and

(i) Multiplying the total of all weighted nudity coefficients by a scaling or weighting constant, which is a mathematical function of the total number of images in the Internet object.

21. A method of classifying an object stored on computer-readable media as recited in claim 15, wherein said step of computing the audio coefficient comprises the steps of:

(a) Converting any audio components in the object stored on computer-readable media into unformatted text using a speech recognition algorithm;

(b) Concatenating the text; and

(c) Evaluating the resultant text data using the method set out in claim 6 above (i.e. text coefficient method), and assigning a text coefficient using the method set out in claim 6 above.

22. A method of classifying an object stored on computer-readable media as recited in claim 15, wherein said step of computing the video coefficient comprises the steps of:

(a) Dividing the video data into image and audio data as required by the media type;

(b) Evaluating the resultant image data using the method set out in claim 23 above (i.e. image coefficient method), and assigning an image coefficient for each frame of video data using the method set out in claim 23 above;

(c) Evaluating the resultant audio data using the method set out in claim 24 above (i.e. audio coefficient method), and assigning an audio coefficient for the audio data using the method set out in claim 24 above; and

(d) Calculating the video coefficient by adding a weighted sum of the image and audio coefficients, with weighting being determined automatically or can be set by the user.

15. A method of classifying an object stored on computer-readable media comprising the steps of computing (i) a text coefficient, (ii) an image coefficient, (iii) an audio coefficient, (iv) a video coefficient.