Data processing: artificial intelligence – Neural network
Reexamination Certificate
2000-01-06
2003-10-14
Patel, Ramesh (Department: 2121)
Data processing: artificial intelligence
Neural network
C706S025000, C706S016000
Reexamination Certificate
active
06633855
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a system, method, and program for filtering content from an incoming data stream being accessed by a viewer program using a neural network to inhibit access to material deemed unacceptable.
2. Description of the Related Art
Although the Internet is heralded as a revolutionary technology capable of empowering people by providing relatively easy access to a tremendous store house of information, it is also derided for allowing unrestricted access to material considered highly inappropriate for children, such as pornography, hate literature, and violence. Many lawmakers and various interest groups have called for government censorship of the Internet because of the availability of such controversial material. One other concern with the Internet has been the ability of children to engage in inappropriate communications with adults on IRC chat lines that not only exposes children to inappropriate material, but also facilitate inappropriate encounters between children and adults. Businesses that network their employee computers to the Internet are also concerned that employees spend time on non-business related Internet activity. Thus, businesses are also interested in controlling which Internet sites their employees may visit to prevent unproductive employee Internet access.
The response of the computer industry to these concerns has been the development of filtering software, such as Net Nanny by Net Nanny LTD., which is described in the “Net Nanny User Guide” (Copyright Net Nanny LTD., 1997). Such prior art filtering software is capable of monitoring Internet software, such as browsers, e-mail and IRC chat room applications and also other application programs, such as word processing and image viewing applications, e.g., Adobe Photoshop, Corel Photo-Paint, and Microsoft Paint. Current filtering software operates by providing a list of Internet sites and words that could be deemed inappropriate to children. To monitor Internet sites, the filtering software developer provides user lists of inappropriate web sites and other Internet sites, such as newsgroups and IRC chat lines, that its in-house researchers have located and deemed inappropriate for a particular age group of children. The purchaser of the filter program, which may be a parent, library, school, etc, may then set the filter software to deny access to all the Internet sites on the provided list.
The current techniques for identifying undesirable web sites rely on the filter developer to provide thorough research on web sites and timely updates. However, there are an immeasurable number of Internet web sites, many of which cannot be located through traditional search engines. Thus, researchers may miss numerous Web sites containing inappropriate material that are not accessible through traditional search techniques. Further, web sites are being added and removed all the time. Thus, a child would have access to inappropriate sites that are added to the Internet between updates of the list.
The prior art method for filtering documents is to scan the document for words on a “hit list” of unacceptable words, and deny access to those documents containing a word on the hit list. This prior art technique for identifying inappropriate documents may screen or deny access to numerous acceptable documents. For instance, if the word “breast” is on the list, then the filtering software would deny access to documents describing cooking recipes for chicken breast or for medical documents on such important issues as breast cancer. In fact, many have criticized the use of word based lists as a filtering tool because it often screens out much useful educational information. For instance, if a list included common hate terms or a discussion of hate groups by a civil rights or anti-hate group, then such filtering software could conceivably deny access to deemed appropriate anti-hate literature.
Businesses are often interested in limiting not only what employees cannot access, like the child filtering product, but also limiting what they can access. For this reason, businesses may use firewall software that would prevent their employees from accessing any Internet sites except those specifically sanctioned or on an approved list. However, such techniques may be problematic because the employee could be barred from readily accessing a work-related site that is not on the approved list, such as a site linked to or mentioned from a site on the approved list.
Thus, there is a need in the art for an improved technique for filtering Internet and computer readable material for inappropriate content, while at the same time providing flexibility to allow access to educational material that may otherwise include words or phrases that are often associated with inappropriate material.
SUMMARY OF THE PREFERRED EMBODIMENTS
To overcome the limitations in the prior art described above, preferred embodiments disclose a method, system, and program for filtering a data object for content deemed unacceptable by a user. A data object requested by a viewer program is received. The data object is processed to determine predefined language statements. Information on the determined language statements is inputted into a neural network to produce an output value. A determination is then made as to whether the output value indicates that the data object is,unacceptable. Viewer program access to the data object is inhibited upon determining that the data object is unacceptable.
In further embodiments, a determination is made as to whether the output value indicates that the data object requires further consideration. If so, information on the data object is logged. User input on a rating indicating the acceptability of the data object is received and the neural network is trained to process the logged data object as input to produce the received rating as output.
The data object may comprise any incoming data stream, such as data from a document or one of multiple packets that form a document requested by the viewer.
In yet further embodiments, inhibiting the viewer program access to the data object upon determining that the data object is unacceptable comprises blocking access to the entire data object. Alternatively, a determination may be made of language statements in the data object that are unacceptable. In such case, inhibiting the viewer program access to the data object upon determining that the data object is unacceptable comprises blocking access to the unacceptable language statements and allowing access to language statements not determined to be unacceptable.
The preferred embodiments provide a filter program that is capable of considering more factors of a document than current filter software that would block access based on the presence of a single unacceptable word. The preferred filter program is capable of considering all unacceptable words as well as acceptable content words. The preferred embodiment filter program utilizes neural networks to train the filter to recognize when a document includes a level of inappropriate language that exceeds an acceptability threshold. Using neural networks allows the filter to consider documents that are not identical to the ones used to train the network; yet what was learned during training is applied to determine whether to inhibit access to a different document. The preferred filter program is an improvement over current art because it is less likely to deny access to acceptable documents that may otherwise include unacceptable words, whereas current filtering programs would reject a document upon the occurrence of a single specified unacceptable word.
Preferred embodiments that filter packets of data, typically transmitted over the Internet, are an improvement over current filter art utilizing lists of unacceptable Internet web sites because the preferred filter program will inhibit access to unacceptable Internet web sites that were not located by searchers creating the list of prohibited web sites or added to the Intern
Hirl Joseph P.
International Business Machines - Corporation
Konrad Raynes Victor & Mannu
Patel Ramesh
Victor David W.
LandOfFree
Method, system, and program for filtering content using... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method, system, and program for filtering content using..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method, system, and program for filtering content using... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3163202