Natural language processing for analyzing internet content and finding solutions to needs expressed in text

• US 9,805,018 B1
• Filed: 07/18/2014
• Issued: 10/31/2017
• Est. Priority Date: 03/15/2013
• Status: Active Grant

• Current Assignee: Steven E. Richfield
• Original Assignee: Steven E. Richfield
• Inventors: Richfield, Steven E.
• Primary Examiner(s): Albertalli, Brain
• Application Number:
  US14/335,685
• Time in Patent Office: 1,201 Days
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 40/163 Handling of whitespace
• CPC Class Codes: G06F 40/205 Parsing
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06Q 30/0241 Advertisements
• CPC Class Codes: G06Q 50/01 Social networking

Abstract

A natural language processing methodology to automatically transform push advertising into pull advertising. Text found in forum, blog, and social media postings throughout the Internet is grammatically analyzed to identify potential customers who have expressed a clear problem. Only parsing rules with the least likely elements present are evaluated. In response, personalized replies are produced that contain pertinent and useful information about a potential product or service. Those replies appear to come from other Internet users, thus converting expressed needs of user/prospects into discussions with sales technicians.

Claims

1. A method of computer-processing natural language text to recognize the presence of words and phrases having certain specific meanings and act upon what has been recognized by sending messages, said method comprising:

a) initializing by storing on an Internet-connected computer processing system parsing rules for which input words contained in textual postings accessed through the Internet are the least-frequently-used of a plurality of words required to satisfy those rules;

b) queuing processing of the parsing rules based on the appearance of the least frequently used word required by the parsing rules, wherein a particular priority queue is pre-selected according to the parsing rule to assure that rules selected by the words that actually occur in each posting are processed in their correct order, said rules including at least one of the rules selected from the list consisting of the correction of spelling errors, the substitution of plain wording for idioms, the division of complex and compound sentences into simple sentences, and to then act on them by outputting responsive messages;

c) receiving at the Internet-connected computer processing system a plurality of input words from textual postings accessed on the Internet;

d) processing the queued rules;

e) determining whether a response is warranted; and

f) where a response is warranted, outputting a responsive message either by directly messaging the author of the textual posting analyzed or by routing responsive messages to a representative.

2. The method of claim 1, wherein during initialization step a) is preceded by the step of storing on the Internet-connected computer processing system a set of input words in a table, along with the ordinal representing their respective positions in the list or a frequency-of-use indicator, the set of stored input words comprising the most commonly used English words, showing their order of frequency of use, thereby providing means to identify the rules for which those words are the least-frequently-used words in the rules by comparing their associated ordinals.

3. The method of claim 1, further including automatically routing messages produced in step f) to the authors of the Internet postings that were processed, by indicating a sender'"'"'s address showing the address of a representative, such that the messages so routed appear to their final recipient to originate directly from representatives.

4. The method of claim 1, wherein step d) includes prioritizing the processing of rules in any convenient order.

5. The method of claim 1, further including:

accessing a plurality of textual postings from the Internet;

tokenizing the text accessed; and

converting the tokenized text to ordinals.

6. The method of claim 5, wherein the step of accessing textual postings is performed by a software program that runs automated tasks over the Internet by a crawling program customized for various Internet domains to extract metadata and relevant posting author information as they crawl, including who made the posting and their contact information, along with other stored metadata indicating what may have been learned about this user from past postings.

7. The method of claim 6, further including the step of running a software module in a user/representative'"'"'s Internet-connected computer that reroutes bidirectional private communications from a central site computer to the Internet to conceal the fact that web accesses are part of a centralized web crawling activity.

8. The method of claim 1, further including using a widely distributed web data extraction program to extract who made the posting and their contact information.

9. The method of claim 8, wherein the step of extracting author information includes accessing information stored in a metadata repository indicating what may have been learned about the posting author from past postings by the same author, and updating information stored in a metadata repository indicating what may have been learned about the posting author from the presently analyzed posting.

10. The method of claim 1, further including recognizing negative statements and weighing negative statements as being more highly negative than positive statements are weighed positively.

11. A method for processing natural language in textual input, as a component of an Internet-connected opportunistic natural language processing AI system, comprising the steps of:

a) converting a standard set of input words to ordinals, the set of input words showing their order of frequency of use, thereby providing means to identify rules by their least-used words by accessing them according to their associated ordinals;

b) associating the input words with parsing rules for which the input words are the least-frequently-used of a plurality of words required to satisfy those rules;

c) accessing textual input;

d) tokenizing the text accessed;

e) converting each word input to an ordinal to identify what rules to perform; and

f) performing the queued rules selected by the words that actually occur in the input words from among the group of rules including at least one of the rules selected from the list consisting of rules to correct spelling errors, rules to substitute plain wording for idioms, rules to divide complex and compound sentences into simple sentences, and act on them by outputting responsive messages.

12. The method of claim 11, wherein said rule to selectively divide complex and compound sentences divides such sentences at conjunctions into simple sentences.

13. The method of claim 11, wherein step c) involves accessing a plurality of textual postings from the Internet.

14. The method of claim 13, further including rerouting bidirectional private communications from a central site computer to the Internet through representatives'"'"' computers to conceal the fact that web accesses are part of a web crawling activity.

15. The method of claim 13, further including routing messages to authors of the textual postings accessed through the accounts of representatives in such a manner that the messages appear to originate directly from sales support personnel without apparent computer generation.