Data processing: artificial intelligence – Knowledge processing system – Knowledge representation and reasoning technique
Reexamination Certificate
2000-02-14
2003-05-06
Starks, Jr., Wilbert L. (Department: 2121)
Data processing: artificial intelligence
Knowledge processing system
Knowledge representation and reasoning technique
C707S793000
Reexamination Certificate
active
06560590
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to the field of human-computer interaction using natural language. In particular, the present invention discloses methods of using a multiple tiered approach to matching a human's natural language query to elements of a corpus of information, in order present the user with relevant sections of information from the corpus.
BACKGROUND OF THE INVENTION
The Internet evolved during the late 20
th
century to a vast infrastructure of communication that affords billions of people with nearly instant access to millions of World Wide Web sites. Large numbers of commercial Webs sites that sell products and services have blossomed to take advantage of this new communication medium. These Internet-based commercial web sites are often referred to as “ecommerce” sites. Many large ecommerce sites on the Web receive tens of thousands of customer inquiries each day.
The ecommerce companies receive customer inquiries through multiple different transport channels on the Internet including email, web forms, and chat or other real-time interactions with a human customer service agent. To keep their customers happy, these ecommerce sites need efficient systems for responding to these voluminous customer inquiries. To provide simple automated customer support many ecommerce sites provide a search engine to the users. A search engine accepts a set of designated keywords and uses those keywords to locate information related to the keywords.
To provide a simple customer response system, many ecommerce Web sites now offer “free-text interactions” across the Internet. Free-text interactions allow a user to enter a multiple word query in the form of a question or command that is handled automatically by a query resolution system running on a computer system. Free-text interactions often involve referring the user to a certain section of the web site or a certain part of the web site containing a pre-defined set of “Frequently Asked Questions” and answers. The frequently asked questions and answers are commonly referred to as an FAQ. Based on the degree of the best match for the multiple word query, the user may be presented with a set of hyperlinks for destinations on the web site. These destinations may include one or more sections of the FAQ. Alternatively, the user might be presented with a customized web page that contains the relevant FAQ entries in it.
A typical search engine allows a user to enter a small number of keywords as search terms that will be searched for in a target database such as the ecommerce web site. Alternatively, a search engine may be used to handle free-text queries, where selected words from the free-text query are matched against an indexed form of a corpus under examination. To select words in the free-text query for matching, a search engine might select all words besides those considered as “stop words.” Stop words are words that occur often in language and do not convey much information in and of themselves. In the English language, stop words include common prepositions, articles, and conjunctions. For example, “on,” “above”, “in,” “and”, “the” are usually considered stop words.
To simplify human and computer interaction at an automated web site, many web site designers have decided to anthropomorphize the search engine of a web site with a computer-based agent. In this manner, the users will interact with the agent as if the agent had the intelligence or verbal communication skills of a human being. In these human-computer interactions with the agent, the customer may be encouraged by the web site to ask a question in its natural form. For example, the site may present the user with a one-line text box with the prompt: “Please ask a question, such as “What does your company do?′.”
Unfortunately, customers may ask questions that the natural language query facility misinterprets and thus does not provide meaningful results. For example, if a natural language query facility strips out the stop words of “What does your company do?” and applies the keywords to a search engine, the search engine will not likely be able to provide meaningful results. One reason this may occur is that the question “What does your company do” does not contain a set of words that conveys the overall meaning of the query the stop words have been removed. For example, the stop word removal system of one embodiment leaves only the word “company”. It is unlikely that a search of the web site's content for the word “company” will provide meaningful results to the original question. In fact, it is likely that the search engine will return a plethora of irrelevant results.
Natural language query systems on the World Wide Web have proven to be quite popular with the general public. However, the current implementations of natural language query systems often yield inaccurate or limited results. It would therefore be desirable to have improved natural language query systems that provide improved results.
SUMMARY OF THE INVENTION
A system for matching natural language queries to web site content is disclosed. The query resolution system returns zero or more links to content that is relevant to the users query. The present invention for query resolution combines two or more types of natural language query resolution methods, where the knowledge base for each of the methods comes from a single knowledge specification.
The various different natural language query resolution methods differ fundamentally in how they match the user query to the web site content. The results of the resolution methods are ranked and all, some, or none of the results of each system may be displayed.
Other objects, features, and advantages of present invention will be apparent from the company drawings and from the following detailed description.
REFERENCES:
patent: 5404295 (1995-04-01), Katz et al.
Gerard Sabah et al; Carmel: A Step Toward Reflection in Natural Language Understanding Systems; Nov. 1993; IEEE; 1063-6730/93; 258-265.
O'Neill Brian
Shwe Michael
Hirl Joseph P.
Johansen Dag
Kana Software, Inc.
Starks, Jr. Wilbert L.
Stattler Johnson & Adeli LLP
LandOfFree
Method and apparatus for multiple tiered matching of natural... 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 and apparatus for multiple tiered matching of natural..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and apparatus for multiple tiered matching of natural... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3064561