AI Lab Report

AI Lab Report

“Google it”; this is a common response people have with one another when someone doesn’t know some information they desire to know. Google is the most popular search engine used by people to find simple and broad information. With all the success Google has had, do you think there will ever be another website that can compete with them? A couple of researchers from the Massachusetts Institute of Technology (MIT) seem to think they have the perfect search engine that can compete with Google. GOOSE (goal-oriented search engine) is the name of the search engine that Hugo Liu, Henry Lieberman, and Ted Selker have developed to battle Google. GOOSE is designed to help the novice or experienced search engine users be able to find any information that is on the World Wide Web. Liu, Lieberman, and Selker feel that when individuals have to search for something difficult search engines only appeal to experienced search engine users. With their system the researchers believe that novice and experienced search engine users will be able to use the system with ease and the results they get when searching will satisfy the user’s original search goals. In this paper, I will present background information on GOOSE, an evaluation of certain aspects of the search engine, and what the future holds for GOOSE. In my opinion I don’t think GOOSE will be able to compete with Google or any other search engine that is currently out there. People aren’t always acceptable to change and Google has already taken over the search engine market.

GOOSE: A Goal-Oriented Search Engine With Commonsense is a project that was developed by Hugo Liu, Henry Lieberman, and Ted Selker at MIT in 2002. GOOSE (goal-oriented search engine) is “an adaptive search engine interface that uses natural language processing to parse a user’s search goal, and uses “common sense” reasoning to translate this goal into an effective query” (Liu, Lieberman & Selker, 2002). GOOSE uses a source of common sense knowledge known as Open Mind, which consists of 400,000 simple facts. The Open Mind Commonsense Project is “an endeavor at the MIT Media Laboratory that aims to allow a web-community of teachers to collaboratively build a database of knowledge using diverse representations, and to explore ways to use this knowledge to make computer applications more intelligent and context aware. With the GOOSE system, “the original query is a natural language statement of the user’s search goal, and the reformulation step involves natural language parsing of this statement, followed by inference to generate the query that will best satisfy this goal” (Liu, Lieberman & Selker, 2002). GOOSE performs inference and composes an entirely new query that would best fulfill the user’s goal (Liu, Lieberman & Selker, 2002).

Hugo Liu, Henry Lieberman, and Ted Selker developed GOOSE as a way for a search engine to adapt to the needs of the user. To set themselves apart from other search engines GOOSE has a different interface. The interface for most if not all search engines is a simple text box, in which a user types in a phrase about something they desire to know about. GOOSE’s user interface consists of a pull down menu and a text box to enter the phrase. Currently with GOOSE a user selects one of the five search goals which are: “I want help solving this problem”, “I want to research…”, “I want to find websites about”, “I want to find other people who”, and “I want specific information about the product/service…” (Liu, Lieberman & Selker, 2002). Users select from an option from the pull down menu and then type a phrase in the text box to complete the query. Here is screen shot of the user interface.

To see how people would respond to this type of interface Liu, Lieberman, and Selker conducted preliminary user tests involving four novice search engine users. The team queried simple tasks on the categories and commonsense sub-domains the system knew how to handle (Liu, Lieberman & Selker, 2002). The queries that the users inputted into GOOSE were also sent to Google. After the search was complete users were asked to rank the relevance of the results on a scale from one to ten. From there results they determined that GOOSE was more effective than Google when it comes to producing first page results that were relevant to their search. Here is a summary of the results from the preliminary user tests.

Search Task # successful inferences Avg. score GOOSE Avg. score Google
Solve household problem 7/8 6.1 3.5
Find someone online 4/8 4.0 3.6
Research a product 1/8 5.9 6.1
Learn more about 5/8 5.3 5.0

Currently, no other work is being done on the GOOSE project. The user interface has some promise to it, but the commonsense knowledge database needs more data in it so users will be able to use the search engine more effectively. From the table we see that GOOSE can compete with Google, but there are some issues the developers of GOOSE need to work on before it is put on the Web for people to use. The Open Mind commonsense knowledge database only has 400,000 facts in it. Minsky estimates that in order for GOOSE to be effective it needs 15 million facts in order for it to compare to the knowledge humans possess. GOOSE also doesn’t recognize brand names or things that are trademarked. Another issue I have with GOOSE is with the pull down menus. Right now GOOSE only offers five categories to search from. In its current state only certain things can be searched for using the interface, but in the future more categories will have to be added to the menu. The work of adding more categories to the interface will be a constant thing and developers will have to be sure to not let categories overlap with one another.

In the future the developers of GOOSE are working on some of the issues I discussed in the previous paragraph and striving to accomplish two goals: personalizing the commonsense, and automatic detection of goal categories (Liu, Lieberman & Selker, 2002). When these goals are accomplished the developers feel GOOSE will be helpful enough to compete with Google. The developers of GOOSE feel like if they incorporate a user’s personalize commonsense they will be able to better interpret the user’s search goals and provide more relevant results (Liu, Lieberman & Selker, 2002). Liu, Lieberman, and Selker would like GOOSE to get to the point where the search results would be relevant to the area the user is currently in like Google currently does with information they receive from users. For example, if a user was looking for a cleaner’s then the top results would be cleaner’s that are close to them. Another way GOOSE developers suggest they might get personalize commonsense is either through an interview wizard, mining information about the user from a homepage, or getting shared information from another application that is also learning personalization’s about the user. The second goal of automatic detection of goal categories requires that the categories be validated and more robust natural language processing will be used (Liu, Lieberman & Selker, 2002).

The technology associated with GOOSE is similar to that of other search engines in which they search thousands of computers to process billions of information to find the data pertinent to what the user searched for. When GOOSE is given a search goal and query it performs four steps before the results are presented to the user. The four steps include: 1) “Parsing the query into a semantic frame, 2) classifying the query into a commonsense sub-domain, 3) reformulating the query through commonsense inference guided by expertise templates, and 4) re-ranking results using commonsense concepts” (Liu, Lieberman & Selker, 2002). The concept of GOOSE is a little more involved than other search engines, but it also relies solely on the Open Mind Commonsense Project. Without Open Mind GOOSE wouldn’t be able to do the things it does.

GOOSE is very applicable to other search engines that are available. Search engines aren’t going anywhere, so with the addition of GOOSE it will just be another search engine amongst others. People no longer go to reference books to find out information; they get on the Internet and find the information they desire. If GOOSE is developed more, it could be better than Google. With the pull down menu option on GOOSE, users have a better chance of getting results that are more relevant to what they are searching for; instead of results that you have to browse through to see if they will be helpful or changing your query once you realize this first query wasn’t helpful.

GOOSE has the potential to be as successful as Google, but I don’t know if that will happen due to the success Google is currently having with search engines. Other search engines that were out before Google, such as Ask Jeeves, aren’t competing with them now. GOOSE has the ability to be helpful to both novice and experienced users. Google has so much power that they would eventually figure out the technology behind GOOSE and figure out a better way to be more beneficial to novice users. Google currently does a pretty good job of predicting what the user might be searching for, so it wouldn’t be that hard for them to become a more effective search engine for novice users when it came to complex queries. GOOSE hasn’t been developed anymore since 2002. The developers have some kinks to work out, but by the time they address those issues I think it will be too late. GOOSE developers would be smart to stop while they are ahead. They should save themselves time, money, and the disappointment that could come as a result of GOOSE not reaching the success Google has.

If I had to rate each category (technology, applicability, status, and potential) they would all get considerable ratings. The technology is similar to that of other search engines, but it is a little more developed with the pull down menu. With that said I would rate it 20 out of 25. GOOSE is very applicable to the real world. People are constantly using search engines to find information, so another one will be beneficial to society. GOOSE will probably be more beneficial to novice users because it doesn’t require its users to formulate their query correctly. For this reason I would rate the applicability of GOOSE 20 out of 25. GOOSE has the potential to be a successful search engine, but I don’t know how accepting people would be with a new search engine. There is also some learning that will have to take place since GOOSE has the pull down menu, so I would rate the potential 20 out of 25. GOOSE hasn’t been touched since 2002, so it’s getting to the point where it might be time to give up on the dream. Technology is constantly changing, so I would rate the status of GOOSE 15 out of 25.

GOOSE had the potential to be as successful if not more successful than Google. Now that a great deal of time has passed since it was developed, I don’t think GOOSE will ever see the potential it could’ve had. Google has taken over the search engine market just like Facebook took over the social networking market when it first came out. Now that I’ve written this paper on GOOSE I will always wonder about what could’ve been for this project.

References

Liu, H., Lieberman, H., & Selker, T. (2002). Goose: A goal-oriented search engine with commonsense. Manuscript submitted for publication, Massachusetts Institute of Technology, Retrieved from http://agents.media.mit.edu/projects/goose/AH2002-goose.pdf.


Comments are closed.