About Lexxe
News and Press
Some international start-up companies are working these days on new search engines, trying to give more accurate answers than Google. Lexxe replies to questions as if it was a real partner for a discussion, and Wink is assisted by surfers. In the future one of these may succeed in defeating the current search giant.
Forster, the international research company, has reported recently that half of those who use the Internet are disappointed by the current search engines. The results are more numerous than ever before, burdened by superfluous links and advertisements, and the users must choose more carefully the search terms and filter the results among thousands of sites that are offered to them, in a process that takes too long.
Some small companies have recently started to build new search engines that will try to respond to users’ expectations. Third generation search engines will not present to a user the maximum amount of information and data on any term that he will put in, but instead will aspire to give him as much as possible accurate answers. To do that, they have to understand better than Google what the user really wants, and the subject of each page on the Internet.
Do such start-up companies have a chance vis a vis Google, that enjoys after all enormous popularity and commands some 50% of the trade in a market that runs six billion dollars annually? “History taught us that Microsoft overcame IBM and succeeded in ruling the world of computing, and that it was Google’s turn to beat Microsoft”, says Dr. Hong Liang Qiao, CEO of Lexxe, which offers one of the promising 3rd generation search engines in this domain. “If your technology is superior and your have enough courage, you have a chance”.
The following list does not exhaust all the companies that try to position themselves in the market for search engines, but is rather focused on the companies that have the greatest chance of success in the next year or two. Any of these is a natural candidate for acquisition by the large search companies. Another common denominator: all these engines are at the initial development stages and their support of Hebrew is very slight, if at all.
Lexxe (Lexxe.com) is a search engine, which answers directly and simply to users’ queries. Like a human discussant, it analyses the question, tries to understand the intention of the user and provides a short answer in a few words. In this, Hong believes, “it overcomes the main weakness of the engines of the second generation, which provide only long lists of links and summaries.”
The idea behind Lexxe is to build it as an intelligent brain that can understand human language (English, for now). It knows how to answer questions such as “Where is Osama Bin Laden hiding?” (in Afghanistan), “How many amino acids are available?” (20), or “Who was the soloist of the Queen Group?” (Freddy Mercury). In addition to the answer, it presents the Internet pages that are directly connected to the question and suggestions for additional search terms that are more focused. In reply to the very same questions, Google and Yahoo refer the user to an endless list of links, out of which he has to “fish” the answer by himself.
The technology of the search engine is based on advanced studies in Artificial Intelligence and Linguistics. Lexxe tries to identify expression within the user’s questions, and to substitute the words of the question to a variety of keywords. Thus, a question like “How old is Ariel Sharon?” will not produce links to pages in which you have the single words “how”, “old”, and “Ariel Sharon”, but instead an accurate answer and pages that refer to the age of Sharon.
Lexxe is not the first engine that aspires to answer questions. The veteran AskJeevs is doing that since 1996’ but is inferior to Lexxe in its language comprehension. A question like “Who did Lee Oswald kill?” cannot be answered by Jeeves, because it cannot process auxiliary verbs like “did”. Google, that sometimes answers questions out of the databases that it scans, makes a mistake in understanding this question and replies: Jack Ruby (Oswald’s murderer). In contrast, Lexxe replies: John Kennedy.
Hakia (hakia.com) and Brainboost (brainboost.com) are two additional search engines which develop similar technology to that of Lexxe and aspire to provide answers instead of results. Hong explains the difference between them and the old engines which tried to answer questions: “Other search engines which purport to provide answers to questions do this by using databases that they build and update themselves, and not by drawing the answer from the Internet”. Lexxe, like Hakia and Brainboost gradually learn to understand Web pages as if a person would read them, and to answer questions without being dependent on any list of facts.
Nevertheless, the dependence of Lexxe on Web pages which do not undergo editing and checking is also detrimental. Some of the answers are quite strange in the best case and erroneous in other cases. On the question, “Why was Rabin murdered?” Lexxe answered “God did not want that he will make peace”. When queried, “Who was Deep Throat?” it answered, “A woman”. In many cases the engine provides a correct answer, but in order that children and other users will be able to trust its first answer and not have to test it further, it needs to be improved. On the question, “Which is the best search engine?” Lexxe indeed provides a correct answer: “Google”.
Using the social network
If the first priority challenge of search engines is to understand the user’s intentions and requirements, the second one is to identify the subject of each page on the Web. When the engines will better understand what each page is talking about, there would be less irrelevant results for each query. An additional genre of third-generation engines tries to solve the problem of inflation in search results and to save on the number of links by using social networks that have become popular in the last couple of years.
Services such as Del.icio.us for keeping bookmarks (favorite sites) on the Web, Flickr for photographs, and Digg.com which collects technology news from users, have created a new dimension on the Internet: sort of a mirror-network in which for each page, picture or news article there is a reflection that provides the descriptions given to them by users, so that they can share these with others and find them more easily in the future. This dimension, sometimes called “The Semantic Web”, is supposed to assist the search engines of tomorrow to better understand the subject of each page on the Internet.
Wink (Wink.com), for example, is a revolutionary search engine which utilizes the information that is aggregated in the Semantic Web to drop results that are irrelevant to the user’s request. The results that are provided by Wink include sites that have been recommended and saved by other users of this service, and are sorted by the number of users who had recommended them. However, as not all the Web is mapped in terms of the Semantic Web, one cannot yet rely on social networks by themselves. As a backup, Wink also provides the results from Google to the requested term.
The name of the algorithm for ranking the search results in Wink is TagRank, which is parallel to Google’s PageRank. TagRank tries to determine which results will appear first, by the number of other users who had described these pages in words that are similar to the search terms. Thus, in the search for the word “Israel”, for example, some pages will be presented on the basis of other users relating them to this country even though the word “Israel” did not appear in them at all.
Any user of Wink can influence future results in the course of his search. If a site in a list fits to a large extent the search term, it can be marked and thus its rank can be upgraded for the next search. The main drawback of engines of this type, such as jookster.com and Wink, that is currently open only for a group of testers, is due to the young age of the Semantic Web. Only a small proportion of Web pages were described until now by users of social networks, and that is why Wink and Jookster are good mainly for finding popular contents that had been propounded over the Web. An additional problem is the “dictatorship of the majority”: popular Web pages will push aside from search results some marginal sites that might indeed contain the desired answer.
Meta-search: The next generation
Services such as MetaCrawler, which combined results from a number of search engines, were the “last cry” in the 90’s, but became less popular as Google has overtaken the market. Indeed, it became pointless to draw results from a number of engines when Google supplied the best results in one place.
In recent years there appeared topical engines that search only in blogs, news sites, images or videos, and at their side other large search engines with good results, such as MSN. One can get results from all these engines together using meta-search engines such as Gada.be that had been created by the technology guru Chris Pirillo. Gada.be draws results from 60 search engines and can provide them on a regular basis by RSS.
The great idea behind Gada.be is to make the search not only multi-engine, but also future-oriented. The support of RSS enables a user who is searching, for example, papers on Alzheimer, not to get only existing articles in this field but also to get updates on every new article that will be published on the Internet. Google, until now, does not support this capability.
|
