照片见相册。

最近 ， wolframalpha上线了，号称是要和GOOGLE 比拼的一个 SEARCH ENGINE， 卖点在于"Computational Knowledge"。

 

我试用了一下，感觉其特点主要在2方面：

1. 对QUERY的结构化表示 ， 比如shanghai beijing distance 会变成 distance: from shanghai to beijing

 这点和SEMANTIC WEB的理念很像， 而这个功能感觉非常像 JACKSON他们以前ISWC做的 KEYWORD QUERY TO SEMANTIC QUERY (不知道我有没有理解错？）

 

2. 做了非常多的SUBENGINE， 比如 化学分子式的显示 ， 地图的显示 ， 乐谱，计算器等等， 搜索的结果是各个SUBENGINE RESULT的 融合。

 

不过，该SEARCH ENGINE目前对NATURAL LANGUAGE QUERY支持的不好， 而且我个人感觉其背后的INDEX也不是全网的INDEX？

 

 

 

Recent news reported that the MSN QnA beta will be shutdown recent days. Actually I touch this site at 2006. At that time, the site was just built and few user use this site. The site has a feature that the question can just have tags but the question can not be categorized.  In my opinion, the lost of category information introduce difficulty for  user browsing existing questions. Maybe it is good for websites, coz website has the property of diversity, more or less it will have some different features, while a question is more focus. Users can hardly focus on one specific field in the mode of MSN QnA. Also, the MSN QnA did not have a lot of advertisement which made it harder to compete against Yahoo! Answers.

 

so what can be improved in current qa services?

 

In SIGIR09, there existed 2 papers i found related to cqa. They still focused on similar question finding and anwer ranking, which was an old topic as I thought. Although these topics did have some effect for improving current cqa site, they can not exploit the real business value of the qa service.

 

so in my point:

 

in general view, one important thing is to extract the real knowledge from these sites, excluding many spam informations(the spam defined here might be just chatting threads like in forum). The "knowledge" defined here is the "General" , "Abstract" Knowledge which can be extracted and be put into wikipedia not just "query-specific" knowledge.

 

in user view, the large value of these site is to  mine user's information according to their question and answers, including users' interests, jobs, or other information. These data may be a great resource for company business.

 

What's your point?