by Andrew M
The article I decided to talk about this week is entitled “Entity Synonyms for Structured Web Search” and is written by Tao Cheng, Hady Lauw and Stelios Paparizos. This article talks about how currently when users run web searches they are not always getting the desired information they want. Currently, developers manually enter in synonyms for search words or use dictionaries and lookup tables. Another method which can be used is called content analysis. This will actually analyze the search and return more accurate data. All the above methods though are costly and time consuming and many times do return the desired results. The author’s proposed solution is to use something named “entity synonyms.” This in effect links search terms to certain entities so that when someone later searches this term the results will be more accurate. This is done by mining previous search data. Two methods are used to mine the search data. These methods are called Intersecting Page Count (IPC) and Intersecting Click Ratio (ICR). IPC measures how close a website is related to what the user was searching for. If a user stays at a page for a while this is seen as a possible match and is logged. ICR measures the amount a user clicks around a website. The more the user clicks around the more the website is seen as a possible match. Next, a cleaning process is started in which all un-needed terms are cleared from the entity synonym. An example would be if a user searched “Looper Trailer.” The search would look for synonyms for the movie “Looper” and return all possible websites. The word “trailer” would be removed because it is un-needed information.
I believe this article is related to what we are talking about in class. Currently, we are going over the Enhanced Entity Relationship model in class and this article is dealing with using entities and how they are related to queries that users run. This article focuses on how web queries can be made more efficient and how this will help to cut costs eventually.
I personally liked this article. It was very informative and I felt gave some good ideas as to how web results may be made more relevant to the user. The only failing point of this article I felt was that at one point they mentioned that their method is being used to a certain extent in Microsoft’s Bing search engine. I personally have found Bing to be a completely useless search engine and it never finds the information I am looking for. This to a certain extent makes me feel that the author’s proposed solution will not work. Other than that, I felt the article was very informative.
Cheng, Tao; Lauw, Hady W.; Paparizos, Stelios; , “Entity Synonyms for Structured Web Search,” Knowledge and Data Engineering, IEEE Transactions on , vol.24, no.10, pp.1862-1875, Oct. 2012