Research on user interaction in the most complete search hint history--read "an eye-tracking Study of user Interactions with Query Auto completion"

Search drop-down prompt (Query Auto completion, abbreviated QAC) nowalmost every search engine essential function, the role of the user in the search box input query words in the process, to show users a series of search queries query for users to choose, to facilitate user input, Shorten user search time, improve the user search experience. In theThere have been a lot of research in this area, such as the context-based, the heat of the pre-measured query based on time series, the personalization sort, and so on, but the study of how the user interacts with the search pull-down hint (hereinafter referred to as QAC) is still a blank.

         Microsoft's Katja Hofmann a few days ago (11.3-11.7) just filled the void. She's in at the CIKM 2014 conference, a Paper--an eye-tracking Study of User Interactions with Query Auto completion was published . The interactive process between user and Qac is studied. through the eye localization technology, the user and Qac's interaction is recorded and analyzed, and a series of interesting conclusions are drawn. This paper, on the basis of reading papers, abstracts the conclusions of the paper concisely. Summed up a little of their own revelation.

practical ideas and conclusions :        1. The user's QAC interaction process is divided into: query examination, query formulation, task Completion3 process. Query examination the process of viewing QAC for the user. During this process, the user notices a hint of QAC and focuses attention on the results of the QAC. Query formulation is the process by which users use QAC. The process chosen based on the results provided by QAC.

Task completion is the behavior of the search results page after the user has selected QAC results. Records the completion of a user's search task.
         2. A series of statistical indicators related to QAC are designed for each process to reflect the quality of the QAC, with detailed indicators such as the following:          TFF records the time interval at which the user knocks the keyboard to the QAC drop-down list, This time is not related to the qac of the drop-down results.

CFT records the time the user is concentrating on the Qac drop-down list, which is related to the sorting of the Qac drop-down results, the better the quality of the sorting results, the more the user 's search intentions related to the query, the less time the user stays. UQ represents a single query number submitted by the user, describing a case constructed with query. Assuming the QAC quality is not good, the user is very difficult to construct a more appropriate query. If you cannot construct a query, you will abandon the search. (The author Opinion: may also be good quality, a query search will find the necessary results)UR means that the user opens a search results page to complete the search and describes the quality of the search results. The less open, the higher the quality of the search results, the sooner the search task is finished. TFC and TCT are similar to each other. It's just a statistic from other angles.
3. QAC has a very strong position bias, the higher the ranking of the query is focused and the probability of clicking, such as the following two graphs:
The horizontal axis of the right shows the sort position of query in Qac, the left vertical indicates the user's attention time, and the right vertical axis indicates the probability of clicking. Each location has two different experiments, with the case of sequencing and randomization in the control QAC. it can be seen from the graph that, regardless of whether the QAC results are sorted, the top position of the query gets the highest attention, the probability of being clicked is significantly higher than the other location of the query. This shows that the top of the query is the most likely to be concerned about. Whether or not the query is out of order, only in the top position, will certainly be concerned about, there is a very strong position bias, also on the side of the QAC to determine which query is finally decided which query is queried as the queries. However, it is also possible to observe that the ordered results are more likely to be clicked than the random sort, because sorting will bring the query that is more consistent with the user's search intent to the front.
4. Although the QAC results are sorted or not, the top-ranked query gets the same attention, but the quality of the search results is different. The data of UQ, UR, TFC and TCT recorded by the experiment are shown. High-quality query search results are better. Enables users to find what they need faster. Ends the search task.

Therefore, it is necessary to put more user-intended query into a more forward position, the addition of such a query exposure to help bring better quality of search results (e-commerce to help improve turnover rate)
5. Summarizes the 3 kinds of pattern:a that users use QAC. Users who rely on QAC search. The result of watching QAC while losing; b. Users who occasionally use QAC; c. From users who do not use QAC, this type of user typically focuses on the keyboard during the input process, without looking at the screen. Computer use is not skilled users, the proportion is small.
6. Summarizes the 2 purposes for which users use QAC: A. spelling or spelling correction; b. Query completion
7. The experiment proves that MRR this evaluation qac the rationality of the quality indicator, which is based on this if. The more likely the result of the sort to be noticed by the user, the less likely it is that the results of the sort will be clicked, and the results from the 3 experiment can be verified.

revelation :1. The sorting of the QAC is very important. Directly affect the quality of search results, for e-commerce search, will affect the conversion rate finally. Therefore, it is possible to introduce the quality-related features of the search results page in the sorting, such as CTR, Deal situation, NDCG and so on. 2. One of the features that users use QAC is a spelling hint. Therefore can increase the spelling error correcting the prompt function, not only simple prefix matches.
references :An eye-tracking Study of the User Interactions with Query Auto completion
Reprint Please indicate the source. This article transferred from : http://blog.csdn.net/huagong_adu/article/details/40951409 This blog moved to:http://ralphadu.com/

Copyright notice: This article Bo Master original articles, blogs, without consent may not be reproduced.

Research on user interaction in the most complete search hint history-read "an eye-tracking Study of users Interactions with Query Auto completion"