Given the popularity of the World Wide Web and the Internet, users can acquire information relating to almost any topic from a large quantity of information sources. In order to find information, users generally apply various search engines to the task of information retrieval. Search engines allow users to find Web pages containing information or other material on the Internet that contain specific words or phrases.
In general, a keyword search can find, to the best of a computer's ability, all the Web sites that have any information in them related to any key words and phrases that are specified. A search engine site will have a box for users to enter keywords into and a button to press to start the search. Many search engines have tips about how to use keywords to search effectively. Typically, such tips aid users to narrowly define search terms, so that extraneous and unrelated information are not returned and the information retrieval process is not cluttered. Such manual narrowing of terms can mitigate receiving several thousand sites to sort through when looking for specific information.
In some cases, search topics are pre-arranged into topic and subtopic areas. For example, “Yahoo” provides a hierarchically arranged predetermined list of possible topics (e.g., business, government, science, etc.) wherein the user will select a topic and then further choose a subtopic within the list. Another example of predetermined lists of topics is common on desktop personal computer help utilities, wherein a list of help topics and related subtopics are provided to the user. While these predetermined hierarchies may be useful in some contexts, users often need to search for/inquire about information outside of and/or not included within these predetermined lists. Thus, search engines or other search systems are often employed to enable users to direct queries, to find desired information. Nonetheless, during user searches many unrelated results are retrieved, since users may be unsure of how to author or construct a particular query. Moreover, such systems commonly require users to continually modify queries, and refine retrieved search results to obtain a reasonable number of results to examine.
It is not uncommon to type in a word or phrase in a search system input query field, and then retrieve several million results as potential candidates. To make sense of the large number of retrieved candidates, the user will often experiment with other word combinations, to further narrow the list.
In general, the search system will rank the results according to predicted relevance of results for the query. The ranking is typically based on a function that combines many parameters including the similarity of a web page to a query as well as intrinsic quality of the document, often inferred from web topology information. The quality of the user's search experience is directly related to the quality of the ranking function, as the users typically do not view lower-ranked results.
In general, the search system will attempt to match or find all topics relating to the user's query input regardless of whether the “searched for” topics have any contextual relationship to the topical area or category of what the user is actually interested in. As an example, if a user who was interested in astronomy were to input the query “Saturn” into a conventional search system, all types of unrelated results are likely to returned including those relating to cars, car dealers, computer games, and other sites having the word “Saturn”. Another problem with conventional search implementations is that search engines operate the same for all users regardless of different user needs and circumstances. Thus, if two users enter the same search query they typically obtain the same results, regardless of their interests or characteristics, previous search history, current computing context (e.g., files opened), or environmental context (e.g., location, machine being used, time of day, day of week).
Tuning the search ranking functions to return relevant results at the top generally requires significant effort. A general approach for modern search engines is to train ranking functions and set function parameters and weights automatically based on examples of manually rated search results. Human annotators can explicitly rate a set of pages for a query according to perceived relevance, and creating the “gold standard” against which different ranking algorithms can be tuned and evaluated. However, explicit human ratings are expensive and difficult to obtain, often resulting in incompletely trained and suboptimal ranking functions.
The following presents a simplified summary in order to provide a basic understanding of some aspects of the claimed subject matter. This summary is not an extensive overview. It is not intended to identify key/critical elements or to delineate the scope of the claimed subject matter. Its sole purpose is to present some concepts in a simplified form as a prelude to the more detailed description that is presented later.
The subject innovation enhances search rankings in an information retrieval system, via employing a user behavior component that facilitates an automatic interpretation for the collective behavior of users, to estimate user preferences for one item over another item. Such preferences can then be employed for various purposes, such as to improve the ranking of the results. The user behavior component can interact with a search engine(s) and include feedback features that mitigate noise which typically accompany user behavior (e.g., malicious and/or irrational user activity.) By exploiting the aggregate behavior of users (e.g., not treating each user as an individual expert) the subject innovation can mitigate noise and generate relevance judgments from feedback of users. The user behavior component can employ implicit or explicit feedback from users and their interactions with results from previous queries. Key behavioral features include presentation features that can help a user determine whether a result is relevant by looking at the result title and description; browsing features like dwell time on a page, manner of reaching search results (e.g., thru other links) deviation from average time on domain, and the like; clickthrough features such as the number of clicks on a particular result for the query. For a given query-result pair the subject innovation provides multiple observed and derived feature values for each feature type.
The user behavior component can employ a data-driven model of user behavior. For example, the user behavior component can model user web search behavior as if it were generated by two components: a “background” component, (such as users clicking indiscriminately), and a “relevance” component, (such as query-specific behavior that is influenced by the relevance of the result to the query).
According to a further aspect of the subject innovation, the user behavior component can generate and/or model the deviations from the expected user behavior. Hence, derived features can be computed, wherein such derived features explicitly address the deviation of the observed feature value for a given search result from the expected values for a result, with no query-dependent information.
Moreover, the user behavior component of the subject innovation can employ models having two feature types for describing user behavior, namely: direct and deviational, where the former is the directly measured values, and latter is deviation from the expected values estimated from the overall (query-independent) distributions for the corresponding directly observed features. Accordingly, the observed value o of a feature f for a query q and result r, can be expressed as a mixture of two components:
o(q,r,j)=C(r,f)+rel(q,r,j)
where C(r, f) is the prior “background” distribution for values of aggregated across all queries corresponding to r, and rel(q, r, j) is the “relevance” component of the behavior influenced by the relevance of the result to the query. For example, an estimation of relevance of the user behavior can be obtained with clickthrough feature, via a subtraction of background distribution from the observed clickthrough frequency at a given position. To mitigate the effect of individual user variations in behavior, the subject innovation can average feature values across all users and search sessions for each query-result pair. Such aggregation can supply additional robustness, wherein individual “noisy” user interactions are not relied upon.
Accordingly, the user behavior for a query-result pair can be represented by a feature vector that includes both the directly observed features and the derived, “corrected” feature values. Various machine learning techniques can also be employed in conjunction with training ranking algorithms for information retrieval systems. For example, explicit human relevance judgments can initially be provided for various search queries and employed for subsequent training ranking algorithms.
In a related aspect, collective behavior of users interacting with a web search engine can be automatically interpreted in order to predict future user preferences; hence, the system can adapt to changing user behavior patterns and different search settings by automatically retraining the system with the most recent user behavior data.
To the accomplishment of the foregoing and related ends, certain illustrative aspects of the claimed subject matter are described herein in connection with the following description and the annexed drawings. These aspects are indicative of various ways in which the subject matter can be practiced, all of which are intended to be within the scope of the claimed subject matter. Other advantages and novel features may become apparent from the following detailed description when considered in conjunction with the drawings.
The various aspects of the subject innovation are now described with reference to the annexed drawings, wherein like numerals refer to like or corresponding elements throughout. It should be understood, however, that the drawings and detailed description relating thereto are not intended to limit the claimed subject matter to the particular form disclosed. Rather, the intention is to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the claimed subject matter.
As used herein, the terms “component,” “system”, “feature” and the like are also intended to refer to a computer-related entity, either hardware, a combination of hardware and software, software, or software in execution. For example, a component may be, but is not limited to being, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and/or a computer. By way of illustration, both an application running on computer and the computer can be a component. One or more components may reside within a process and/or thread of execution and a component may be localized on one computer and/or distributed between two or more computers.
The word “exemplary” is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects or designs.
Furthermore, the disclosed subject matter may be implemented as a system, method, apparatus, or article of manufacture using standard programming and/or engineering techniques to produce software, firmware, hardware, or any combination thereof to control a computer or processor based device to implement aspects detailed herein. The term computer program as used herein is intended to encompass a computer program accessible from any computer-readable device, carrier, or media. For example, computer readable media can include but are not limited to magnetic storage devices (e.g., hard disk, floppy disk, magnetic strips . . . optical disks (e.g., compact disk (CD), digital versatile disk (DVD) . . . ), smart cards, and flash memory devices (e.g., card, stick). Additionally it should be appreciated that a carrier wave can be employed to carry computer-readable electronic data such as those used in transmitting and receiving electronic mail or in accessing a network such as the Internet or a local area network (LAN). Of course, those skilled in the art will recognize many modifications can be made to this configuration without departing from the scope or spirit of the claimed subject matter.
Turning initially to
The user behavior component 104 can interact with the ranking component. For a given query the user behavior component 104 retrieves the predictions derived from a previously trained behavior model for this query, and reorders the results for the query such that results that appeared relevant for previous users are ranked higher. For example for a given query q, the implicit score ISr can be computed for each result r from available user interaction features, resulting in the implicit rank Ir for each result. A merged score SM(r) can be computed for r by combining the ranks obtained from implicit feedback, Ir with the original rank of r, Or:
The weight wI is a heuristically tuned scaling factor that represents the relative “importance” of the implicit feedback. The query results can be ordered in by decreasing values of SM(r) to produce the final ranking. One particular case of such model arises when setting wI to a very large value, effectively forcing clicked results to be ranked higher than unclicked results—an intuitive and effective heuristic that can be employed as a baseline. In general, the approach above assumes that there are no interactions between the underlying features producing the original web search ranking and the implicit feedback features. Other aspects of the subject innovation relax such assumption by integrating the implicit feedback features directly into the ranking process, as described in detail infra. Moreover, it is to be appreciated that more sophisticated user behavior and ranker combination algorithms can be employed, and are well within the realm of the subject innovation.
o(q,r,j)=C(r,j)+rel(q,r,f)
where C(r, f) is the prior “background” distribution for values off aggregated across all queries corresponding to r, and rel(q, r, j) is the component of the behavior influenced by the relevance of the results. For example, an estimation of relevance of the user behavior can be obtained with clickthrough feature, via a subtraction of background distribution (e.g., noise) from the observed clickthrough frequency at a given position. To mitigate the effect of individual user variations in behavior, the subject innovation can average direct feature values across all users and search sessions for each query-URL pair. Such aggregation can supply additional robustness, wherein individual “noisy” user interactions are not relied upon. Accordingly, the user behavior for a query-URL pair can be represented by a feature vector that includes both the directly observed features and the derived, “corrected” feature values.
Likewise, the browsing feature 420 can capture and quantify aspects of the user web page interactions. For example, the subject innovation can compute deviation of dwell time from expected page dwell time for a query, which allows for modeling intra-query diversity of page browsing behavior. Such can further include both the direct features and the derived features, as described in detail supra. Likewise, clickthrough features 430 are an example of user interaction with the search engine results. For example, clickthrough features can include the number of clicks for a query-result pair, or the deviation from the expected click probability.
As illustrated in
In order to provide a context for the various aspects of the disclosed subject matter,
With reference to
The system bus 918 can be any of several types of bus structure(s) including the memory bus or memory controller, a peripheral bus or external bus, and/or a local bus using any variety of available bus architectures including, but not limited to, 11-bit bus, Industrial Standard Architecture (ISA), Micro-Channel Architecture (MSA), Extended ISA (EISA), Intelligent Drive Electronics (IDE), VESA Local Bus (VLB), Peripheral Component Interconnect (PCI), Universal Serial Bus (USB), Advanced Graphics Port (AGP), Personal Computer Memory Card International Association bus (PCMCIA), and Small Computer Systems Interface (SCSI).
The system memory 916 includes volatile memory 920 and nonvolatile memory 922. The basic input/output system (BIOS), containing the basic routines to transfer information between elements within the computer 912, such as during start-up, is stored in nonvolatile memory 922. By way of illustration, and not limitation, nonvolatile memory 922 can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM), or flash memory. Volatile memory 920 includes random access memory (RAM), which acts as external cache memory. By way of illustration and not limitation, RAM is available in many forms such as synchronous RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct Rambus RAM (DRRAM).
Computer 912 also includes removable/non-removable, volatile/non-volatile computer storage media.
It is to be appreciated that
A user enters commands or information into the computer 912 through input device(s) 936. Input devices 936 include, but are not limited to, a pointing device such as a mouse, trackball, stylus, touch pad, keyboard, microphone, joystick, game pad, satellite dish, scanner, TV tuner card, digital camera, digital video camera, web camera, and the like. These and other input devices connect to the processing unit 914 through the system bus 918 via interface port(s) 938. Interface port(s) 938 include, for example, a serial port, a parallel port, a game port, and a universal serial bus (USB). Output device(s) 940 use some of the same type of ports as input device(s) 936. Thus, for example, a USB port may be used to provide input to computer 912, and to output information from computer 912 to an output device 940. Output adapter 942 is provided to illustrate that there are some output devices 940 like monitors, speakers, and printers, among other output devices 940 that require special adapters. The output adapters 942 include, by way of illustration and not limitation, video and sound cards that provide a means of connection between the output device 940 and the system bus 918. It should be noted that other devices and/or systems of devices provide both input and output capabilities such as remote computer(s) 944.
Computer 912 can operate in a networked environment using logical connections to one or more remote computers, such as remote computer(s) 944. The remote computer(s) 944 can be a personal computer, a server, a router, a network PC, a workstation, a microprocessor based appliance, a peer device or other common network node and the like, and typically includes many or all of the elements described relative to computer 912. For purposes of brevity, only a memory storage device 946 is illustrated with remote computer(s) 944. Remote computer(s) 944 is logically connected to computer 912 through a network interface 948 and then physically connected via communication connection 950. Network interface 948 encompasses communication networks such as local-area networks (LAN) and wide-area networks (WAN). LAN technologies include Fiber Distributed Data Interface (FDDI), Copper Distributed Data Interface (CDDI), Ethernet/IEEE 802.3, Token Ring/IEEE 802.5 and the like. WAN technologies include, but are not limited to, point-to-point links, circuit switching networks like Integrated Services Digital Networks (ISDN) and variations thereon, packet switching networks, and Digital Subscriber Lines (DSL).
Communication connection(s) 950 refers to the hardware/software employed to connect the network interface 948 to the bus 918. While communication connection 950 is shown for illustrative clarity inside computer 912, it can also be external to computer 912. The hardware/software necessary for connection to the network interface 948 includes, for exemplary purposes only, internal and external technologies such as, modems including regular telephone grade modems, cable modems and DSL modems, ISDN adapters, and Ethernet cards.
As used herein, the terms “component,” “system” and the like are intended to refer to a computer-related entity, either hardware, a combination of hardware and software, software, or software in execution. For example, a component may be, but is not limited to being, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and/or a computer. By way of illustration, both an application running on computer and the computer can be a component. One or more components may reside within a process and/or thread of execution and a component may be localized on one computer and/or distributed between two or more computers. The word “exemplary” is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other aspects or designs.
Furthermore, the disclosed subject matter may be implemented as a system, method, apparatus, or article of manufacture using standard programming and/or engineering techniques to produce software, firmware, hardware, or any combination thereof to control a computer or processor based device to implement aspects detailed herein. The term computer program as used herein is intended to encompass a computer program accessible from any computer-readable device, carrier, or media. For example, computer readable media can include but are not limited to magnetic storage devices (e.g., hard disk, floppy disk, magnetic strips . . . optical disks (e.g., compact disk (CD), digital versatile disk (DVD). . . ), smart cards, and flash memory devices (e.g., card, stick). Additionally it should be appreciated that a carrier wave can be employed to carry computer-readable electronic data such as those used in transmitting and receiving electronic mail or in accessing a network such as the Internet or a local area network (LAN). Of course, those skilled in the art will recognize many modifications can be made to this configuration without departing from the scope or spirit of the claimed subject matter.
What has been described above includes various exemplary aspects. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing these aspects, but one of ordinary skill in the art may recognize that many further combinations and permutations are possible. Accordingly, the aspects described herein are intended to embrace all such alterations, modifications and variations that fall within the spirit and scope of the appended claims.
Furthermore, to the extent that the term “includes” is used in either the detailed description or the claims, such term is intended to be inclusive in a manner similar to the term “comprising” as “comprising” is interpreted when employed as a transitional word in a claim.
This is an application claiming benefit under 35 U.S.C. 119(e) of U.S. Provisional Patent Application Ser. No. 60/778,650 filed on Mar. 2, 2006. The entirety of this application is hereby incorporated herein by reference.
Number | Date | Country | |
---|---|---|---|
60778650 | Mar 2006 | US |