Biasing queries to determine suggested queries
专利汇可以提供Biasing queries to determine suggested queries专利检索，专利查询，专利分析的服务。并且Applying a bias when determining a suggested search query. Examples of biases that can be applied include, but are not limited to, temporal biases and monetization biases. Temporal biasing involves increasing a weight associated with a search query, based on a temporal attribute associated with the query. Search queries may also have associated with them a parameter such as frequency, count, etc. One of these parameters may serve as a weight. In order to determine suggested search queries, the temporal attribute is used to modify or bias the parameter (e.g., frequency parameter). Thus, the weight of the search query is modified to temporally bias the query. The suggested search queries are determined based on the biased parameter.，下面是Biasing queries to determine suggested queries专利的具体信息内容。
What is claimed is:
This application is a continuation-in-part and claims priority of U.S. patent Ser. No. 11/313,525, entitled INTERACTIVE SEARCH ENGINE, filed by Richard Kasperski, on Dec. 20, 2005, the contents of which are herein incorporated by reference as if originally set forth herein.
This application is a continuation-in-part and claims priority of U.S. patent Ser. No. 11/212,044, entitled SPECULATIVE SEARCH RESULTS FOR A SEARCH QUERY, filed by Stephen Hood et al., on Aug. 24, 2005, the contents of which are herein incorporated by reference as if originally set forth herein.
A portion of the disclosure of this patent document contains material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document of the patent disclosure, as it appears in the United States Patent and Trademark patent file or records, but otherwise reserves all copyright rights whatsoever.
The present invention relates to search queries, and in particular, to determining suggested search queries by biasing search queries.
A search engine is a computer program that helps a user to locate information. Using a search engine, a user can enter one or more search query terms and obtain a list of resources that contain or are associated with subject matter that matches those search query terms. While search engines may be applied in a variety of contexts, search engines are especially useful for locating resources that are accessible through the Internet. Resources that may be located through a search engine include, for example, files whose content is composed in a page description language such as Hypertext Markup Language (HTML). Such files are typically called pages. One can use a search engine to generate a list of Universal Resource Locators (URLs) and/or HTML links to files, or pages, that are likely to be of interest.
The search engine typically has an interface allowing users to specify search criteria and an interface displaying the search results. Typically, the search engine orders the search results prior to presenting the search results interface to the user. The order usually takes the form of a “ranking,” where the document with the highest ranking is the document considered most likely to satisfy the interest reflected in the search criteria specified by the user. The user is sent one (or more) search results pages based on the ranking. However, the user must still spend considerable time and effort processing the search results to determine whether the search query produced adequate search results. If the user is not satisfied with the results, the user forms a new search query and repeats the process.
Thus, the search process is often an iterative task in which the user forms a search query, determines whether the voluminous search results are adequate, and then re-formulates the search query, if needed. Therefore, the user experience with search engines is often frustrating and time consuming.
The approaches described in this section are approaches that could be pursued, but not necessarily approaches that have been previously conceived or pursued. Therefore, unless otherwise indicated, it should not be assumed that any of the approaches described in this section qualify as prior art merely by virtue of their inclusion in this section.
The present invention is illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which like reference numerals refer to similar elements and in which:
In the following description, for the purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present invention. It will be apparent, however, that the present invention may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to avoid unnecessarily obscuring the present invention.
Typically, when forming a search of the World Wide Web, or the like, a user will formulate a search query and then enter a carriage return, or click on a “search button”, to initiate a search request based on the search query. The act that initiates the search request also serves to indicate that the search query has been “submitted”. Prior to the act that initiates the search, the search query is “not-yet-submitted”. Embodiments of the present invention perform a search and provide a speculative search result based on a not-yet-submitted search query prior to the user explicitly initiating the search request.
Techniques are described herein for providing a speculative search result for a search query prior to finalization of the search query. According to one embodiment, in response to receiving a not-yet-submitted search query from a client node, a speculative search result is provided to the client node for the not-yet-submitted search query prior to receiving an indication from the client node that said search query is completely formed. The speculative search result may be displayed on the same web page on the client node as the not-yet-submitted search query, while the search query is still being entered by the user. As the user continues to formulate the search query, a new speculative search result may be provided to the user.
The search request and search result may pertain to the World Wide Web; however, the techniques described herein are not so limited. More generally, the search request and result may pertain to any searchable data in any format. For example, the data may be a user's address book, saved links, personally stored images (such as .jpeg files, gif files, etc.), video, audio (.mp3 files, .wmv files, etc.), contact cards (e.g., v-cards), calendar objects, word processing documents, graphics files, or computer code (object files and source code).
To limit the frequency with which new speculative search results are provided during the entry of the search query, a determination may be made as to whether the not-yet-submitted search query meets criteria for initiating a search. Such criteria are referred to herein as “speculative search initiation criteria”. The speculative search initiation criteria may be designed to limit the speculative search results to highly relevant results. For example, a criterion may be that the user has entered a complete word in the not-yet-submitted search query.
The speculative search initiation criteria can be used to ensure that the user is not inundated with frequently changing search results as the search query is formed. Moreover, requiring the satisfaction of speculative search initiation criteria helps ensure that the search engine is not inundated with search requests. Further, performing no search until satisfaction of speculative search initiation criteria may increase the likelihood that a speculative search result will have a sufficient relevancy to warrant its return to the user. For example, a speculative search result that is based on some partially formed words may have a low relevancy. However, there are situations in which a speculative search result based on a partially formed word may be highly relevant. Techniques are provided to determine which partial search queries, whether a partial word or not, are likely to lead to a relevant speculative search result.
One technique to determine whether the not-yet-submitted search query meets criteria for initiating a search and is thus sufficiently well-formed is to time how long it has been since the user last entered a character in the search request. If the timer has expired, then a search is initiated. The period of the timer can be dynamically adjusted to adapt to the user. For example, if user types too fast, then a search may be initiated too infrequently. Alternatively, if the user types slowly, then a search may be initiated too frequently.
Another technique to determine whether the not-yet-submitted search query meets speculative search initiation criteria is to determine if a given character is included in the search query. For example, if the user enters a space character, it is likely that the user has completed a word, wherein the search query may be sufficiently well-formed to initiate a search. As used herein, a “word” may be either a word that occurs in a natural language, such as English, or a special notation, such as an abbreviation, acronym, product code, URL, etc.
Another technique to determine whether the not-yet-submitted search query is sufficiently well-formed to initiate a search is to determine if the search query matches or nearly matches a phrase in a dictionary that comprises pre-defined phrases. The pre-defined phrases may be phrases that are likely combinations of words in search requests. In one embodiment, the dictionary is based on popular queries. In one embodiment, the dictionary is used to limit the number of search queries that are performed by determining if the not-yet-submitted search query matches, or nearly matches, a term or phrase in the dictionary, as a condition for performing a search.
The determination as to whether the not-yet-submitted search query is sufficiently well-formed to initiate a search is not limited to the techniques described herein. Moreover, the determination can be made using any combination of the techniques. The determination may be made at the client node, although this is not a requirement. In one embodiment, the client node has an AJAX (Asynchronous JavaScript+XML) engine to facilitate the determination as to whether to forward the not-yet-submitted search query to a search engine. The client node may make the determination by other techniques. In another embodiment, the determination of whether to initiate the search is made at a node other than the client, such as at the search engine.
After the speculative search initiation criteria have been satisfied, pre-defined search rules are applied to provide the user with one or a few highly relevant speculative search results, which may be displayed on the same web page as the not-yet-submitted search query. For example, the pre-defined rules may include performing a first search that is likely to produce a highly relevant speculative search result. If the first search does not produce a speculative search result that meets a certain relevance threshold, then the user is not presented with the search results. Instead, a second search is performed. The second search is performed based on the current state of the not-yet-submitted search query, which may be different than the state of the not-yet-submitted search query upon which the first search was performed. If the results of the second search do not meet the relevance threshold, the results of the second search are not presented to the user. The process of repeatedly searching based on the not-yet-submitted search query may continue until the search query is submitted. Until the search query is submitted, the user is only presented with the results of speculative searches when those results meet the relevance threshold.
The different searches can employ different search techniques and/or may search different databases. This ordered searching, and the use of a relevancy threshold, provides the user with highly relevant results and avoids overloading the user with results, as no results need be returned to the user if the speculative search result relevancy or other criterion is deemed too low.
In
The speculative search result 106a is one that is determined to have sufficient relevancy to merit providing it to the user. For example, it is undesirable to inundate the user with results having low relevancy. Embodiments of the present invention provide various techniques for determining what speculative search result has sufficient relevancy to merit returning to the user. Prior to completing the search query 101a, the user may click on the hypertext link 106a in the speculative search result 106a, if desired.
The speculative search result 106a may be updated as the user further enters the search query.
The speculative search result is not necessarily updated each time the user enters a new character in the search query 101. Referring now to
Referring now to
As used throughout this description, a not-yet-submitted search query means a search query for which no indication has been received, by a user interface for entering the search query, that the search request is complete. For example, when forming a search of the world wide web, or the like, a user will typically enter a carriage return or click on a “search button” to initiate the search request, thus indicating the search query is complete.
In block 204, a determination is made that the not-yet-submitted search query meets the speculative search initiation criteria. When the speculative search initiation criteria are satisfied, a search is initiated, in block 206. If the not-yet-submitted search query is not sufficiently well-formed to satisfy the speculative search initiation criteria, then the process 200 returns to block 202 to receive a further entry of the not-yet-submitted search query.
In block 206, a search is performed on the not-yet-submitted search query. The search may be performed in any convenient technique and may include several stages. An embodiment for performing the search is described in
In block 208, a speculative search result is provided to the user. Exemplary speculative search results 106 are depicted in
If the user selects the speculative search result, in block 210, then content related to the speculative search result is provided, in block 212. For example, if user selects one of the links in the speculative search results 106 depicted in
If the user provides an indication that the search query is complete, then the user is provided search results that might be returned in a conventional search. Thus, if an indication that the search query is complete is received in block 214, then a complete search result is provided to the user in block 216. For example, if the user clicks on a search button (
The process 200 returns to block 202 if the user does not indicate that the search query is complete, wherein a further entry to the not-yet-submitted search query is received. The process 200 continues and may finish when the user selects one of the speculative search results for further information or indicates that the search query is complete.
After the decision to initiate a speculative search has been made, and before the user has explicitly initiated the search request, pre-defined search rules are applied to provide the user with one or a few highly relevant search results, which may be displayed on the same web page as the still being formed search query.
In block 304, a determination is made as to whether the search query triggers a pre-defined search result that may include an organized collection of information related to the search query. For example, based on an analysis of many searches, a set of pre-defined search results are formed that include various links related to a search query. As an example, the speculative search result for the stock depicted in
If the search query matches (or nearly matches) a pre-defined search result in block 304, then the pre-defined search result is provided in block 306. For example, a web server transfers the pre-defined search result to the client node, wherein an application executing on the client node is able to display the result on the same web page as the search query being formed.
In block 308, the not-yet-submitted search query is analyzed to determine if it may be a search query to obtain a navigational link, such as a URL. For example, many users employ a search query to obtain a URL if the user is not certain of the URL. By analyzing the not-yet-submitted search query, a determination can be made if the search query appears to be an attempt to locate a URL. In some cases, the user may type a search query that would be expected to logically relate to a URL; however, the actual URL is different than would be expected. Built in intelligence can correct such a situation, such that the user is provided with a URL to a highly relevant web site.
If the determination in block 308 is that the search query may be for a navigational link, then the navigational link is provided to the user in block 310. For example, the search engine forwards the web link to the client node, wherein an application executing on the client node displays the navigational link on the same web page as the search query.
If the not-yet-submitted search query does not trigger either a pre-defined search result or a navigational link, then a determination is made as to whether to provide a fallback search result to the user, in block 312. The relevancy of a potential fallback result can be assessed prior to providing it to the user, such that the user is not inundated with results having low relevancy. In one embodiment of the present invention, a dictionary is used to filter the search query prior to performing the fallback search. Thus, if the search query does not match, or nearly match, a term or phrase in the dictionary, the process 200 goes to block 314 in which a fallback search is not performed. In one embodiment, the phrases in the dictionary are learned from a statistical analysis of searches.
If test of block 312 determines that a fallback search should be performed, then one or more fallback search results are selected for providing to the user, in block 316. The fallback search results may include a title of a web page, one or more abstracts of a page, and a Uniform Resource Locator (URL) of a page. The fallback result can provide a single result (or limited number of results) of a web search or partial web search. Moreover, the scope of the search can be adjusted to impact the expected relevancy of search results. Fallback search results are described in greater detail hereafter.
Rather than provide uncommon search results, which may have a low relevancy, the fallback search can be confined to a set of common search results. For example, the results of the top 50,000 searches are tracked, wherein a fallback search result is only provided of the not-yet-submitted search query triggers one of the popular search results. Any convenient number of popular search results may be used. Thus, embodiments of the present invention avoid providing the user with a non-relevant search result just because a search based on a not-yet-submitted search query triggers an obscure web page. The fallback search results can be based on any set of user searches ranging from everyone who uses the search engine to only the user who is forming the current search query.
In order to simplify the user experience, a limited number of fallback results are shown on the search query page. For example, only a single result may be provided to the user in one embodiment. Typically, results of searchers are accorded a relevancy. In one embodiment, the search result with the highest relevancy is provided. Alternatively, a few results can be provided based on an analysis of the relevancy. In one embodiment, the search result is only provided if its relevancy exceeds a threshold. Multiple results can be provided if the spread in relevancy between top results is close. The threshold can be dynamic. An abstract can be shown with the result to help guide the user as to its relevancy.
Relevancy may be determined by a variety of techniques. For example, a page's relevance score may be based on how many words, from the search query, are contained in the page, and/or how frequently those words occur in the page. A page's relevance score may also take into account some “page quality metrics” like a citation index, the authoritativeness of the page source, etc. A page's relevance may also be defined relative to the user who submitted the query. For example, relevance may be based on the language used to submit the query (e.g., English or French). For another example, relevance may be based on properties of the user (e.g., the user's geographic region, gender, or social group). Thus, relevance scores for pages that result from query submitted by a user in the United Kingdom may be determined differently than relevance scores for pages that result from the same query submitted by a user in the United States.
In order to improve the quality of a search query, an embodiment of the present invention suggests one or more alternatives to a search query that is being formed, prior to the user indicating that the search query is complete. For example, a user may enter a preliminary search query, and the suggestion mechanism may determine that the preliminary search query has a possible spelling error. The suggestion mechanism then prompts the user with an alternative spelling suggestion, in accordance with an embodiment of the present invention. Moreover, the spelling suggestion is provided prior to the user indicating that the search query is complete. The user may click on a provided link associated with the spelling suggestion to cause the spelling in the search box to be automatically corrected. Further, responsive to the user clicking the link associated with the suggested spelling, the user will be presented with one or more search results based on the query with the suggested spelling. The search results may be in the form of navigational links to web pages that satisfy the query.
In another embodiment, the user need not provide any specific input to be presented with the alternative query suggestions. Rather, the user is prompted to try another search query based on an analysis of the not-yet-submitted search query. The suggested alternative search query is provided responsive to the not-yet-submitted search query, while the user is still entering the search query.
In one embodiment, a history of a user's search requests is tracked and used to determine a relevant search result. For example, if the user frequently accesses a certain weather report, a link may be automatically provided upon the user typing in the beginning of the search query.
One type of speculation to enhance the user experience is to speculate as to what the not-yet-submitted search query will turn out to be and proactively search based on the speculated complete search query. For example, a database that stores previously submitted user search queries can be searched for queries that contain the partially formed search query in order to speculate as to what the complete search query might be. For example, if the partially formed search query is “britney sp”, then a predicted search query might be “britney spears”. Alternatively, information that is user specific such as an address or phone book can be searched with the partially formed search query to speculate as to what the complete search query might be. For example, if the phrase, “john telep” is entered by the user as a not-yet-submitted search query, the user's telephone/address book can be searched to find the telephone number of anyone named John. The telephone number is provided to the user responsive to the not-yet-submitted search query. Other user databases could be searched as well, such as e-mail, note pad, favorite places, history, etc.
Embodiments of the present invention analyze the search query for keywords to determine what type of a search to perform. For example,
In one embodiment, an interface to a search engine is provided to assist the user 1) by predicting what the user is searching for based on the character or characters the user has entered so far in the query field of the interface, and 2) by providing search results to the user via the interface without the user having to formally issue the intended query. For each character entered into the query field, that portion of the query entered so far is automatically issued to a query predictor that determines a set of one or more predicted queries that correspond to the portion of the query. The set of predicted queries is determined based on one or more factors. Such factors may include the frequency of previously issued queries alone and/or when the previous queries were issued.
In one embodiment, the search mechanism determines which of the possible predicted search queries is “most likely” the query intended by the user. The most likely predicted query is processed by a search engine to obtain search results. Both the predicted queries and the search results are provided to the user via the interactive user interface. The predicted queries, when displayed to the user, may be ordered based on their popularity (frequency-based) alone or also on their temporal relevance (time-based).
If the user is not interested in the search results based on the most likely predicted query, then the user may select any query in the set of predicted queries. When the user selects a different predicted query in the list, the search results are then updated to display the search results that pertain to the different predicted query.
In addition to displaying the predicted queries and search results to the user via the user interface, other dynamic data may be provided that relate to the most likely predicted query but are not necessarily obtained by the search results, such as advertisements and other related links to websites.
A search engine executes the particular predicted query 608 from the set of predicted queries and sends the search results 612 to the user to be displayed, for example in results page 610. It is thus possible that the user only has to enter one or a few characters before the actual intended query is determined and the results of the intended query are displayed. Therefore, search button 606 may never have to be selected in order for an intended query to be issued.
In one embodiment of the invention, the portion of the query entered so far by a user is sent from the user's web browser to a query predictor across a network. This may occur for each character, or sequence of characters, entered by the user. The query predictor examines the one or more characters and makes one or more predictions on what the intended query is. The one or more predictions are in the form of one or more completed queries, each of which was a previously issued query. These one or more predictions are sent to and displayed on the user's computer; effectively assisting the user in formulating a query before the user is finished typing the entire intended query in the query field.
In one embodiment, the query predictor operates according to the assumption that it is highly probable that a user intends to issue a query that at least one other person has issued previously. Thus, the initial set of “candidate” suggested queries may be a set of queries that were previously received. By using the information about previously received queries, a highly interactive search engine may assist the user in formulating a query, or perhaps aid the user in refining a query by listing other possible variants of the query that the user would be interested in.
To use previously received queries as candidates for suggested queries, the query predictor must have access to a repository of previously received queries. According to one embodiment, every previously issued query is saved and logged because, if the query was valuable to one user, it has potential value for another user.
It would be inefficient to “suggest” to each user a list that includes all previously-received queries. Therefore, the set of suggested query candidates is filtered to present only those suggested queries that are likely to correspond with the user's intent. One of those factors may be, for example, the frequency with which each candidate search query was previously submitted. In order to use frequency as a selection factor, repository of previously-received queries also includes information about how frequently each of the queries was received. Query completion is based, at least in part, on a frequency with which a search query was previously submitted, in one embodiment.
In one embodiment, the query predictor extends to other languages and is not exclusive to English. The query predictor may also support other types of strings, such as product names and part numbers where the user may only know a small piece of it.
The query predictor thus has a searchable database of previously-received queries that the query predictor may access once the query predictor has received one or more characters from the user. Based on the partial query, the query predictor determines one or more completed queries from the database of queries that match lexically. However, instead of simply completing the partial query lexically and presenting to the user only those queries that begin with the character or characters in the partial query, other queries are also found that contain the lexical completion anywhere in the predicted query. For example, if a user enters the string “th”, not only may “theory of evolution” be a predicted query sent to the user, but also “string theory” or “music theory,” each of which are not simple lexical completions of “th.”
As previously discussed, the process of selecting which of the candidate queries to suggest may be based on the frequency with which each candidate search query was submitted. In one embodiment, query prediction is based on both frequency and a temporal attribute of a search query. For example, query prediction is based on how many times a query was submitted and when the query was submitted.
For example, many previously issued queries may begin with “th.” The most useful queries are likely the ones that not only have issued most often (high frequency), but also those that have been issued most recently (temporal relevance). Therefore, in one embodiment of the invention, the query predictor determines which queries to present to a user by filtering the suggested query candidates based on the frequency with which they were issued in a recent time interval and an older time interval, with the queries issued in the recent time interval weighted more heavily. For example, the fact that a query was previously received five times today may count twice as much as the fact that the query was received five times last week. The most recently issued queries are weighted more heavily based on the premise that a user is more likely to be interested in a subject that many other people are interested in at roughly the same time.
In one embodiment, the time component is determined by searching at least two databases, one for relatively recent queries (a “recent database”) and one for relatively older queries (an “older database”), and then scaling the results from searching the recent database and weighting them accordingly. Scaling is performed to account for differences between the size of the recent database and the older database. Thus, scaling is performed to normalize the data in the two databases. Weighting is performed to favor (or bias) the more recent queries over the older queries.
The relatively recent queries and the relatively older queries are stored in the same database, in one embodiment. The recent and older queries may be stored in separate partitions of the database. However, other techniques can be used to logically differentiate the recent from the older queries, such as associating an attribute (e.g., timestamp) with each search query.
As an example, although “renewable energy sources” may appear in the older database five times more often than “nuclear energy,” the partial query “ener” will cause the query predictor to select “nuclear energy” as the particular predicted query if “nuclear energy” was issued much more frequently in the last week due to a hypothetically recent announcement by Congress that 100 nuclear reactors will be constructed. Moreover, the frequency associated with recently issued queries is adjusted upward to temporally bias the recent search queries, in one embodiment. Therefore, a query in the recent database will be selected over a query having the same frequency in the older database, in this embodiment.
In one embodiment, a characteristic associated with stored search queries (such as frequency) is biased based on another characteristic. For the purpose of explanation, the characteristic that is biased is referred to herein as a “parameter” associated with the search query, and the characteristic that is used to bias the parameter is referred to as an “attribute” associated with the search query.
In one embodiment, this “biasing” is performed ahead of time, so that the biased parameter can be used to filter the suggested query candidates for a query without having to incur the biasing overhead during the processing of the query. In other words, parameter values associated with search queries (e.g. frequency values) are “pre-conditioned” based on an attribute (such as when the search queries were submitted).
For the following example, assume that an “older database” stores search query information for queries submitted over the last 19 months and a “recent database” stores search query information for queries submitted in the last week. Each search query has associated with it a frequency parameter that is based on how many times the search query was submitted by users over a certain time period (e.g., query submissions per hour). As a particular example, prior to biasing, the query “nuclear energy” has a frequency parameter of 100 in the older database and 200 in the recent database. In order to give the more recent search queries more weight, the frequency parameter associated with search queries in the recent database are biased, in one embodiment. For example, the frequency parameter for “nuclear energy” in the recent database is adjusted to 240. Thus, a temporal bias is given to search queries that were recently submitted. The bias can be for any other time period.
Further, the bias can be other than temporal, such as a monetization bias. For example, a search query is assigned a monetization attribute based on its association with revenue generation. As a particular example, a search query that is associated with shopping might be assigned a relatively high monetization attribute.
The bias can be based on many other attributes associated with a query. For example, assume that a database has a language tag for each query, in addition to the frequency of the query. At runtime, the language of the query and the language of the user can be used to modify the frequency value to give preference to queries that are in the language of the other. As another example, instead of language, country of origin of the search query can be used, with a bias given towards queries that have the same country of origin as the user.
A still further example of an attribute upon which to base a bias is some ontology. Examples are health, science, sports, politics, movies, television, music. When the user is searching for health based items, queries that are suggested to a user can be biased towards health based items, for example. This requires knowledge of a user preference or characteristic. A persistent state of the user is kept, such as the category of the last query the user clicked on, to indicate what their category of interest is, in one embodiment.
Furthermore, rather than biasing a frequency parameter, another parameter associated with a search query can be biased, such as count. More generally, the queries can be ordered by any first factor and then re-ordered by a second factor. For example, the queries can be ordered by a monetization attribute and then biased by a temporal attribute. Thus, any of the attributes discussed herein can be used a parameter that is biased or conditioned by some other factor.
As another example, a probability parameter associated with a query can be biased with some attribute. For example, queries that are most probable are selected and the probabilities are conditioned bases on some attribute such as time or monetization. More generally, the parameter that is biased can be any metric where the metric indicates importance for the user and the metric has the property that A>B implies A is better than B or that says A<B is better.
Pointwise mutual information is still another example of a parameter that can be biased or conditioned on an attribute. Equation 3 describes a way to calculate how tightly the words of a particular query are bound to each other.
pwi(A0 . . . Ak)=log2(prob(A0,A1, . . . ,Ak)/(pr(A0)*pr(a1)* . . . *pr(Ak)) Equation 3:
This pointwise mutual information can be used to rank the queries such that the words in the first query are the most highly bound, etc. The ranking of these queries can then be modified by some attribute such as time, monetization, etc.
Query prediction would be less effective if a moderately popular query in both the small and large databases were not used to scale. If a query was popular only in the large database and but not in the small database, then the scaling factor would be skewed. For example, if the query “floppy disk” were used as the scaling factor and it was queried many times in the history of the large database but was queried only a few times in the previous week, for the simple reason that no one produces or uses floppy disks anymore, then the ratio between the large and small databases would be enormous. This would skew the results of a partial query by heavily weighting relatively recent queries to the detriment of relatively older, and potentially more valuable, queries.
A similar problem would exist if a new query were used as the scaling factor that was only issued in the past week but rarely issued in the history of the large database. For example, “nuclear energy” may be an infrequently issued query in the past. But, because of a hypothetically recent announcement by Congress that 100 nuclear reactors will be constructed, the query “nuclear energy” will likely be issued thousands, if not hundreds of thousands of times. In that case, the scaling factor would be quite small; and when a query in the small database is weighted against the queries in the large database, then relatively older predicted queries, rather than relatively newer, and potentially more valuable, predicted queries, would most likely be returned to the user.
Therefore, referring to
At this point, a weight is applied to the potential queries in the small database versus the potential queries in the large database. This is performed by multiplying the result of the scaled small database number by ⅔ and adding it to the result of multiplying the number of times the potential query was found in the large database by ⅓ (see steps 708-712). Applying the weight, in effect, applies a bias in favor of the more recent search query. This is a type of temporal bias as it is based on when the queries were submitted. Steps 702-712 are performed for each potential query determined by the query predictor. When there are no more potential queries to process (714), all the potential queries are then compared with each other (step 716) based on their respective values determined for each potential query at step 712. The two or more queries (e.g., ten) with the highest values become the predicted queries, which are subsequently sent to the user.
Many variations of the process in
Fb=a*Sf*Fs+(1−a)Fl Equation 1:
In the equation 1, “Sf” is a scaling factor, “Fs” is the frequency of the query in the small database, “Fl” is the frequency of he query in the large database, “a” is a blending value. In the example flow of
Other variations of the flow depicted in
Fb=G(Fs, Ds, Fl, Dl, Ud) Equation 2:
In equation 2, “Fs” is the frequency of the query in the small database, “Fl” is the frequency of he query in the large database. “Ds” is more data from the small database for the current query. As examples, Ds is a temporal attribute, a monetization attribute, or both. “Dl” is more data from the large database for the current query. As examples, Dl is a temporal attribute, a monetization attribute, or both. “Ud” is some data that describes the user. As example, Ud may describe a user interest such as shopping, or sports, or news, etc. Ud might also describe the user's location, language preference, profession, etc. “G” is a function that maps the given parameters to the biased frequency. Often biasing that does not involve Ud can be done at database build time resulting in a database DBp that is directly derived from a database DB but with the biasing now contained in the value stored in the DB.
In one embodiment of the invention, the search engine component processes the particular predicted query (i.e. the most likely intended predicted query) that a user would be interested in. The particular predicted query is processed to obtain search results. The search engine that may be used for this purpose is common in the art and requires no further description.
The search results obtained by the search engine are sent to and displayed on the user's computer. If the particular predicted query is the user's intended query, the search results based on the particular predicted query may appear on the user's monitor even before the user enters another character in the query field and very likely before the user finishes entering the full intended query. If the particular predicted query is not the user's intended query, then the user may select a different predicted query in the list or continue typing, at which time a new set of search results, based on the selected or new particular predicted query, will be displayed via the user interface.
A user at a client 800 enters a partial query in a web browser 802. The partial query 812 is sent to a front end server 804 over a network 850. Front end server 804 is not a necessary element in any embodiment of the invention. Its main purpose is to add security to the interactive search engine system. Network 850 is also not a required element in any embodiment, but is merely illustrated to show one approach in which the invention may be implemented. Network 850 may be a local area network (LAN), a wide area network (WAN), or the Internet. Front end server 804 forwards partial query 812 to a query predictor 806, discussed above, which processes the partial query.
Front end server 804, query predictor 806, and a search engine 808, or any combination thereof, may be implemented on the same device. However, for the purpose of illustration and simplification, they each reside on different devices.
Query predictor 806 determines a set of one or more predicted queries based on the partial query and sends them 814 back to front end server 804. Along with the set of predicted queries, query predictor 306 sends additional data indicating which of the predicted queries in the set is the particular predicted query. Either query predictor 806 determines which predicted query is the particular predicted query or web browser 802 is given sufficient information to make that determination. Front end server 804 then forwards the predicted queries 814 and the data indicating the particular predicted query to client 800 over network 850 to be displayed on web browser 802.
Upon receipt of the set of predicted queries, web browser 802 sends particular predicted query 816 over network 850 to front end server 804, which forwards particular predicted query 816 to search engine 808. Search engine 808, described above, processes the particular predicted query to obtain search results. The search results 818 are finally sent to front end server 804, which forwards them 818 to client 800 over network 850.
One advantage of this implementation is that the predicted queries are sent immediately to the user as soon as they are determined. However, this implementation also illustrates the possibility that for every character the user types into the query field of his web browser, there are two complete round trips that a communication has to make between client 800 and front end server 804.
A user at client 800 enters a partial query in a web browser 802. Partial query 812 is sent to front end server 804 over a network 850. Front end server 804 forwards partial query 812 to query predictor 806, which processes the partial query.
Query predictor 806 determines a set of one or more predicted queries based upon the partial query and sends them 814 to front end server 804. Instead of immediately forwarding the predicted queries to client 800, front end server 804 retains the predicted queries and sends search engine 808 the particular predicted query 816. Again, along with the set of predicted queries, query predictor 806 sends additional data indicating which of the predicted queries in the set is the particular predicted query. Either query predictor 806 determines which predicted query is the particular predicted query or front end server 804 is given sufficient information to make that determination.
Search engine 808 processes the particular predicted query to obtain search results. The search results 818 are sent to front end server 804, at which time front end server 804 forwards both predicted queries 814 and search results 818 to client 800 over network 850.
In the absence of front end server 804, query predictor 806 sends the predicted queries 814 and to search engine 808, which subsequently sends the predicted queries 814 and search results 818 to client 800 over network 850.
One advantage of this implementation is that there is less communication (i.e., traffic) between client 800 and front end server 804. However, the predicted queries may not display on the user's web browser 802 as quickly as in the previous embodiment because the predicted queries must “wait” for the search results to be produced and sent to front end server 804 before the predicted queries are forwarded to client 800.
A user at client 800 enters a partial query in a web browser 802. Partial query 812 is sent to front end server 804 over network 850. Front end server 804 forwards the partial query 812 to query predictor 806, which processes the partial query.
Query predictor 806 determines a set of one or more predicted queries based upon the partial query and sends them 814 to front end server 804. Again, along with the set of predicted queries, query predictor 806 sends additional data indicating which of the predicted queries in the set is the particular predicted query. Either query predictor 806 determines which predicted query is the particular predicted query or front end server 804 is given sufficient information to make that determination.
Instead of “holding on” to the predicted queries, as in the last embodiment, front end server 804 sends the predicted queries 814 to client 800 over network 850 and sends particular predicted query 816 to search engine 808 at substantially the same time. It is also possible for query predictor 806 to send the particular predicted query to search engine 808 directly.
Search engine 808 processes the particular predicted query to obtain search results. The search results 818 are sent to front end server 804, at which time front end server 804 forwards search results 818 to client 800 over network 850. In the absence of front end server 804, query predictor 806 sends both the predicted queries 814 and the particular predicted query 816 to search engine 808, after which search engine 808 sends predicted queries 814 and search results 818 to client 800 over network 850.
In the absence of front end server 804, query predictor 806 sends both predicted queries 814 and the particular predicted query 816 to search engine 808, which subsequently sends predicted queries 814 and search results 818 to client 800 over network 850.
The advantage of this implementation compared to the embodiment described in
In one embodiment of the invention, as illustrated in
In one embodiment, the set of predicted queries are listed, beginning at the top, in order of the most likely predicted query to the least likely predicted query.
In addition to the user interface, query predictor, and search engine described above, the interactive search engine may be modified in many ways to alter the look, feel, and responsiveness of the search experience.
Tabs
For instance, the user interface includes tabs, such buttons or links 622 in
Keywords
Often when a query is issued, the order of words in the query is unimportant. As alluded to earlier, the issued query does not have to be in English. In other embodiments, not only are other natural languages supported, but also non-natural strings, such as product names and part numbers where the user may only know a portion of the non-natural string. Therefore, the term “word” as used herein may include an English word, a Korean word, or a product number.
When a user enters two or more words in the query field, the user is not necessarily concerned that the search engine returns a link to a web accessible document that contains the two or more words in the order that they were entered. Rather, the user is interested in a web accessible document that merely contains those words, in whatever order they are found.
For example, a user enters “solar wind water power” in the query field. The user does not particularly care about the order. The user is rather interested in queries that contain the words “solar,” “wind,” “water,” and “power” somewhere in the query. The query predictor determines what words are important in the query and which words are not important, and then predicts queries based on the important words instead of predicting queries based simply on a matching substring.
Delay Results
In another embodiment, the step of displaying the predicted queries and/or the search results is delayed. Instead of immediately returning predicted queries, the query predictor “waits” until certain criteria is satisfied (such as the lapse of a specified amount of time or when a few characters are entered, or both) before the predicted queries and search results are displayed. This additional step of waiting assumes that the user may not be sure what he/she wants to query on. Thus, the predicted queries are delayed until the interactive user interface determines, based on the waiting criteria, that this is what the user truly intends to query on. Once the waiting criteria are satisfied, the partial query is processed by the query predictor and the search engine, as described above.
Other Dynamic Data
There are additional ways to aid users other than to predict the intended query and return the appropriate search results. In another embodiment, advertisements that appear on the interactive user interface change based on the particular predicted query returned from the query predictor. Thus, every time the particular predicted query changes, new advertisements that relate to the query are posted on the user interface and advertisements that related to an older and non-relevant query are deleted from the user interface. For instance, if a user types “elli” and the query predictor determines “elliptical” as the particular predicted query, advertisements that relate to exercise equipment will appear on the user interface.
In addition to advertisements, other dynamic information may be useful to the user when submitting a query. In another embodiment, information relating to a particular predicted query, but not found in the search results, are displayed to the user via the user interface. Extending the “theory” example used above, the query predictor determines that “theory” is the particular predicted query for the partial query “th” entered by the user. The query predictor, or perhaps another program, determines that “theory” is associated with “string theory,” “music theory,” and “math theory” and returns these related subjects to be displayed in the form of predicted queries or in a different form on the user interface. For short queries like “theory,” this additional information happens to be the same set as what the query predictor would produce.
However, if the user entered “interna” in the query field and the query predictor determined that the particular predicted query is “international trade” then the query predictor, in addition to the predicted queries, would return queries that are not lexical completions of “international trade,” but rather queries related to the topic of international trade. Such queries could be on GATT, WTO, UN, US trade policies, etc. A program separate from the query predictor could also perform this function.
Clearly, this aspect of the invention is not performing query prediction, but rather is providing the user with dynamic, related, and hopefully helpful information. A principle in providing advertisements, additional queries, and other related information is to keep everything that is displayed via the user interface consistent with what the query predictor “believes” is the user's intent, which the query predictor determines from the partial query.
Techniques have been described for predicting what a user search query will be when completed. The search query thus predicted is referred to herein as the “predicted search query”. According to one embodiment, after determining the predicted search query, the search engine goes on to determine and provide to the user one or more alternatives to the predicted search query. Such alternative queries are referred to herein as “alternative search queries”. Each of the alternative search queries is based, at least in part, on the predicted search query, but differs in some manner from the predicted search query.
The set of “candidate suggested queries” for any given query may include both predicted search queries and alternative search queries. Thus, the candidate queries that are selected for display to the user (the “suggested queries”) may include both predicted search queries and alternative search queries.
Unlike the predicted search query, the alternative search queries are not predictions of what the user query will look like when completed. In fact, alternative search queries are often something that the partially-formed user search query is unlikely to complete to. For example, in response to the user entering a search query of “brittany sp”, the predicted search query may be “brittany spears”, but the alternative search query may be “britney spears”. In this example, it is highly unlikely that the completed version of “brittany sp” would be “britney spears”, since “britney spears” could only be produced if the user backtracks and changes the spelling.
In accordance with an embodiment of the present invention, the alternative search queries are determined as follows: A search query is received from a client node. Prior to receiving an indication from the client node that the search query is completely formed, the following steps are performed: 1) a predicted search query is determined by predicting what the search query will be when completed; and 2) an alternative search query that differs from the predicted search query is determined based on the predicted search query. The alternative search query is provided to the client node.
The alternative search query may be based on an alternative spelling (or spelling suggestion) of the predicted search query. The alternative search query may be a closely related search query to the predicted search query. For example, if the predicted search query is an acronym, the alternative may be an expansion of the acronym. Another example of closely related search queries is “movie times” and “show times”.
Biasing to Suggest Search Queries: Overview
As discussed herein, there are numerous ways that a search engine can suggest a search query to a user. As used herein, the term “suggested search query” means any search query that is suggested to a user. For example, the previously discussed, predicted search queries and the alternative search queries are suggested search queries. Further examples of suggested search queries include, but are not limited to, suggested alternative spellings and “also try” suggestions. An also try suggestion is based on search queries that are closely related with one another. An example of an also try query is suggesting the related search query “movie times” to a user search query of “show times”. Queries can be suggested not only in a network environment, but also by an electronic device. For example, cellular telephones, personal digital assistants, and television systems can suggest queries to users via a user interface.
Techniques are disclosed herein for applying a bias when determining a suggested search query. Examples of biases that can be applied include, but are not limited to, temporal biases and monetization biases. Temporal biasing involves increasing a weight associated with a search query, based on a temporal attribute associated with the query. An example of a temporal attribute is when the query was submitted. Search queries may also have associated with them a parameter such as frequency, count, etc. One of these parameters may serve as a weight. For example, as discussed herein, a determination of a predicted query completion can be based on a frequency parameter. In order to determine suggested search queries, the temporal attribute is used to modify or bias the parameter (e.g., frequency parameter), in one embodiment. Thus, the weight of the search query is modified to temporally bias the query. The suggested search queries are determined based on the biased parameter, in one embodiment. In another embodiment, the suggested search queries are determined based on the temporal attribute associated with each search query without biasing a parameter such as frequency.
Monetization biasing involves increasing a weight associated with a search query based on monetization information associated with the query. For example, a search query is assigned a monetization attribute based on its association with revenue generation. Suggested search queries are determined based on monetization biasing, in one embodiment.
On-the-Fly Biasing
In one embodiment, one or more suggested queries are provided to a client node, based on “on-the-fly” biasing of search queries. As used herein, “on-the-fly” biasing means that determining a suggested search query for a particular user search query involves performing a biasing operation after receiving the particular user search query. A method of on-the-fly biasing involves the following steps, in accordance with an embodiment of the present invention.
A user search query is received from a client node. A set of un-biased suggested query candidates is generated. For example, a database of submitted search queries is maintained, with each search query having a frequency parameter associated therewith. A list of one or more search queries from this database is generated, based on the frequency parameter. The list may also be based on the user search query. Thus, a set of un-biased suggested query candidates is generated.
A set of biased suggested query candidates is generated by biasing a parameter that is associated with the un-biased suggested query candidates. The parameter may be a frequency, a count, etc. that is associated with the un-biased suggested query candidate. The frequency parameter associated with each query is biased, as described below, to generate a set of biased suggested query candidates.
The biasing of the parameter is based on an attribute associated with each un-biased suggested query candidate. For example, each suggested query candidate has a temporal attribute associated therewith that reflects when the suggested query candidate was previously received. As another example, the biasing is based on a monetization attribute that reflects a monetary value that is associated with the un-biased suggested query candidate. Based on the biased parameter of the biased suggested query candidates, one or more of the suggested query candidates are selected to be suggested queries for the user search query. For example, the biased suggested query candidates with the highest value for the biased parameter are selected to be suggested queries to forward to the client node. At least one of the suggested queries is provided to the client node.
Pre-conditioning of Data
In one embodiment, one or more suggested queries are provided to a client node, based on “pre-conditioned” biasing of a parameter associated with search queries, by performing the following steps. As used herein “pre-conditioned” biasing means that determining a suggested search query for a particular user search query involves performing a biasing operation prior to receiving the particular user search query. A method of pre-conditioned biasing involves the following steps, in accordance with an embodiment of the present invention.
Prior to receiving a user query, a biased parameter is generated for each of a plurality of suggested query candidates. The parameter may be a frequency, a count, etc. that is associated with search queries. For example, a database of submitted search queries is maintained, with each search query having a frequency parameter associated therewith. The frequency parameter associated with each query is biased, as described below, to generate a biased frequency parameter. The search queries in this database serve as a pool of candidates for suggested search queries. Therefore, a set of suggested query candidates is generated, with each candidate having a biased parameter associated therewith.
The biasing of the parameter is based on an attribute associated with each suggested query candidate. For example, each suggested query candidate has a temporal attribute associated therewith that reflects when the suggested query candidate was previously received. As another example, the biasing is based on a monetization attribute that reflects a monetary value that is associated with the suggested query candidate.
After the biasing has been completed, the user search query is received from a client node. Based, at least in part, on the biased parameters associated with the suggested query candidates, one or more of the suggested query candidates are selected to be suggested queries for the user search query. For example, suggested query candidates with the highest value of the biased parameter, given the user search query, are selected to be suggested queries for the user search query. At least one of the suggested queries is provided to the client node.
Selecting a Suggested Query Based on Temporal or Monetization Attributes
In one embodiment, suggested queries are determined based on either temporal attributes or monetization attributes. The determination of the suggested search query is determined on the temporal attribute or the monetization attribute without biasing a parameter, in this embodiment. For example, the temporal attribute is not used to bias a frequency parameter, in this embodiment. A method of suggesting queries determined based on either temporal attributes or monetization attributes involves the following steps, in accordance with an embodiment of the present invention.
A user search query is received from a client node. A set of suggested query candidates are determined. For example, a list of search queries is generated based on a frequency parameter associated with each search query. For example, search queries with the highest value for the frequency parameter are selected, given the user search query. The list of search queries may also be generated based on either the temporal attribute or the monetization attribute.
Based on a set of one or more factors, one or more of the suggested query candidates is selected to be suggested queries for the user search query. The set of one or more factors includes at least one of (a) temporal information associated with each of the suggested query candidates and (b) monetization information associated with each of the suggested query candidates. At least one of the suggested queries is provided to the client node. For example, queries that have a higher value for the monetization information are selected.
The suggested search query 904 “when is mother's day” is a prediction of what the user search query 902 will be when completed. This prediction is based on a temporal bias. For example, if the present time is near mother's day, then the likelihood that the user will complete the user query 902 to “when is mother's day” is much higher than it would be at a different time of the year. However, at a different time of the year, the same user query 902 might be predicted to complete to “when is father's day”. A temporal bias is applied to stored search queries in order to make the prediction, in accordance with an embodiment of the present invention. For example, the frequency with which the search query “when is mother's day” was submitted in the last week is adjusted upwards to provide a temporal bias in favor of this search query.
Additional suggested search queries are presented below “when is mother's day”. Not all of the suggested search queries are on the list because of the temporal bias. For example, the search query “when is father's day” may be in both the previously described older database, which does not receive a temporal bias, and the previously described recent database, which does receive a temporal bias. The query “when is father's day” might appear in its location on the list of suggested queries based on its frequency in the older database. Thus, hypothetically, “when is father's day” did not make the list due to a temporal bias of that query.
The user interface 900 has search results 906 that contain links to documents (e.g., web pages) that satisfy one or more of the search queries. In this example, the search results 906 pertain to the predicted search query “when is mother's day”, which has the highest relevancy ranking of the suggested search queries 904. These search results are provided to the user by default without the user requesting search results for that particular search query, in one embodiment. However, search results can be provided for any of the search queries. Note that there is no button for the user to submit a search query, in this example user interface 900. The user can change which search results are displayed by scrolling through the list of search queries (including the mirrored search query 903), in one embodiment. The user interface 900 has other areas for displaying results related to the search queries, such as sponsor provided areas 910.
Process Flow for Applying a Bias to Determine a Suggested Query
As an example, the suggested search query “when is mother's day” appears in both the recent database and the older database. This search query has a parameter associated with it such as frequency, count, etc. This parameter is biased by an attribute based on time, monetization, etc. As a particular example, a frequency parameter in the recent database is adjusted to provide a temporal bias. However, the frequency parameter in the older database is not adjusted.
More generally, a bias is applied to a parameter that is associated with search queries stored in a database. As previously discussed, the parameter that is biased might be frequency, count, etc. As previously discussed, the bias might be temporal, monetary, etc. If optional step 1002 is performed, the data are said to be “pre-conditioned”. However, the bias can be applied after a user search query is received. In this case, the biasing is said to be performed “on-the-fly”.
In step 1004, a search query is received from a client node. For example, the example user interface 900 of
In step 1006, a set of suggested query candidates is determined. For example, if the data are pre-conditioned the following might hypothetically occur. The query “when is mother's day” is extracted from the recent database and added to the set of suggested query candidates based on a temporally biased frequency parameter. The search query “when is father's day” might be extracted from the older database based on the value of its un-biased frequency parameter. The search query “when is mother's day” might also be extracted from the older database based on its un-biased frequency parameter.
If the data are not pre-conditioned, the following might hypothetically occur, in step 1006. The search queries can still be stored based on different time periods. For example, there may still be a recent database and an older database, although the frequency parameter in these databases is un-biased. The search query “when is father's day” might be added to the list of suggested query candidates based on the value of its un-biased frequency parameter in either or both databases. The search query “when is mother's day” might also be added to the list based on its un-biased frequency parameter in either or both databases.
In optional step 1008, a bias is applied to a parameter for one or more suggested query candidates. This step is referred to as performing biasing “on-the-fly”. As previously discussed, the parameter that is biased might be frequency, count, etc. As previously discussed, the bias might be temporal, monetary, etc. The on-the-fly biasing can be performed with or without pre-conditioning the data.
For example, consider the example in which the data were not pre-conditioned, and “when is father's day” and “when is mother's day” are on the list of suggested query candidates based on their un-biased frequency parameters. The frequency parameter associated with the suggested query candidates from the recent database is adjusted to temporally bias those suggested query candidates. For example, the frequency parameter associated with “when is mother's day” from the recent database is increased to create a temporal bias for that recent search query.
With on-the-fly biasing, the bias may also be based on a user characteristic. For example, if it is known that a user is interested in shopping, then the monetization attribute may be increased. If the user is interested in news, then the temporal attribute can be increased. For example, a database stores suggested query candidates, with each query candidate having a monetization attribute, in one embodiment. If it is determined that a user is interested in shopping, then the monetization attribute associated with all queries can be increased to favor those queries even more.
In step 1010, one or more suggested query candidates are selected to be suggested queries to provide to the client node. This selection is based on the biased parameter. For example, the suggested queries 904 depicted in
In one embodiment, suggested queries are ranked in accordance with the relative weight of the suggested queries. For example, the weight is based on a frequency parameter as modified by a temporal attribute or a monitization attribute. The suggested query candidates are then sent to the client node 800 and displayed in the user interface 900, as shown in
In optional step 1012, search results that satisfy one or more of the suggested queries are determined and provided to the client 800. Results that satisfy one or more of the suggested queries may be displayed as search results 906 in other regions of the user interface 900, such as a sponsor provided area 910.
Database Operation
Biasing a parameter such as frequency, count, etc. that is associated with a search query with an attribute such as a temporal attribute or a monitzation attribute allows for a single value to be associated with the query stored in a database. This allows for very fast searches for suggested queries by allowing search queries to be placed in a database by strata. For example, assuming that search queries a ranked by a biased frequency parameter, searches can be categorized by very high frequency, moderately high frequency, medium frequency, low frequency, very low frequency. When a search for a predicted completion to a user query is performed, for example, the search starts at the very high frequency strata and works down until a sufficient number of potential matches are found. Thus, the database can be searched very efficiently.
Hardware Overview
Computer system 1100 may be coupled via bus 1102 to a display 1112, such as a cathode ray tube (CRT), for displaying information to a computer user. An input device 1114, including alphanumeric and other keys, is coupled to bus 1102 for communicating information and command selections to processor 1104. Another type of user input device is cursor control 1116, such as a mouse, a trackball, or cursor direction keys for communicating direction information and command selections to processor 1104 and for controlling cursor movement on display 1112. This input device typically has two degrees of freedom in two axes, a first axis (e.g., x) and a second axis (e.g., y), that allows the device to specify positions in a plane.
Computer system 1100 may be used to implement the techniques described herein. According to one embodiment of the invention, those techniques are performed by computer system 1100 in response to processor 1104 executing one or more sequences of one or more instructions contained in main memory 1106. Such instructions may be read into main memory 1106 from another computer-readable medium, such as storage device 1110. Execution of the sequences of instructions contained in main memory 1106 causes processor 1104 to perform the process steps described herein. In alternative embodiments, hard-wired circuitry may be used in place of or in combination with software instructions to implement the invention. Thus, embodiments of the invention are not limited to any specific combination of hardware circuitry and software.
The term “computer-readable medium” as used herein refers to any medium that participates in providing instructions to processor 1104 for execution. Such a medium may take many forms, including but not limited to, non-volatile media, volatile media, and transmission media. Non-volatile media includes, for example, optical or magnetic disks, such as storage device 1110. Volatile media includes dynamic memory, such as main memory 1106. Transmission media includes coaxial cables, copper wire and fiber optics, including the wires that comprise bus 1102. Transmission media can also take the form of acoustic or light waves, such as those generated during radio-wave and infra-red data communications.
Common forms of computer-readable media include, for example, a floppy disk, a flexible disk, hard disk, magnetic tape, or any other magnetic medium, a CD-ROM, any other optical medium, punchcards, papertape, any other physical medium with patterns of holes, a RAM, a PROM, and EPROM, a FLASH-EPROM, any other memory chip or cartridge, a carrier wave as described hereinafter, or any other medium from which a computer can read. Various forms of computer readable media may be involved in carrying one or more sequences of one or more instructions to processor 1104 for execution. For example, the instructions may initially be carried on a magnetic disk of a remote computer. The remote computer can load the instructions into its dynamic memory and send the instructions over a telephone line using a modem. A modem local to computer system 1100 can receive the data on the telephone line and use an infra-red transmitter to convert the data to an infra-red signal. An infra-red detector can receive the data carried in the infra-red signal and appropriate circuitry can place the data on bus 1102. Bus 1102 carries the data to main memory 1106, from which processor 1104 retrieves and executes the instructions. The instructions received by main memory 1106 may optionally be stored on storage device 1110 either before or after execution by processor 1104.
Computer system 1100 also includes a communication interface 1118 coupled to bus 1102. Communication interface 1118 provides a two-way data communication coupling to a network link 1120 that is connected to a local network 1122. For example, communication interface 1118 may be an integrated services digital network (ISDN) card or a modem to provide a data communication connection to a corresponding type of telephone line. As another example, communication interface 1118 may be a local area network (LAN) card to provide a data communication connection to a compatible LAN. Wireless links may also be implemented. In any such implementation, communication interface 1118 sends and receives electrical, electromagnetic or optical signals that carry digital data streams representing various types of information.
Network link 1120 typically provides data communication through one or more networks to other data devices. For example, network link 1120 may provide a connection through local network 1122 to a host computer 1124 or to data equipment operated by an Internet Service Provider (ISP) 1126. ISP 1126 in turn provides data communication services through the world wide packet data communication network now commonly referred to as the “Internet” 1128. Local network 1122 and Internet 1128 both use electrical, electromagnetic or optical signals that carry digital data streams. The signals through the various networks and the signals on network link 1120 and through communication interface 1118, which carry the digital data to and from computer system 1100, are exemplary forms of carrier waves transporting the information.
Computer system 1100 can send messages and receive data, including program code, through the network(s), network link 1120 and communication interface 1118. In the Internet example, a server 1130 might transmit a requested code for an application program through Internet 1128, ISP 1126, local network 1122 and communication interface 1118.
The received code may be executed by processor 1104 as it is received, and/or stored in storage device 1110, or other non-volatile storage for later execution. In this manner, computer system 1100 may obtain application code in the form of a carrier wave.
In the foregoing specification, embodiments of the invention have been described with reference to numerous specific details that may vary from implementation to implementation. Thus, the sole and exclusive indicator of what is the invention, and is intended by the applicants to be the invention, is the set of claims that issue from this application, in the specific form in which such claims issue, including any subsequent correction. Any definitions set forth herein for terms contained in such claims shall govern the meaning of such terms as used in the claims. Hence, no limitation, element, property, feature, advantage or attribute that is not expressly recited in a claim should limit the scope of such claim in any way. The specification and drawings are, accordingly, to be regarded in an illustrative rather than a restrictive sense.
| 标题 | 发布/更新时间 | 阅读量 |
|---|---|---|
| 将实际对象通用实时链接至机器、网络、互联网或软件服务的综合系统和方法 | 2020-05-13 | 808 |
| 用于货币化匿名化的上下文的技术 | 2020-05-17 | 145 |
| 软件开发套件平台 | 2020-05-08 | 442 |
| 一种生态工程建设效益评估方法 | 2020-05-15 | 858 |
| 用于跟踪容器的方法和系统 | 2020-05-14 | 346 |
| 基于弹性证券化的股息收益货币的方法、装置和计算机可读介质 | 2020-05-13 | 608 |
| 用于活动资金的出资公开募集系统 | 2020-05-11 | 650 |
| 一种供热系统的阀门控制电路及利用该电路的控制系统 | 2020-05-15 | 947 |
| SYSTEM AND METHOD FOR HYBRID SAAS VIDEO EDITING | 2020-05-16 | 167 |
| AUTOMATIC WIRELESS GRID ACCESS SHARING AND MONETIZATION FROM A MOBILE DEVICE APPLICATION THAT RECEIVES CREDENTIALS OVER A WIRELESS CHANNEL | 2020-05-15 | 454 |
高效检索全球专利专利汇是专利免费检索,专利查询,专利分析-国家发明专利查询检索分析平台，是提供专利分析，专利查询，专利检索等数据服务功能的知识产权数据服务商。
我们的产品包含105个国家的1.26亿组数据，免费查、免费专利分析。
专利汇分析报告产品可以对行业情报数据进行梳理分析，涉及维度包括行业专利基本状况分析、地域分析、技术分析、发明人分析、申请人分析、专利权人分析、失效分析、核心专利分析、法律分析、研发重点分析、企业专利处境分析、技术处境分析、专利寿命分析、企业定位分析、引证分析等超过60个分析角度，系统通过AI智能系统对图表进行解读，只需1分钟，一键生成行业专利分析报告。
