CN105808590A - Search engine realization method as well as search method and apparatus - Google Patents

Abstract

The invention discloses a search engine realization method as well as a search method and apparatus. The method comprises the steps of obtaining search expression histories input by a user in a continuous search process and selected target results; according to the search expression histories, determining an expression keyword sequence; and according to the expression keyword sequence and the target results, generating a semantic net, wherein the semantic net comprises connection relationships and weights between expression keywords and between the expression keywords and the target results. According to the search engine realization method as well as the search method and apparatus provided by the invention, the statistics is made based on the expression keywords input by the user and the selected target results to establish the semantic net, the distances and the weight relationships between the keywords and between the keywords and the target results are obtained, and when the user searches for the target results, services can be provided for the user based on the semantic net, so that the service information search time of the user is shortened and more accurate and efficient information services are provided for the user.

Description

Search engine realizes method, searching method and device

Technical field

The present invention relates to internet arena, particularly relate to a kind of search engine and realize method, searching method and device.

Background technology

Search engine technique is as what Internet technology development got up, and Main Function is when information explosion, the retrieval of reply magnanimity information, meets people and obtains the demand of information.

One core aspect of search engine technique is to adopt the algorithm of all kinds of advanced persons to calculate user to input the dependency of information in key word or problem and backstage mass data index, the algorithm taked can evaluate the relation of numerous information and key word from all angles, and evaluation result is exactly the list according to sizes related of the information.After user obtains this list, it is necessary to oneself actively screens suitable Search Results.User is in the process of filtered list, progressively understand the using method of search engine, understand the skill of key word input, the input that can adjust key word carrys out expressing information demand again, take part in search procedure from this angle people, the feature of the information service systems such as search engine, service ability are sightless for user.But prior art needs to rely on user the judgement of result comes the using method of learning system and adjusts the key word of input, and to reach search purpose, search data means are relatively not intelligent, and user obtains time length, search experience is low, compare waste user time.

Therefore, it is necessary to propose a kind of search engine to realize method, user search information time can be saved, provide the user the information service of more precise and high efficiency.

Summary of the invention

To solve one of the disclosure technical problem is how to provide a kind of search engine to realize to save user search information time, provides the user the information service of more precise and high efficiency.

The disclosure provides a kind of search engine to realize method, including:

Obtain the search expression history of input in one continuous search procedure of user and the objective result of selection；Expression keyword sequence is determined according to search expression history；According to expressing keyword sequence and objective result generative semantics net for search engine, semantic net includes expressing between key word and expresses the annexation between key word and objective result and weight.

Alternatively, it is thus achieved that the search expression history of input and the objective result of selection in one continuous search procedure of user, including: obtain user's input express key word, adjustment express key word and the objective result selected according to Search Results.

Alternatively, also comprising determining that service field according to before expressing keyword sequence and objective result generative semantics net, service field includes sole user, group, vertical field or integrated information service field；Judge that expressing of the input of user under service field is expressed key word and be whether complete information service process according to the process expressing Keyword Selection objective result in keyword sequence.

Alternatively, semantic net includes expressing between key word and expresses the annexation between key word and objective result and weight, including:

If the expression key word of user's input is unsatisfactory for semantic web updates rule, then cast out described expression key word；

If the expression key word of user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are less than similarity threshold, the expression key word of user's input is added as the new keywords in semantic net in semantic net, and record the objective result that user selects, generate or the multi-stress of the annexation weight that updates in described new keywords and semantic net between other key words and described new keywords and objective result；

If the expression key word of user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are be more than or equal to similarity threshold, generate or the multi-stress of key word and objective result in update semantics net, multi-stress is the shortest path factor according to the key word in semantic net to objective result and weight factor is determined, weight factor is arrive the number of times normalization of objective result according to key word in the semantic net of historical statistics to determine.

Alternatively, judge that expressing of the input of user under service field is expressed key word and be whether complete information service process according to the process expressing Keyword Selection objective result in keyword sequence, comprise determining that user input reaches the process of the key word search targets result number of clicks to each objective result；If the number of clicks of objective result is more than information service threshold value, then service process is defined as complete service process；

And/or

The process of the objective result selected after user input reaches key word is defined as complete service process.

The disclosure additionally provides a kind of searching method, comprises determining that user input reaches the expression key word in formula；Determining and express the start node that key word mates in semantic net, wherein, semantic net generates based on the expression key word in user search history and objective result；Determining objective result according to start node in semantic net, objective result is determined according to distance and the weight of start node to objective result；Prompting objective result.

Alternatively, the method also includes: determine prompting key word according to start node in semantic net；Prompting prompting key word.

Alternatively, in semantic net, determine that prompting key word includes according to start node:

If start node is not the endpoint node of semantic net, start node is selected to develop to the path of objective result, it is determined that the node on path is as prompting key word.

Alternatively, if start node is the endpoint node of semantic net, objective result corresponding for start node is prompted to user；If start node is not the endpoint node of semantic net, select the weight factor objective result more than weight threshold of objective result；Select the path factor that start node develops to objective result more than the path of path threshold；Determine that other nodes on path are as prompting key word；Prompting prompting key word and/or multiple objective result；Wherein, the node in semantic net is the set of same or analogous key word composition, and weight factor is determine according to the number of times normalization of key word arrival objective result in the semantic net of historical statistics.

Disclosure one search engine realizes device, including: acquisition module, the objective result of search expression history and selection for obtaining input in one continuous search procedure of user；Determine module, for determining expression keyword sequence according to search expression history；Generation module, for according to expressing keyword sequence and objective result generative semantics net for search engine, semantic net includes expressing between key word and expresses the annexation between key word and objective result and weight.

Alternatively, expression key word, the expression key word of adjustment and the objective result according to Search Results selection that acquisition module inputs for obtaining user.

Alternatively, this device also includes:

Module is determined in service field, is used for determining that service field, service field include sole user, group, enterprise or industry；Information service process judge module, for judging that expressing of the input of user under service field is expressed key word and be whether complete information service process according to the process expressing Keyword Selection objective result in keyword sequence.

Alternatively, evaluation module, if the expression key word for user's input is unsatisfactory for semantic web updates rule, then cast out described expression key word；

If the expression key word that evaluation module is additionally operable to user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are less than similarity threshold, the expression key word of user's input is added as the new keywords in semantic net in semantic net, and record the objective result that user selects, generate or the multi-stress of the annexation weight that updates in described new keywords and semantic net between other key words and described new keywords and objective result；

If the expression key word that evaluation module is additionally operable to user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are be more than or equal to similarity threshold, generate or the multi-stress of key word and objective result in update semantics net, multi-stress is the shortest path factor according to the key word in semantic net to objective result and weight factor is determined, weight factor is arrive the number of times normalization of objective result according to key word in the semantic net of historical statistics to determine.

Alternatively, information service process judge module is used for: determine that user input reaches the process of the key word search targets result number of clicks to each objective result；If the number of clicks of objective result is more than information service threshold value, then service process is defined as complete service process；And/or the process of the objective result selected after user input reaches key word is defined as complete service process.

Disclosure one searcher, including: express key word analysis module, for determining that user input reaches the expression key word in formula；Start node analysis module, expresses, for determining, the start node that key word mates in semantic net, and wherein, semantic net generates based on the expression key word in user search history and objective result；Determining module, for determining objective result according to start node in semantic net, objective result is determined according to distance and the weight of start node to objective result；Reminding module, is used for pointing out objective result.

Optionally, it is determined that module is additionally operable in semantic net to determine prompting key word according to start node；Reminding module is additionally operable to prompting prompting key word.

Alternatively, reminding module is used for: if start node is not the endpoint node of semantic net, select start node to develop to the path of objective result, it is determined that the node on path is as prompting key word.

Alternatively, if reminding module is the endpoint node of semantic net for start node, objective result corresponding for start node is prompted to user；If and/or reminding module is for the endpoint node that start node is not semantic net, selects the weight factor objective result more than weight threshold of objective result；Select the path factor that start node develops to objective result more than the path of path threshold；Determine that other nodes on path are as prompting key word；Prompting prompting key word and/or multiple objective result, wherein, the node in semantic net is the set of same or analogous key word composition, and weight factor is determine according to the number of times normalization of key word arrival objective result in the semantic net of historical statistics.

The search engine that the disclosure provides realizes method, searching method and device, carry out statistics set up semantic net based on the expression key word of user's input and the objective result of selection, obtain the distance between key word and key word and between key word and objective result and weight relationship, when user search objective result, information on services can be provided the user based on this semantic net, save user search information time, provide the user the information service of more precise and high efficiency.

Accompanying drawing explanation

Fig. 1 illustrates that the search engine of one embodiment of the invention realizes the flow chart of method.

Fig. 2 illustrates that the continuous search procedure of one embodiment of the invention obtains the schematic diagram of objective result.

Fig. 3 illustrates that the search engine of another embodiment of the present invention realizes the flow chart of method.

Fig. 4 illustrates that the search engine of one embodiment of the invention realizes the flow chart of optimization method.

Fig. 5 illustrates the flow chart of the searching method of one embodiment of the invention.

Fig. 6 illustrates the flow chart of the searching method of another embodiment of the present invention.

Fig. 7 illustrates the flow chart of the searching method of another embodiment of the present invention.

Fig. 8 illustrates that a sphere of learning one of one embodiment of the invention teaches the schematic diagram of the semantic net of group.

Fig. 9 illustrates the structural representation of the search engine system of one embodiment of the invention.

Figure 10 illustrates that the search engine of one embodiment of the invention realizes the structured flowchart of device.

Figure 11 illustrates that the search engine of another embodiment of the present invention realizes the structured flowchart of device.

Figure 12 illustrates the structured flowchart of the searcher of one embodiment of the invention.

Figure 13 illustrates that the search engine of an alternative embodiment of the invention realizes the structured flowchart of device.And

Figure 14 illustrates the structured flowchart of the searcher of an alternative embodiment of the invention.

Detailed description of the invention

With reference to the accompanying drawings the present invention is described more fully, the exemplary embodiment of the present invention is wherein described.

Search engine is understood that, for customer problem, the core objective that search engine is pursued.Generally, in the service process of search engine, there are three core links.First core link is the understanding for problem, and search engine understands user by all kinds of algorithms and method asks to put question to the core point of (QUERY) it is known that what user looked for is what.Second core link is: how background information resource to be organized, is indexed, and makes background information can store with the mode adapting to information inquiry in the way of expressing information essence by a kind of；3rd core link is: search need and Search Results directly mate, and makes to mate between user's request with storage information, and carries out reasonable computation, and result of calculation is namely in the face of the demand of user profile, the information that reflection Essence of Information is expressed again.

Embodiment of the present invention basic thought is at information service fields such as search engines, introduce semantic network technology, propose a kind of can for individual, group, field application search engine realize method, the method formed be a kind of can perception field change and a certain period class problem expression change semantic network, and this semantic net can according to individual, group, field the application autonomous evolution of information service, ultimately form towards particular person, group, field understanding service-Engine.According to this semantic network, in user search process, the diversity that different user describes can be overcome for same thing difference, even if different user adopts different expression for same thing, it is also possible to obtain identical objective result or prompting key word.The method of the embodiment of the present invention be also equipped with according to time and information requirement, expression, generation change, understand the ability of information, adapt to the search of different times and express and search is understood and objective result recommends ability.

Fig. 1 illustrates that the search engine of one embodiment of the invention realizes the flow chart of method.As it is shown in figure 1, the method specifically includes that

Step S100, it is thus achieved that the search expression history inputted in one continuous search procedure of user and the objective result of selection；

In one embodiment, it is possible to the search procedure according to user, expression key word, the expression key word of adjustment and the objective result according to Search Results selection that acquisition user inputs.

For same problem, the identification of different expression waies, the search history record of user application, key word is expressed in the history input corresponding to Learning Service success, if servicing successfully, what record inputted continuously expresses key word and corresponding objective result.The record that search engine system inputs for the user search in certain time, analyzing the adjustment that in record, user search input is expressed, final goal result and the difference of this objective result corresponding that study user selects in several continuous input processes within a period of time express key word.

It should be noted that the search expression of the present invention not merely refers to input word, it will be understood by those in the art that user can input search expression in many ways, for instance, it is possible to it is the modes such as audiovideo.Such as, based in the search procedure of voice dialogue, for a problem, user's possibility can have polyglot expression way, if user have selected objective result according to the Search Results of system, generally can stop search and check objective result, it is possible to user voice be expressed and the objective result of selection gets off as log recording, in order to statistical analysis generative semantics net.

Step S102, determines expression keyword sequence according to search expression history；

In search routine, user need to while learning the knowledge in strange field, constantly adjust input to express simultaneously, required information can be obtained in a strange field, the objective result of key word and selection in record and study user search process, by learning different user, same problem there is is different expression, record the process of multiple user search, it is possible to extract the annexation between key word and key word.

Step S104, according to expressing keyword sequence and objective result generative semantics net for search engine, semantic net includes expressing between key word and expresses the annexation between key word and objective result and weight.

In one embodiment, it is possible to determine the multi-stress of this key word according to the crucial weight with objective result of distance and expression expressing key word and objective result.

The multi-stress that expression key word path and key word and objective result weight are determined decides how a start node obtains arriving the target information finally provided the user.On the one hand, annexation between analysis of key word and key word, obtain a key word and be connected to the shortest path of another key word, on the other hand, calculate key word linking objective result weighted value on certain paths, path weight value is to set up in semantic net process and in use procedure after foundation, expresses according to new user and the selection of objective result is continued to optimize adjustment and obtains.

In one embodiment, it is possible to record several users in several times search procedure, record user expresses key word according to one and adjusts another and express key word number of times, and using this number of times as connection weight.This connection weight refers to from a key word to next key word, the number of times that user selects, it is possible to be weighted or normalization and this weight of obtaining.

In one embodiment, this multi-stress is to determine according to path between the key word in semantic net and objective result and weight composite weighted.So can determine, according to this multi-stress, the relation expressing key word and objective result.In search processes, if user's input is this key word, then provide the user service according to this multi-stress.

In one embodiment, by learning multiple search procedures of multiple user, objective result based on the input key word of multiple users, adjustment key word and selection, it is possible to generate the relational network of keyword sequence and objective result, it is possible to this network is referred to as semantic net.When generating this semantic net, record each annexation expressing key word and according to the number of times expressing key word and obtaining objective result, and through historical statistics using this number of times normalization weight factor as this key word Yu objective result.

The semantic network of the embodiment of the present invention is according to expressing key word and the network of objective result composition, it forms one " key word and objective result network " according to the meaning of key word, and each node in this key word network all has annexation with adjacent node, key word is the basic component of this semantic net, this semantic net follows lexical semantics principle, the incompatible expression node of synset can be used, and with a number of relationship type, these key word nodes are associated, constitute a key word and objective result semantic net.This semantic net can be may apply in natural language understanding field, machine translation, input method field and searching engine field by this semantic net formed.

The search engine of the embodiment of the present invention realizes method, carry out statistics set up semantic net based on the expression key word of user's input and the objective result of selection, obtain the distance between key word and key word and between key word and objective result and weight relationship, in user search process, service is provided the user based on this semantic net, user search information time can be saved, provide the user the information service of more precise and high efficiency.

In one embodiment, user is before obtaining final result, and the information of input is probably change.In a continuous print information access process, there is the process that a result returned according to system learns how to express in user；User returns result by understanding, and also has a process of cognition to oneself demand.By recording user's input, input correspondence every time returns result and final result, after carrying out feature extraction, sets up a semantic network or semantic network data base.

Fig. 2 illustrates that the continuous search procedure of one embodiment of the invention obtains the schematic diagram of objective result, as shown in Figure 2, in input 1 to input n, user inputs different key word of expressing and wishes to search the good market buying general merchandise, first time inputs well-known general merchandise and shows multiple objective result, user non-selected any search display result, in second time input, user adjusts expression and inputs other several key words, multiple Search Results of display, with non-selected per family, in n-th inputs, user inputs famous general merchandise, result in search shows Yaohan's result, user clicks Shanghai the first Yaohan, enter relevant store information page and obtain the information that user needs.General merchandise, famous general merchandise, Yaohan and Shanghai the first Yaohan form a search history record, and wherein, general merchandise, famous general merchandise, Yaohan belong to the expression key word inputted continuously, and Shanghai the first Yaohan belongs to objective result.

With reference to Fig. 2, the expression key word that general merchandise, famous general merchandise, Yaohan input continuously as user is defined as keyword sequence.

Based on statistical result repeatedly, it is thus achieved that general merchandise is to the weight of famous general merchandise, and famous general merchandise is to the weight of Yaohan, and Yaohan is to the weight of Shanghai the first Yaohan.

Based on Fig. 2 semantic net formed, when using the search engine based on this semantic net or input method as a user, if search " general merchandise ", search engine can point out key word " famous general merchandise ", " Yaohan " according to this semantic net, after user clicks and determines search " general merchandise ", browser preferentially exports the search information that " Shanghai the first Yaohan " is relevant, for user.

The search engine of the embodiment of the present invention realizes method, it is possible to solving people for the diversity in problem understanding, description, the adaptability that individual variation brings standard product is not enough.Excavate based on the record in search engine interaction, understand service field knowledge and concept, need when solving user search request in many ways to adjust to obtain objective result problem.

Fig. 3 illustrates that the search engine of another embodiment of the present invention realizes the flow chart of method, as it is shown on figure 3, the method includes:

Step 300, it is thus achieved that the search expression history inputted in one continuous search procedure of user and the objective result of selection；

Step 301, determines expression formula key word according to search expression history；

Step S302, it is determined that service field, service field includes sole user, group etc.；

For example, at information service field, it is possible to for individual, group, vertical field or a comprehensive service field, information service system record user inputs key word and finally selects result.This service field can be conventional internet field, mobile Internet field, it is also possible to be voice information services platform.

Record user inputs the objective result information of key word and corresponding selection, so by the accumulation of long a period of time, just can obtain key word and report a corresponding relation net of enterprise, search result information or objective result.It is possible to further use record user in information service system process to input the change of key word user.

Step S303, it is judged that under service field, whether expressing of user's input is expressed key word in keyword sequence and be complete information service process according to the process expressing Keyword Selection objective result.

In one embodiment, judge that expressing of the input of user under service field is expressed key word and be whether complete information service process according to the process expressing Keyword Selection objective result in keyword sequence, including: within a time cycle, such as 1 day, 1 month, 1 year etc., it is determined that user input reaches the process of the key word search targets result number of clicks to each objective result；If the number of clicks of objective result is more than information service threshold value, then service process is defined as complete service process.

In one embodiment, it is possible to the process of the objective result selected after user input reaches key word is considered as complete service process.Can passing through additive method and judge that whether service process is complete service process, clearly expressing current Search Results such as user is exactly that it thinks objective result of acquisition；Or user stopped search after selecting certain objective result.

One complete information service process can be expressed as, and first user inputs key word, then observes search result list, if the user while number N, the N of the result clicked in list, each Search Results that record user clicks and click Search Results >=0.In the present invention, we definable N is the information service factor, in generic services process, when N is not zero, it is believed that this is a complete information service process；Under some specific environment or in a certain field, N can take an empirical value of system, thinks a complete information service process more than this empirical value, is a complete information service process less than not thinking of this empirical value.In a complete information service process, user converts key word, but the reason not producing Search Results has multiple.

For example, attend a banquet for voice information services platform the search procedure of service personal, in customer service system, forward direction user refers to make a phone call or access the user of means inquiry message by other, the user that attends a banquet refers to contact staff, its exchanging and oneself understanding according to user and user, input key word and some feature representations, scan for information in background service information.This conversion key word is likely to be produced by following reason:

1, forward direction user is for the diversity of information representation, the expression of forward direction user is its understanding required problem described according to self, and whether user's (attending a banquet) is correct from starting to be likely to and do not know about this understanding expression in service, whether store corresponding information on backstage, thus resulted in keyword search first and do not found satisfactory result.

2, user's (attending a banquet) passes through the understanding to system and have adjusted input key word alternately, and rely on the accumulation of service experience, take out the core point of (forward direction user) information requirement, but this central point exists certain change in service process.

So by long term accumulation, just can obtain key word and objective result such as reports a corresponding relation net of enterprise.The key word that last Ceng Shi end user (attending a banquet) of this net finally enters, the final goal information that user's (attending a banquet) that this key word is corresponding selects.Through long-term study and optimization, based on record repeatedly and adjustment, when this information network reaches after to a certain degree, it is possible to apply in reality system.Namely after user inputs those key words being likely not to have report result, for instance, according to conventional information search or search method, it is possible to after not reporting the key word of result, according to the report enterprise that the key word of semantic net acquisition finish node is corresponding with this key word.

In one embodiment, semantic net includes expressing between key word and expresses the annexation between key word and objective result and weight, including:

If the expression key word of user's input is unsatisfactory for semantic web updates rule, then cast out described expression key word；If the expression key word of user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are less than similarity threshold, the expression key word of user's input is added as the new keywords in semantic net in semantic net, and record the objective result that user selects, generate or the multi-stress of the annexation weight that updates in described new keywords and semantic net between other key words and described new keywords and objective result；If the expression key word of user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are be more than or equal to similarity threshold, generate or the multi-stress of key word and objective result in update semantics net, multi-stress is the shortest path factor according to the key word in semantic net to objective result and weight factor is determined, weight factor is determined in the number of times normalization arriving objective result according to key word in semantic net through historical statistics.

In one embodiment, weight factor is reach the number of times of objective result according to key word in semantic net, calculates through historical statistics, normalization and obtains.It is also conceivable to other factors and determine this weight factor, for instance consider time factor, in time period result often, might not be many in the next time period.

In one embodiment, semantic web updates rule can be newly-increased key word rule, reservation key word rule, delete key word rule, more new keywords multi-stress rule etc..

In one embodiment, if the key word that user is newly inputted, be unsatisfactory for semantic web updates rule, then system casts out this key word；Newly-increased node rule can be described below: 1, user is whether the input of a series of continuous print obtains objective result；2, the dependency of continuous input content judges；Or dispersion judges；Continuously whether input for same goal description, or whether input conceptually identical continuously, or continuously input on expressing, whether similar etc. 3, user's input can extract key word or the expression etc. that system can reuse；If the technology that user's input meets the newly-increased node of semantic net judges；Simultaneously, user input reaches key word with original key word similarity in semantic net less than similarity threshold, namely may determine that this is that a new user expresses, such as: key word, in semantic net, then add expressing if key word is as the new keywords in semantic net of user's input, and record the objective result that user selects.

In one embodiment, for individual, group, vertical field, area or a selected service orientation set information service factor.And according to this information service factor, selected information service process.

Record in an information service process, the Search Results that the key word of user's input and correspondence select in search result list.

In one embodiment, semantic web updates rule can be passed through key word screening and determine, relies on an evaluation function, it is judged that whether this word can be kept as the newly-increased node of semantic net.

In one embodiment, this evaluation function can take following principle:

1, in the service dictionary of service accumulation, it is judged that whether user input reaches whether key word is a complete word, or be a complete expression.Evaluation algorithm has multiple, for instance adopt basic binary or ternary participle method, it is judged that whether the word after participle is identical with the word in dictionary.Here we record this and are evaluated as α₁, wherein 0≤α₁≤ 1, if α₁Less, then illustrate relatively low with the Keywords matching in semantic net dictionary, α₁More big, then illustrate high with the matching degree of key word in semantic net.

Generally, it is believed that word is to be made up of two words, or three word compositions.If it is considered to be made up of two words, in short, from lead-in, each two word is divided into a word；Remove lead-in then one word of each two word afterwards, so, just obtain a component word, divide again after reality can also remove some function words.It is also contemplated that seriality between word in actually used, or context relation, and utilize the expression key word of above-mentioned binary, ternary participle legal principle solution user input.

It should be noted that, although the present invention illustrates how to understand the expression key word of user's input with binary, ternary participle method for example, but the invention is not restricted to this, this participle understanding method is only the segmenting method of an example of present invention citing, it is true that multiple segmenting method can be adopted in embodiments of the present invention, or the method for the concept extraction of natural language processing field or feature extraction, obtain a key word or the feature representation of corresponding input.

In one embodiment, if user's input is a complete expression, as having the sentence of SVO composition, such as " general merchandise of where doing shopping is relatively good ", or short sentence such as " famous general merchandise ", or short key word such as " Yaohan ", then can pass through segmenting method and obtain " famous " " general merchandise ", " famous general merchandise ", " Yaohan ", and then based on user location " Shanghai ", it is determined that the annexation of key word is " general merchandise " → " famous general merchandise " → " Yaohan ".

In one embodiment, can also according to for determining semantic net, Address Recognition Network Based and time record, user location and search time can be obtained, based on different areas, time, adjust semantic net so that search engine can be better understood from expressing key word and setting up the relation of key word annexation, key word and objective result according to key word of user.

2, judge whether the key word that user is currently entered reaches Keywords matching with the user input recorded and there is certain matching relationship.This key word recorded can be the key word of semantic net node in the present invention.Can adopting coupling and similarity two kinds of rules of coupling completely, the needs that select of specific rules are determined according to concrete service field.It is such as the semantic net serving personal user, it is possible to adopt complete matched rule；It is such as serve the rule that the semantic net in group or vertical field can adopt similarity to be master；It is such as the semantic net that a service area is bigger, then preferably with similarity rule.Here we record this and are evaluated as α₂, wherein 0≤α₂≤1。

Work as α₁During less than certain threshold value, then it is assumed that this key word is not the information representation of a standard, and system casts out this key word.

Work as α₁More than certain threshold value, and α₂Time again less than certain threshold value, it is believed that user's input is a qualified key word that can be used by system.So this node is added in the relevant position in semantic net.Record the follow-up information whether having key word to continue with, and recording subsequent user selection simultaneously.

Work as α₁More than certain threshold value, and α₂Time again more than certain threshold value, then it is assumed that certain node of user's input and semantic net matches, at this moment can through calculating obtaining the multi-stress considering distance and weight that select answer from this node to end user, be designated as β.

It should be noted that α₁、α₂Size be as the criterion with actual selection, different size of value can be selected in different field, the size of the two value is not done specifically defined by the present invention.

In one embodiment, multi-stress β is by β₁、β₂Two factors calculate and to get.Wherein, factor-beta₁, it is the shortest path calculating this key word to final report information.Factor-beta₂, it is calculate this key word to the final weight coefficient reporting information, weight here is reach the number of times of objective result according to key word in semantic net, calculates through historical statistics, normalization and obtains.It is also conceivable to other factors and determine this weight factor, for instance consider time factor, in time period result often, might not be many in the next time period.

Optimized algorithm COMPREHENSIVE CALCULATING factor-beta can be adopted₁And factor-beta₂Multi-stress β.For example, it is possible to first select the expression of several shortest paths, in the expression of shortest path, the expression that selecting paths weight is maximum, and then determine multi-stress β；Can also first select the expression that several weights are maximum, the expression that selecting paths is the shortest in the expression that weight is maximum, and then determine multi-stress β；The composite weighted that can also consider path and weight determines multi-stress etc..Arranging two class data with the height of multi-stress to select for user: the first kind is the Search Hints key word recommended for user, Equations of The Second Kind is the possible Search Results recommended for user.

Search engine differentiates that the degree of correlation ultimate principle of a key word and webpage is to adopt pagerank technology, does not account for the calculating process that user uses and expresses behavior in prior art.In one embodiment, it is possible to the search service system dependency annexation of key word and objective result and the weight relationship system of being added to applied differentiates in computing formula, improves correlation computations process and the result of formula.If namely occurring in that this key word in webpage, it is first considered that it is correlated with, degree of correlation sorts according to the significance level of webpage, this significance level is calculated by pagerank model, if namely a webpage is in an important website, and points to or link the quantity more many (being also contemplated for linking the webpage significance level of this webpage) of this webpage.

It should be noted that, although the present invention is with α₁For threshold value, it is judged that whether the expression of user's input is a normal expression, uses α₂For threshold value, it is judged that the key word of input and the similarity of key word in semantic web data storehouse.But the present invention is to α₁、α₂Size do not do concrete restriction, those skilled in the art can set the size of the two value according to practical situation.

Based on existing search technique, the semantic net that the embodiment of the present invention is formed is a space net structure expressing key word and objective result, this cancellated generation is the record by user's use procedure and refines generation, therefore can improve the calculating process being arrived an objective result list by a key word, namely can improve traditional pagerank technology.

So, by learning the search procedure of user, objective result based on the key word of user's input, the key word of adjustment and finally selection, and set up semantic net by matching degree method and similarity based method, this semantic net includes the relation between the relation between key word and key word and key word and objective result.

The search engine of the embodiment of the present invention realizes method, people can be obtained for the diversity in problem understanding, description, eliminate individual variation and bring the problem that the adaptability of standard product is not enough, solve in real world applications can not by formed relative normalized semantic net and with the problem that this semantic net is combined with specific field application.

Fig. 4 illustrates that the search engine of one embodiment of the invention realizes the flow chart of optimization method, and as shown in Figure 4, the method includes:

Step S401, determine service field.

In one embodiment, first determining the field to service, this service field can make the fields such as some people, a certain group, the such as hotel service, medical services of a certain field, Chinese character input；For a certain pervasive field, can also specifically not be limited to certain specifically field.

Step S402, determine the information service factor.

The system recommendation result of step S403, acquisition user search or use.

Step S404, user's input and characteristic key words Similarity Measure.

When performing Similarity Measure, it is necessary to perform step S405 query semantics gateway keyword, information representation storehouse, according to semantic net key word, information representation storehouse, it is judged that expressing and the similarity of key word in existing semantic net of user's input.

Step S406, judge whether to reach to evaluate threshold value.

Can judge whether the expression that user inputs is a normal expression according to semantic net dictionary, the determination methods of similarity can be adopted, judge whether the key word that user inputs is a normal expression, if a normal expression, then optimize semantic net according to the key word of user's input and the objective result of selection, if not normal expression, then give up this key word, enter common search routine, namely perform step S407.

Step S407, not up to threshold value, flow process terminates.

Step S408, user application search key word and final selection (or non-selected) result are formed and optimize semantic net.

The similarity of the existing key word of step S409, judgement and semantic net.

Step S410, semantic net key word, information representation storehouse.

Step S411, judge whether to reach semantic net threshold value.

Step S412, strengthening semantic net relevant oriented link.

Step S413, in semantic net, add new node and objective result.

Same problem is had different expression by different user, and service field knowledge is difficult to be easily absorbed on a cognitive level by the user simultaneously.User while learning the knowledge in strange field, need to constantly adjusting input simultaneously and expressing, could obtain required information in a strange field.Systematic learning inputs the record of certain time for user search, analyzes the change that in record, user search input adjusts, and user is in a continuous input process in study, the selection of final result and the difference of the objective result that should select is expressed key word.Meanwhile, the method provided according to this patent embodiment maybe can also differentiate backstage can diversity between quality of service information, the selection according to user, according to using frequency, information is carried out an Optimal scheduling.

For same problem, identify that user's difference expresses key word, apply this historical record, the history input expression that Learning Service success is corresponding, as expressed key word by the history input obtaining objective result that adjusts of key word, conceptual understanding and feature extraction can be carried out together with information on services content with backstage, form a kind of input prompt method based on semantic net or the possible input key word prompting of user and the prompting of corresponding objective result.

The search engine of the embodiment of the present invention realizes method, and based on the conceptual understanding to service field knowledge, the record in search engine interaction excavates, the problem understanding user search request.Based on user, the information of input is understood, use the mode of a kind of semantic net optimizing autonomous evolution to be optimized system.

Search engine service system or the information recommendation system such as the method that can be realized by above-mentioned search engine is applied to cell-phone customer terminal, handheld device, internet device, voice information services are attended a banquet.So, by the search service system history serve log such as user, groups of users, service field are analyzed, the valuable available information that can be used by user later or system is extracted.The aspects such as Service Source dispatching patcher can also be applied to, it is possible to differentiate the class of business of user's request, and reasonably dispatch backstage according to the feature of class of business and current information service and be provided that the resource of information service, make background service maximized.In user's service afterwards, the user that the offer of the above-mentioned valuable available information extracting is later can be utilized, Consumer's Experience can be improved, strengthen the service effectiveness of service enterprise, and then obtain bigger commercial interest, be conducive to service provider to carry out better resource distribution simultaneously, save cost of serving.The search engine of the embodiment of the present invention realizes method can go out user's search engine easy to use and related telecommunications information service system according to the capability development of the ability of information service firm network, service, it is possible to realize the products such as method the exploitation product such as similar Siri, the service robot that make new advances according to this search.

In one embodiment, propose a kind of can for individual, group, field the searching method of application, the method formed be a kind of can perception field change and a certain period class problem expression change semantic network, according to this network, the diversity that different user describes can be overcome for same thing difference, make different user for the different description of same thing, it is also possible to obtain identical result.

Fig. 5 illustrates the flow chart of the searching method of one embodiment of the invention.As it is shown in figure 5, the method specifically includes that

Step S501, determine that user input reaches the expression key word in formula.

Judge that user input reaches whether key word is a complete word, or whether be a complete expression.Evaluation algorithm has multiple, for instance adopt basic binary or ternary participle method, it is judged that whether the word after participle is identical with the word in dictionary.

In one embodiment, the expression of user's input potentially includes multiple key word, understands the meaning of expression of user's input based on the key word in semantic net, it is determined that user input reach in expression key word.

In one embodiment, it is possible to adopt the evaluation of estimate α in above-described embodiment₂, wherein 0≤α₂≤ 1, if α₂Less, then illustrate relatively low with the Keywords matching in semantic net dictionary, α₁More big, then illustrate high with the matching degree of key word in semantic net.May determine that user is intended to the key word of search according to this evaluation of estimate.

Step S502, determining and express the start node that mate in semantic net of key word, wherein, this semantic net is based on the expression key word in user search history and objective result generation.

In one embodiment, semantic net interior joint can be the set of multiple equivalent in meaning or close key word.

By understanding the expression analyzing user, obtain after expressing key word, this expression key word is matched with the key word in semantic net, the one or more start node of determination in semantic net.Key word and the matching relationship of key word in semantic net can be judged according to above-mentioned similarity based method.

The selection of start node, has several computational methods.One of which computational methods are to calculate the similarity between the expression and some node (key word or certain expression characteristic) that input, if being all word, just calculating the number of same word between word and word, and whether order is correlated with, namely Keywords matching calculates.If expressing, it is possible to adopt the comparison between syntactic analysis, semantic analysis, key word vector to calculate.

In one embodiment, it is judged that whether the expression key word of user's input exists and the matching relationship of semantic net interior joint, and semantic net node can be a key word, it is also possible to be multiple same or like crucial combinations.Can adopting the rule such as coupling and similarity coupling completely, the needs that select of specific rules are determined according to concrete service field.

For example, if serving the semantic net of personal user, it is possible to adopt complete matched rule；If the rule that the semantic net serving group or vertical field can adopt similarity to be master；If the semantic net that a service area is bigger, then preferably with similarity rule.

It should be noted that in specific implementation process, it is possible to the situation according to information service resource, select Similarity measures rule flexibly.

If the key word that the expression according to user's input is determined and the similarity of key word in semantic net are more than certain value, then select this key word in semantic net as start node, and then provide the user service.

Step S503, determining objective result according to start node in semantic net, objective result is determined according to distance and the weight of start node to objective result；

Step S504, prompting objective result.

Semantic net determines after start node, according to relation between key word and key word in semantic net, and the relation between key word and objective result, objective result can be provided the user and show on the display apparatus, if there being multiple objective result, point out objective result according to the weight sequencing of objective result.

The searching method of the embodiment of the present invention, the expression formula of user's input can be analyzed, expression keyword query semantic net according to user's input, semantic net points out key word and information service according to the annexation of key word and key word and the weight of key word and objective result for user, user search information time can be saved, provide the user the information service of more precise and high efficiency.

Fig. 6 illustrates the flow chart of the searching method of another embodiment of the present invention.In the step S601 of Fig. 6～step S603 and Fig. 5, corresponding step S501～step S503 is essentially identical, repeats no more for brevity herein, and as shown in Figure 6, the method also includes:

Step S604, determines prompting key word according to start node in semantic net；

Step S605, prompts the user with prompting key word；

Step S606, points out objective result.

Determine in semantic net after start node, it is determined that this start node develops to multiple possible paths of multiple destination nodes, according to the relation between key word and key word, provide the user user and may want to the key word of input, and point out this key word for user.

In one embodiment, in semantic net, determine that prompting key word includes according to start node: if start node is not the endpoint node of semantic net, select start node to develop to the path of objective result, it is determined that the node on path is as prompting key word.

In one embodiment, if start node is the endpoint node of semantic net, objective result corresponding for start node is prompted to user；If start node is not the endpoint node of semantic net, select the weight factor objective result more than weight threshold of objective result；Select the path factor that start node develops to objective result more than the path of path threshold；Determine that other nodes on path are as prompting key word；Prompting prompting key word and/or multiple objective result；Wherein, the node in semantic net is the set of same or analogous key word composition, and weight factor is determine according to the number of times normalization of key word arrival objective result in the semantic net of historical statistics.

Objective result can be the Enterprise Lists of service in one embodiment, after user inputs the key word of this network arbitrary node, if this key word is in last layer of semantic network, just can the Enterprise Lists triggered by this key word or information on services, arrange according to the height of weight or the frequency, the search service system dependency this relation system of being added to can also applied differentiates in computing formula, improves correlation computations process and the result of formula.

If attending a banquet, user inputs keyword search available enterprise information on services, when this key word is in the non-final node of network, then can calculate and be developed to the path of objective result by this node, obtain the several paths with final goal result reached on threshold value, obtain on path simultaneously and changed key word, now realize the semantic net of formation in method based on above-mentioned search engine, search engine system can give user's prompting of attending a banquet search input possible with this key word or objective result.

After inputting key word by search engine, the semantic net that method generates is realized based on search engine, attend a banquet or user obtains one and is likely to provide the user service or meet the service list of customer information requirement, attend a banquet by linking up (or user oneself selects in lists) with user, have selected the final company information reported for user.

In this service process, by the accumulation of serve log, a multi-to-multi key word correspondence can be obtained and report the list of enterprise.Further, owing to obtaining the key word adjusted in information seeking processes of attending a banquet, a limited network structure of the number of plies can therefore be formed.Therefore, in information service process, when clearly inputting a key word when attending a banquet, the backstage once selected enterprise provided can just be judged.Therefore, it can attend a banquet every time scan for time, those are triggered by this key word, once provide the user before the company information of satisfactory service is arranged in, saved user and searched the information of related service enterprise.

Fig. 7 illustrates the flow chart of the searching method of another embodiment of the present invention.As it is shown in fig. 7, the method specifically includes that

Step S701, user input key word or problem.

Step S702, match calculating for input and semantic net node.

Matching primitives can be carried out, it is also possible to carry out matching primitives according to the semantic net that artificial statistics is formed according to the expression of the semantic net that above-mentioned search engine implementation is formed with user's input.In one embodiment, perform step S711, search semantic gateway keyword, information representation storehouse, carry out matching primitives according to key word in this semantic net, it is judged that whether the expression of user's input is a normal expression.

In one embodiment, if user's input is an expression,

Step S703, judge whether to reach to evaluate threshold value.

If not up to threshold value, step S712, entrance search routine.

Step S705, selection semantic net start node.

Step S706, calculating start node report the weight distance of information to target.

Step S707, semantic net key word, information representation storehouse.Inquire about this semantic net key word, whether information representation storehouse exists recommended keywords and information on services, if not finding, performing step S712 and entering search routine, if finding, performing step S708.

Step S712, entrance search routine.

When user only selects the key word of recommendation, after not selecting the objective result recommended, the search system needs search click commands according to user, again search in data base.Search Results now arranges according to system relatedness computation algorithm, the dependency that can semantic net be accumulated is as general search system, one input parameter of relevancy algorithm, general search system is relative to the search system for the search system of semantic net, it is possible to refer to the existing system being not based on semantic net.

Step S708, recommend the semantic net key word on threshold value and corresponding objective result.

Step S709, user judge recommendation results and select.

The method that can adopt similarity is mated with key word in semantic net.If the expression of user's input is a word, then the key word directly and in semantic net carries out matching primitives, the similarity according to the key word of user's input with semantic net key word, it is determined that the start node of search.If user's input is a sentence, according to certain segmenting method participle, it is determined that user input reach in multiple key words, it is possible in semantic net, select multiple start node, according to node each in semantic net and weight distance between objective result, recommend objective result for user.

Step S713, record user select, and utilize negative feedback adjusting and optimizing semantic net, and enter search routine.

After user input reaches, according to semantic net recommend or user oneself select objective result with negative feedback mode adjust above-mentioned semantic net.

Step S710, record user select and optimize semantic net.

The searching method that the embodiment of the present invention provides is a kind of method that searching request is understood and result is recommended that Kernel-based methods knowledge concepts extracts, mark prompting can be understood based on the searching request of domain knowledge and service-aware simultaneously, search system can when user scans for inputting, prompt the user on how to correctly enter information according to the information content feature that backstage can service, or the backstage important service information very first time is fed back to user interface.

The Information Requirement Characteristic of the searching method perception user of the embodiment of the present invention and the change of customer information requirement environment, the searching request of user is carried out the sightless integrity mark of user, the Search Results making search engine more conforms to the demand of user, and adapts to the requirement of demand environment.

Fig. 8 illustrates that a sphere of learning one of one embodiment of the invention teaches the schematic diagram of the semantic net of group, based on this semantic net, if semantic net being taught 1 node include multiple key word such as " Di Mubainasi-Li ", " Di Muli ", " Bai Nasi-Li ", " TimLee ", " Berners-Lee " " TimBerners-Lee ", if user inputs the name such as " Di Mubaina-Li " of professor 1, according to this semantic net, the key word that user is inputted by search system matches with the key word in semantic net dictionary, find that the word of user's input compares alike with " Di Mubainasi-Li ", therefore the node that can select professor 1 place in semantic net is start node, objective result according to being connected with this start node provides the user information on services.

Professor 1 is in the endpoint node of semantic net, it is joined directly together with objective result, have the objective result of annexation to include " university " " professor " " author " " semantic net " " computer network system " with this start node, if the weight distance of professor 1 and above-mentioned objective result be " 100 " " 200 " " 300 " " 400 " " 1000 ", the objective result recommended for user according to weight distance be ordered as " computer network system " " semantic net " " author " " professor " " university ".Prompting key word such as " professor 2 " " field 1 " " works 1 " " works 3 " " works 2 " " student 1 " " laboratory 1 " can also be provided the user.Considering the relation of path and weight, in this semantic net, each calculates the multi-stress of each key word and the objective result being associated, and this multi-stress is according to weight and what distance aggregative weighted obtained.If after judgement start node, directly recommend objective result or prompting key word for user according to the multi-stress size of this start node with each objective result.The method of the concrete method recommending objective result or prompting key word is identical with the recommendation method in above-mentioned searching method, repeats no more herein.

The searching method of the embodiment of the present invention, relate to in the information search inquiry system service processes such as search engine, for specific user individual, user application group, question answering in specific vertical industry service process evaluation, extraction and the knowledge accumulation etc. of application message can be accumulated.According in service process, for the record that particular problem is answered, by similar answer answer, the initiation problem triggering this answer can be obtained in turn.Further, after user inputs similar initiation problem, do not need the adjustment through user's key word, directly provide the user possible search answer.

Fig. 9 illustrates the structural representation of the search engine system of one embodiment of the invention.As it is shown in figure 9, search engine realizes and the module of optimization side includes:

User's usage data record module 905, record key word and selection objective result information, for instance, record in the information service process of a definition, key word, the key word of adjustment and the final Search Results selected that user inputs.

Serve log storehouse 906, deposits the information that user's usage data record module 905 records.

User uses information analysis evaluation module 907, is used for evaluating user and inputs the similarity of key word and key word in keyword criteria storehouse and semantic net node key word.

Candidate keywords, problem add up storehouse 908, and user deposits the key word of standard, and the key word in this dictionary can be used to evaluation calculation α₁。

Key word, information representation storehouse 909, key word and objective result based on user count key word, information representation storehouse.

Semantic net is formed and optimizes computing module 910, according to user search behavior, forms a reasoning semantic net, and this semantic net is constantly optimised in Reusability process.

Recommending module 911 based on semantic net, select the key word recommended or select the objective result recommended, the search certain to user is pointed out, and after a user input, is prompted to user by the Search Results understanding the information Perception user analyzing user's input possible.

User's application side includes such as lower module:

User inputs key word module 901, for receiving the search key word of user's input；

Information search recommending module 902, it is possible to recommend based on semantic net information and be supplied to user.

Recommendation results represents module 903, and user's application represents different contents according to different result of calculation, for instance recommended keywords, it is recommended that the result that key word is corresponding, final Search Results etc..

Enter search routine module 904, if the non-selected search engine of user realizes the result that flow process is recommended, then enter general search flow process, perform common search.

When user only selects the key word of recommendation, after not selecting the Search Results recommended, search engine realizes and optimizes side needing the search click commands according to user, again search in data base, Search Results now arranges the relatedness computation algorithm originally set according to system, and the dependency of semantic net accumulation can input parameter as the one of system relevancy algorithm.

The search engine system of the embodiment of the present invention, it is not substitute existing search system completely, it acts on mainly two: the semantic net being the input according to user and search engine realization, recommend the information on services relevant to input key word for user, go out through history service successfully search objective result record COMPREHENSIVE CALCULATING during the information on services that this is correlated with.Another effect is to recommend to be currently entered relevant input key word to user, correct user error or there is no the input of Search Results or provide the user with possible correctly entering, when user does not select key word and the Search Results of these recommendations, system can input still according to the input of user.

The sequence of the Search Results for carrying out according to user's input, it is possible to select the accumulation considering semantic net to adjust ranking results, it is also possible to be left out the accumulation of semantic net, be ranked up according to the relatedness computation method of system self.

The search engine system of the embodiment of the present invention, record is searched for according to for a certain individual, groups of users, field or the information on services in a period of time, study user search use habit, refines the semanteme between a kind of key word, information representation mode and final goal result and conceptual relation；The information that simultaneously backstage can be serviced carries out conceptual understanding, extraction, forms concept characteristic network corresponding between user's input with final service result.When user inputs a searching request, according to the method or apparatus that this search engine system provides, recommend the information on services expression relevant to this searching request and the background information that can service for user.It is possible to further realize servicing concept extraction and the application of data for search engine history, provide the user a kind of input method, information cuing method or information recommendation system.

Figure 10 illustrates that the search engine of one embodiment of the invention realizes the structured flowchart of device, and as shown in Figure 10, this search engine realizes device 1000 and includes:

Acquisition module 1001, the objective result of search expression history and selection for obtaining input in one continuous search procedure of user；

Determine module 1002, for determining expression keyword sequence according to search expression history；

Generation module 1003, for according to expressing keyword sequence and objective result generative semantics net, semantic net includes expressing between key word and expresses the annexation between key word and objective result and weight.

The search engine of the embodiment of the present invention realizes device, carry out statistics set up semantic net based on the expression key word of user's input and the objective result of selection, obtain the distance between key word and key word and between key word and objective result and weight relationship, in user search process, service is provided the user based on this semantic net, user search information time can be saved, provide the user the information service of more precise and high efficiency.

In one embodiment, expression key word, the expression key word of adjustment and the objective result according to Search Results selection that acquisition module inputs for obtaining user.

Figure 11 illustrates that the search engine of another embodiment of the present invention realizes the structured flowchart of device, and as shown in figure 11, this search engine realizes device 1100 and includes:

Module 1104 is determined in service field, is used for determining that service field, service provider include sole user, group；

Information service process judge module 1105, for judging that expressing of the input of user under service field is expressed key word and be whether complete information service process according to the process expressing Keyword Selection objective result in keyword sequence.

In one embodiment, this search engine realizes device 1100 and also includes: evaluation module 1106, if the expression key word for user's input is unsatisfactory for semantic web updates rule, then casts out described expression key word；

If the expression key word that evaluation module 1106 is additionally operable to user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are less than similarity threshold, the expression key word of user's input is added as the new keywords in semantic net in semantic net, and record the objective result that user selects, generate or the multi-stress of the annexation weight that updates in described new keywords and semantic net between other key words and described new keywords and objective result；

If the expression key word that evaluation module 1106 is additionally operable to user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are be more than or equal to similarity threshold, generate or the multi-stress of key word and objective result in update semantics net, multi-stress is the shortest path factor according to the key word in semantic net to objective result and weight factor is determined, weight factor is arrive the number of times normalization of objective result according to key word in the semantic net of historical statistics to determine.

In one embodiment, information service process judge module 1105 is used for:

Determine that user input reaches the process of the key word search targets result number of clicks to each objective result；If the number of clicks of objective result is more than information service threshold value, then service process is defined as complete service process, or the process of the objective result selected after user input reaches key word is defined as complete service process.

Figure 12 illustrates the structured flowchart of the searcher of one embodiment of the invention, and as shown in figure 12, this searcher 1200 includes:

Express key word analysis module 1201, for determining that user input reaches the expression key word in formula；

Start node analysis module 1202, expresses, for determining, the start node that key word mates in semantic net, and wherein, semantic net generates based on the expression key word in user search history and objective result；

Determining module 1203, for determining objective result according to start node in semantic net, objective result is determined according to distance and the weight of start node to objective result；

Reminding module 1204, is used for pointing out objective result.

The searcher of the embodiment of the present invention, it is possible to the expression query semantics net according to user's input, is expressed as user according to the input of user and points out key word and information service, it is possible to saves user search information time, provides the user the information service of more precise and high efficiency.

In one embodiment, it is determined that module is additionally operable in semantic net to determine prompting key word according to start node；

Reminding module is additionally operable to prompting prompting key word.

In one embodiment, reminding module is used for:

If start node is not the endpoint node of semantic net, start node is selected to develop to the path of objective result, it is determined that the node on path is as prompting key word.

In one embodiment, reminding module, if being the endpoint node of semantic net for start node, objective result corresponding for start node is prompted to user；

And/or

If reminding module is for the endpoint node that start node is not semantic net, select the weight factor objective result more than weight threshold of objective result；Select the path factor that start node develops to objective result more than the path of path threshold；Determine that other nodes on path are as prompting key word；Prompting prompting key word and/or multiple objective result, wherein, the node in semantic net is the set of same or analogous key word composition, and weight factor is determine according to the number of times normalization of key word arrival objective result in the semantic net of historical statistics.

Figure 13 illustrates that the search engine of an alternative embodiment of the invention realizes the structured flowchart of device.It can be possess the host server of computing capability, personal computer PC or portable portable computer or terminal etc. that search engine realizes device 1300.Implementing of computing node is not limited by the specific embodiment of the invention.

Search engine realizes device 1300 and includes processor (processor) 1310, communication interface (CommunicationsInterface) 1320, memorizer (memory) 1330 and bus 1340.Wherein, processor 1310, communication interface 1320 and memorizer 1330 complete mutual communication by bus 1340.

Communication interface 1320 is used for and network device communications, and wherein the network equipment includes such as Virtual Machine Manager center, shares storage etc..

Processor 1310 is used for performing program.Processor 1310 is probably a central processor CPU or application-specific integrated circuit ASIC (ApplicationSpecificIntegratedCircuit), or is configured to implement one or more integrated circuits of the embodiment of the present invention.

Memorizer 1330 is for storing documents.Memorizer 1330 is likely to comprise high-speed RAM memorizer, it is also possible to also include nonvolatile memory (non-volatilememory), for instance at least one disk memory.Memorizer 1330 can also be memory array.Memorizer 1330 is also possible to by piecemeal, and block can by certain rule sets synthesis virtual volume.

In a kind of possible embodiment, said procedure can be the program code including computer-managed instruction.This program is particularly used in: obtain the search expression history of input in one continuous search procedure of user and the objective result of selection；Expression keyword sequence is determined according to search expression history；According to expressing keyword sequence and objective result generative semantics net, semantic net includes expressing between key word and expresses the annexation between key word and objective result and weight.

In one embodiment, it is thus achieved that the search expression history of input and the objective result of selection in one continuous search procedure of user, including:

Expression key word, the expression key word of adjustment and the objective result according to Search Results selection that acquisition user inputs.

In one embodiment, also include according to before expressing keyword sequence and objective result generative semantics net:

Determine that service field, service provider include sole user, group；

Judge that expressing of the input of user under service field is expressed key word and be whether complete information service process according to the process expressing Keyword Selection objective result in keyword sequence.

In one embodiment, semantic net includes expressing between key word and expresses the annexation between key word and objective result and weight, including:

If the expression key word of user's input is unsatisfactory for semantic web updates rule, then cast out described expression key word；

If the expression key word of user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are less than similarity threshold, the expression key word of user's input is added as the new keywords in semantic net in semantic net, and record the objective result that user selects, generate or the multi-stress of the annexation weight that updates in described new keywords and semantic net between other key words and described new keywords and objective result；

If the expression key word of user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are be more than or equal to similarity threshold, generate or the multi-stress of key word and objective result in update semantics net, multi-stress is the shortest path factor according to the key word in semantic net to objective result and weight factor is determined, weight factor is arrive the number of times normalization of objective result according to key word in the semantic net of historical statistics to determine.

In one embodiment, judge that expressing of the input of user under service field is expressed key word and be whether complete information service process according to the process expressing Keyword Selection objective result in keyword sequence, comprise determining that user input reaches the process of the key word search targets result number of clicks to each objective result；If the number of clicks of objective result is more than information service threshold value, then service process is defined as complete service process, or the process of the objective result selected after user input reaches key word is defined as complete service process.

The search engine of the embodiment of the present invention realizes device, expression formula and selection objective result according to user's input carry out statistics and set up semantic net, service is provided the user, it is possible to save user search information time, provide the user the information service of more precise and high efficiency based on this semantic net.

Figure 14 illustrates the structured flowchart of the searcher of an alternative embodiment of the invention.Searcher 1400 can be possess the host server of computing capability, personal computer PC or portable portable computer or terminal etc..Implementing of computing node is not limited by the specific embodiment of the invention.

Searcher 1400 includes processor (processor) 1410, communication interface (CommunicationsInterface) 1420, memorizer (memory) 1430 and bus 1440.Wherein, processor 1410, communication interface 1420 and memorizer 1430 complete mutual communication by bus 1440.

Communication interface 1420 is used for and network device communications, and wherein the network equipment includes such as Virtual Machine Manager center, shares storage etc..

Processor 1410 is used for performing program.Processor 1410 is probably a central processor CPU or application-specific integrated circuit ASIC (ApplicationSpecificIntegratedCircuit), or is configured to implement one or more integrated circuits of the embodiment of the present invention.

Memorizer 1430 is for storing documents.Memorizer 1430 is likely to comprise high-speed RAM memorizer, it is also possible to also include nonvolatile memory (non-volatilememory), for instance at least one disk memory.Memorizer 1430 can also be memory array.Memorizer 1430 is also possible to by piecemeal, and block can by certain rule sets synthesis virtual volume.

In a kind of possible embodiment, said procedure can be the program code including computer-managed instruction.This program is particularly used in: determine that user input reaches the expression key word in formula；Determining and express the start node that key word mates in semantic net, wherein, semantic net generates based on the expression key word in user search history and objective result；Determining objective result according to start node in semantic net, objective result is determined according to distance and the weight of start node to objective result；Prompting objective result.

In one embodiment, also include: in semantic net, determine prompting key word according to start node；Prompting prompting key word.

In one embodiment, in semantic net, determine that prompting key word includes according to start node: if start node is not the endpoint node of semantic net, select start node to develop to the path of objective result, it is determined that the node on path is as prompting key word.

In one embodiment, if start node is the endpoint node of semantic net, objective result corresponding for start node is prompted to user；If start node is not the endpoint node of semantic net, select the weight factor objective result more than weight threshold of objective result；Select the path factor that start node develops to objective result more than the path of path threshold；Determine that other nodes on path are as prompting key word；Prompting prompting key word and/or multiple objective result；Wherein, the node in semantic net is the set of same or analogous key word composition, and weight factor is determine according to the number of times normalization of key word arrival objective result in the semantic net of historical statistics.

The searcher of the embodiment of the present invention, it is possible to the expression query semantics net according to user's input, is expressed as user according to the input of user and points out key word and information service, it is possible to saves user search information time, provides the user the information service of more precise and high efficiency.

Description of the invention provides for example with for the purpose of describing, and is not exhaustively or limit the invention to disclosed form.Many modifications and variations are obvious for the ordinary skill in the art.Selecting and describing embodiment is in order to principles of the invention and practical application are better described, and makes those of ordinary skill in the art it will be appreciated that the present invention is thus design is suitable to the various embodiments with various amendments of special-purpose.

Claims (18)

1. a search engine realizes method, it is characterised in that including:

Obtain the search expression history of input in one continuous search procedure of user and the objective result of selection；

Expression keyword sequence is determined according to described search expression history；

According to described expression keyword sequence and described objective result generative semantics net for search engine, described semantic net includes the annexation between described expression key word and between described expression key word and described objective result and weight.

2. method according to claim 1, it is characterised in that obtain the search expression history of input in one continuous search procedure of user and the objective result of selection, including:

Obtain user input initially express key word, adjustment express key word and according to Search Results select objective result.

3. method according to claim 1, it is characterised in that also include according to before described expression keyword sequence and described objective result generative semantics net:

Determine that service field, described service field include sole user, group, vertical field or integrated information service field；

Judge that expressing of the input of user under described service field is expressed key word and be whether complete information service process according to the process of described expression Keyword Selection objective result in keyword sequence.

4. method according to claim 3, it is characterized in that, judge that expressing of the input of user under described service field is expressed key word and be whether complete information service process according to the process of described expression Keyword Selection objective result in keyword sequence, including:

Determine that user input reaches the service process of the key word search targets result number of clicks to each objective result；If the number of clicks of objective result is more than information service threshold value, then described service process is defined as complete service process；

And/or

The process of the objective result selected after user input reaches key word is defined as complete service process.

5. method according to claim 1, it is characterised in that described semantic net includes the annexation between described expression key word and between described expression key word and described objective result and weight, including:

If the expression key word of user's input is unsatisfactory for semantic web updates rule, then cast out described expression key word；

If the expression key word of user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are less than similarity threshold, the expression key word of described user input is added as the new keywords in semantic net in described semantic net, and record the objective result that described user selects, generate or the multi-stress of the annexation weight that updates in described new keywords and semantic net between other key words and described new keywords and objective result；

If the expression key word of user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are be more than or equal to similarity threshold, generate or update the multi-stress of key word and objective result in described semantic net, described multi-stress is the shortest path factor according to the described key word in semantic net to objective result and weight factor is determined, weight factor is arrive the number of times normalization of objective result according to key word in the semantic net of historical statistics to determine.

6. a searching method, it is characterised in that including:

Determine that user input reaches the expression key word in formula；

Determining the start node that described expression key word mates in semantic net, wherein, described semantic net generates based on the expression key word in user search history and objective result；

Determining objective result according to described start node in described semantic net, described objective result is determined according to the described start node distance to described objective result and weight；

Point out described objective result.

7. method according to claim 6, it is characterised in that also include:

Described semantic net determines prompting key word according to described start node；

Point out described prompting key word.

8. method according to claim 7, it is characterised in that described in described semantic net according to described start node determine prompting key word include:

If described start node is not the endpoint node of described semantic net, described start node is selected to develop to the path of described objective result, it is determined that the node on described path is as prompting key word.

9. the method according to claim 7 or 8, it is characterised in that:

If the endpoint node that described start node is described semantic net, objective result corresponding for described start node is prompted to user；

If described start node is not the endpoint node of described semantic net, select the weight factor objective result more than weight threshold of objective result；Select the path factor that described start node develops to described objective result more than the path of path threshold；Determine that other nodes on described path are as prompting key word；Point out described prompting key word and/or the plurality of objective result；

Wherein, the node in described semantic net is the set of same or analogous key word composition, and weight factor is determine according to the number of times normalization of key word arrival objective result in the semantic net of historical statistics.

10. a search engine realizes device, it is characterised in that including:

Acquisition module, the objective result of search expression history and selection for obtaining input in one continuous search procedure of user；

Determine module, for determining expression keyword sequence according to described search expression history；

Generation module, for according to described expression keyword sequence and described objective result generative semantics net, described semantic net includes the annexation between described expression key word and between described expression key word and described objective result and weight.

11. device according to claim 10, it is characterised in that

Expression key word, the expression key word of adjustment and the objective result according to Search Results selection that described acquisition module inputs for obtaining user.

12. device according to claim 10, it is characterised in that also include:

Module is determined in service field, is used for determining that service field, described service field include sole user, group, vertical field or integrated information service field；

Information service process judge module, for judging that expressing of the input of user under described service field is expressed key word and be whether complete information service process according to the process of described expression Keyword Selection objective result in keyword sequence.

13. device according to claim 10, it is characterised in that including:

Evaluation module, if the expression key word for user's input is unsatisfactory for semantic web updates rule, then casts out described expression key word；

Described evaluation module is additionally operable to the expression key word of user's input and meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are less than similarity threshold, express key word as the new keywords in semantic net what described semantic net added described user input, and record the objective result that described user selects and generate or the multi-stress of the annexation weight that updates in described new keywords and semantic net between other key words and described new keywords and objective result；

If the expression key word that described evaluation module is additionally operable to user's input meets semantic web updates rule, the expression key word of user's input and the similarity of key word in semantic net are be more than or equal to similarity threshold, generate or update the multi-stress of key word and objective result in described semantic net, described multi-stress is the shortest path factor according to the described key word in semantic net to objective result and weight factor is determined, weight factor is arrive the number of times normalization of objective result according to key word in the semantic net of historical statistics to determine.

14. device according to claim 12, it is characterised in that described information service process judge module is used for:

Determine that user input reaches the process of the key word search targets result number of clicks to each objective result；If the number of clicks of objective result is more than information service threshold value, then described service process is defined as complete service process；

And/or

The process of the objective result selected after user input reaches key word is defined as complete service process.

15. a search engine device, it is characterised in that including:

Express key word analysis module, for determining that user input reaches the expression key word in formula；

Start node analysis module, for determining the start node that described expression key word mates in semantic net, wherein, described semantic net generates based on the expression key word in user search history and objective result；

Determining module, for determining objective result according to described start node in described semantic net, described objective result is determined according to the described start node distance to described objective result and weight；

Reminding module, is used for pointing out described objective result.

16. device according to claim 15, it is characterised in that

Described determine that module is additionally operable to determine prompting key word according to described start node in described semantic net；

Described reminding module is additionally operable to point out described prompting key word.

17. device according to claim 16, it is characterised in that described reminding module is used for:

If described start node is not the endpoint node of described semantic net, described start node is selected to develop to the path of described objective result, it is determined that the node on described path is as prompting key word.

18. the device according to claim 16 or 17, it is characterised in that:

Described reminding module, if for the endpoint node that described start node is described semantic net, being prompted to user by objective result corresponding for described start node；

And/or

If described reminding module is for the endpoint node that described start node is not described semantic net, select the weight factor objective result more than weight threshold of objective result；Select the path factor that described start node develops to described objective result more than the path of path threshold；Determine that other nodes on described path are as prompting key word；Point out described prompting key word and/or the plurality of objective result,

Wherein, the node in described semantic net is the set of same or analogous key word composition, and weight factor is to determine according to the number of times normalization of key word arrival objective result in the semantic net of historical statistics.