JP2016115294A - Information retrieval method and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an information retrieval method and an information retrieval device.SOLUTION: An information retrieval method includes: receiving a first query and acquiring a guidance sentence of intention explanation in accordance with the first query; receiving a second query updated in accordance with the guidance sentence of intention explanation and acquiring a retrieval result in accordance with the second query; replying a retrieval result to a client.SELECTED DRAWING: Figure 1

Description

  Embodiments of the present disclosure generally relate to the technical field of computer networks, and more particularly, to an information search method and an information search device.

  With the development of computer technology, Internet applications have become more and more popular, allowing users to obtain the required Internet resources through search engines. Currently, with related search methods, the user first enters a query, the search engine retrieves the search results associated with the query, returns the search results to the client, and the user is required from the finally returned search results. Get a resource.

  However, according to the above search method, an accurate answer is not provided by the search engine because the query input by the user is an unclear or incomplete phrase. Even if the same resources are provided, if the user is not familiar with the domain to which the query belongs, the user will need to modify the query many times in order to perform the search, thus increasing the search cost. Even if the cost is high, the obtained search results still cannot satisfy the user's request.

  The embodiments of the present disclosure are intended to solve, at least in part, at least one of the problems existing in the prior art.

  Therefore, the first object of the present disclosure is to guide and clarify the user's search intention, improve the accuracy of the search engine to identify the user's search intention, satisfy the user's requirements, and enhance the user experience. It is to provide an information retrieval method that can be improved.

  A second object of the present disclosure is to provide an information search device.

  In order to achieve the above object, an embodiment of the first aspect of the present disclosure provides an information search method. The information search method according to the embodiment of the present disclosure receives a first query, acquires a guidance sentence for an intention explanation according to the first query, and is updated according to the guidance sentence for the intention explanation. Receiving a query, obtaining a search result according to the second query, and returning the search result to the client.

  According to the information search method according to the embodiment of the present disclosure, the guidance sentence of the intention explanation provided by the search engine is acquired, the first query is updated according to the guidance sentence of the intention explanation, and the search intention of the user is clear. Search results are provided according to an updated query, such as a second query. Thus, the accuracy of the search engine for identifying the user's search intention is improved and the user's requirements are met, thus improving the user experience.

  An embodiment of the second aspect of the present disclosure provides an information retrieval device. An information search device according to an embodiment of the present disclosure receives a first query and is configured to acquire a guidance sentence of an intention description according to the first query; and the intention description A second acquisition module configured to receive the second query updated according to the guidance sentence and to acquire the search result according to the second query, and to respond the search result A response module.

  According to the information search device according to the embodiment of the present disclosure, the guidance sentence of the intention explanation provided by the search engine is acquired, the first query is updated according to the guidance sentence of the intention explanation, and the search intention of the user is clear. Search results are provided according to an updated query, such as a second query. Thus, the accuracy of the search engine for identifying the user's search intention is improved and the user's requirements are met, thus improving the user experience.

  An embodiment of the third aspect of the present disclosure provides a computer readable storage medium. The computer-readable storage medium includes a computer program for executing an information retrieval method according to an embodiment of the first aspect of the present disclosure, which is executed on a computer.

  Further aspects and advantages of the embodiments of the present disclosure will become apparent in part by being given in part in the following description, or may be learned from the examples of the present disclosure.

5 is a flowchart illustrating an information search method according to an embodiment of the present disclosure. 5 is a flowchart showing the construction of a sentence library in the information search method according to the embodiment of the present disclosure. It is a flowchart which shows acquisition of the guidance sentence of intention description according to the 1st query in the information search method according to the embodiment of the present disclosure. 5 is a flowchart illustrating an information search method according to an embodiment of the present disclosure. It is a schematic diagram which shows the acquisition effect of the interactive sentence pair based on the community-based Q & A resource in the information search method according to the embodiment of the present disclosure. It is a schematic diagram which shows the acquisition effect of the guidance sentence of the intention description in the information search method according to embodiment of this indication. It is a schematic diagram which shows the acquisition effect of the search result according to the 2nd query updated according to the guidance sentence of the intention description in the information search method according to embodiment of this indication. 1 is a block diagram illustrating an information retrieval device according to an embodiment of the present disclosure. FIG.

  Various aspects and advantages of embodiments of the present disclosure will become apparent and more readily appreciated from the following description, with reference to the accompanying drawings.

  Embodiments of the present disclosure will be described in detail by way of example. In the embodiments of the present disclosure, elements having the same or similar functions and elements having the same or similar functions will be denoted by the same reference numerals throughout the description. The embodiments described herein according to the drawings are illustrative and non-limiting examples of the disclosure.

  FIG. 1 is a flowchart illustrating an information search method according to an embodiment of the present disclosure. As shown in FIG. 1, this information retrieval method includes the following steps.

  In step S101, the first query is received, and the guidance sentence for the intention explanation is acquired according to the first query.

  In some embodiments of the present disclosure, the first query can include terms and sentences. The sentence library may be established before the intent explanation guidance text is obtained according to the first query so that the search engine can search the sentence library to obtain the intent explanation guidance text. In some embodiments, the step of establishing a sentence library may be avoided if the sentence library is pre-established.

  In step S102, the second query updated according to the guidance text of the intention explanation is received, and the search result is acquired according to the second query.

  In some embodiments of the present disclosure, the search engine may obtain a second query that is updated according to the intent description guidance text, and obtain a search result according to the second query. For example, the search engine displays the intent explanation guidance text to the user through the client, and the user enters the updated query (eg, the second query) in the input box provided by the search engine, and then the search engine May be searched according to a query that is updated, and a search result may be obtained.

  In step S103, the search result is returned to the client.

  In some embodiments of the present disclosure, after obtaining the search results, the search engine can display the search results to the user by returning the search results to the client. The user can acquire necessary resources based on the displayed search results.

  As described above, the information retrieval method further includes establishing a sentence library. In some embodiments of the present disclosure, as shown in FIG. 2, establishing a sentence library includes the following steps:

  In step S201, the first sentence is acquired according to the first information.

  In some embodiments of the present disclosure, the search engine divides the first information and determines whether the first information belongs to a sentence of a preset type according to a preset feature amount. It may be detected. When the first information belongs to a preset type of sentence, the first sentence may be acquired according to the first information. In some embodiments of the present disclosure, the first information includes a sentence in the community user's knowledge exchange platform, the preset type of sentence includes various questions, and the preset feature quantity is an interrogative dictionary. Or the feature-value of a question sentence can be included.

  In step S202, the second sentence is acquired according to the second information related to the first information.

  In some embodiments of the present disclosure, the search engine divides the second information and determines whether the second information belongs to a predetermined type of sentence according to a predetermined feature amount. It may be detected. When the second information belongs to a preset type of sentence, the second sentence may be acquired according to the second information related to the first information.

  In step S203, at least one sentence pair in the sentence library is established according to the first sentence and the second sentence.

  In some embodiments of the present disclosure, the search engine, after obtaining the first and second sentences, at least one sentence pair according to the first and second sentences to establish or expand a sentence library. And a sentence pair may be stored in the sentence library.

  In some embodiments of the present disclosure, the search engine may search for resources in the community user's knowledge exchange platform. The resource includes a plurality of questions (first information) provided by the user and answers (second information) provided by other users to these questions. After it is determined that both the question and the answer belong to a preset type of sentence, the search engine determines from at least one question (first sentence) and the corresponding answer (second sentence). A sentence pair may be acquired. Thus, a sentence library is constructed. For example, one user asks Baidu Knows what to do if the baby has a fever of 38.5 ° C and another user asks, “It depends on how old the baby is. ". The search engine may acquire a sentence pair including the question and the answer, or may store the sentence pair in a sentence library. Thus, a sentence library is established or expanded.

  In some embodiments of the present disclosure, the search engine may filter the sentence pairs in the sentence library after establishing at least one sentence pair in the sentence library. In some embodiments of the present disclosure, the second sentence in the sentence library is filtered according to preset filtering rules. For example, the first sentence “A place that is interesting to travel on October 1” and the second sentence “Where is an interesting place on October 1” have similar meanings and belong to the same question. Should be filtered.

  In some embodiments of the present disclosure, a user may enter a first query in an input box provided by a search engine after establishing a sentence library. Then, the search engine may search the sentence library according to the first query and obtain a guidance sentence for the intention explanation. In some embodiments of the present disclosure, as shown in FIG. 3, obtaining the intent explanation guidance text according to the first query includes the following steps.

  In step S301, the feature amount information of the first query is acquired, and a first sentence related to the feature amount information is acquired according to the feature amount information.

  In some embodiments of the present disclosure, the search engine first obtains feature information of the first query, and searches the reverse index generated according to the feature information of all the first sentences in the sentence library. Then, the feature amount information may be matched according to the reverse index, and thus the first sentence related to the feature amount information may be obtained.

  In some embodiments of the present disclosure, the search engine retrieves the first sentence when the semantic similarity between the first sentence and the first query is greater than a preset threshold initially. May be. In some embodiments of the present disclosure, the first sentence may be filtered if the semantic similarity between the first sentence and the first query is initially less than a preset threshold. . That is, the first sentence is unrelated to the first query.

  In step S302, the sentence library is searched according to the first sentence, and the second sentence corresponding to the first sentence is acquired.

  In some embodiments of the present disclosure, after obtaining a first sentence, the search engine searches a sentence library according to the first sentence, and obtains a second sentence corresponding to the first sentence. Also good. In some embodiments of the present disclosure, the first sentence may correspond to a plurality of second sentences, and thus searching the sentence library for the second sentence corresponding to the first sentence. May be obtained by

  In some embodiments of the present disclosure, the second sentence is obtained when the first query does not include a preset type keyword of the second sentence. In some embodiments of the present disclosure, the second sentence is filtered if the first query includes a preset type keyword of the second sentence, such as the interrogative word “what”. The

  In some embodiments of the present disclosure, the second sentence is obtained when the semantic similarity between the first query and the second sentence is less than a second preset threshold. In some embodiments of the present disclosure, the second sentence is when the semantic similarity between the first query and the second sentence is greater than a second preset threshold, ie, the first If the query and the second sentence belong to the same type of sentence, they are filtered.

  In step S303, a guidance sentence for intent explanation is acquired according to the second sentence.

  In some embodiments of the present disclosure, a plurality of second sentences are obtained according to second information related to the first information, and the search engine has obtained a second sentence corresponding to the first sentence. Later, the second sentence may be sequenced and clustered.

In some embodiments of the present disclosure, the second sentence is sequenced based on Equation 1.
Here, Qb represents a second sentence (accordingly, Qbs represents the plurality of Qb), P _i represents the voting value of the i-th tree, N is the representative of the number of trees.
Formula 1 is a random forest model formula for calculating the score of Qb (ie, the second sentence). Qbs may be sequenced according to the score. The search engine then clusters the second sentence according to semantic similarity.

  In some embodiments of the present disclosure, the search engine clusters second sentences having the same or similar meaning to obtain a clustered second sentence (ie, type). For example, the sentence “How old is a baby” and the other sentence “How old is a baby” have the same meaning and may be clustered. In some embodiments of the present disclosure, clustering may be performed by at least one of a K-means clustering algorithm (hard clustering algorithm) and a hierarchical clustering algorithm.

  Thereafter, a number of clustered second sentences may be obtained, and the search engine sequences the clustered second sentences to obtain an intent description guidance sentence.

In some embodiments of the present disclosure, the score of the clustered second sentence (ie, type) may be calculated according to Equation 2.
Here, c represents a type, | c | represents the number of Qb in the type, and λ represents a weight value for probability evaluation of the type (0 <λ <1). In some embodiments of the present disclosure, after the score of the clustered second sentence (ie, type) is calculated, the clustered second sentence (ie, type) is sequenced according to the score. The intent guidance sentence with the highest score is obtained and returned to the client for display to the user, whereby the user updates the first query according to the intent explanation guidance sentence Is possible. In the present embodiment, the three intention explanation guidance texts are provided as examples only. In some embodiments of the present disclosure, one or more intent guidance text may be obtained without particularly limiting the present disclosure.

  According to the information search method according to the embodiment of the present disclosure, the guidance sentence for the intention explanation is provided by the search engine, the first query is updated according to the guidance sentence for the intention explanation, and the search result is like the second query. As a result, the search intention of the user becomes clear. In addition, the accuracy of the search engine for identifying the user's search intent is improved and the user's requirements are met, thus improving the user experience.

  FIG. 4 is a flowchart illustrating an information search method according to an embodiment of the present disclosure. This information retrieval method includes the following steps.

  In step S401, an interactive question library is established.

  In some embodiments of the present disclosure, the search engine may establish an interactive question library by searching for questions among community-based Q & A resources. Community-based Q & A resources can include various knowledge exchange platforms such as “Baidu Knows”, “Search Ask”, “Sina iAsk”, etc. Community-based Q & A resources may interact multiple times for the same question. In some embodiments of the present disclosure, questions asked by users are referred to as Qa, and answers provided by other users to question Qa are referred to as Qb. For example, as shown in FIG. 5, the user A asks the question Qa “How to handle when the baby has a fever of 38.5 ° C.”, and the user B gives the answer Qb “It depends on how old the baby is” provide. The question Qa of “how to handle when the baby has a fever of 38.5 ° C.” and the answer Qb of “it depends on how old the baby is” can form a pair of interactive sentences.

  In some embodiments of the present disclosure, the search engine may establish an interactive sentence library by obtaining interactive sentence pairs each consisting of a plurality of Qa and Qb.

  In some embodiments of the present disclosure, as shown in FIG. 5, a complete dialog (such as a sentence pair) for a single question is primarily contained in a question description field, a question supplement field, and an answer field. The question description field contains the main question points, generally brief and short, that are asked by the user. The question supplement field includes additional information provided by the user to supplement and further describe the question asked by the user. The answer field contains answers to this question provided by other users.

  In some embodiments of the present disclosure, the search engine separates the question description field and the answer field, respectively, and sentences in the question description field and the answer field are interrogated according to the question dictionary and the question feature. You may detect whether it is a question. If the sentence is determined to be a question-type question, a Qa-Qb interactive sentence pair consisting of sentence Qa in the question description field and sentence Qb in the answer field is obtained.

  Because question and answer phrases in community-based Q & A resources are free, sentence pairs need to be filtered to ensure the validity of interactive sentence pairs. In some embodiments of the present disclosure, Qa-Qb interactive sentence pairs may be filtered according to filtering rules. In some embodiments of the present disclosure, the filtering rules indicate that the core point of Qb is in the question supplement field, the semantic similarity between Qb and Qa is very large, Qb includes descriptive interrogatives such as “how” and “what to” and inappropriate words. As shown in Table 1, Qa is shown in the first column, Qb is shown in the second column, and the reason for filtering the Qa-Qb interactive sentence pair is shown in the third column.

  In some embodiments of the present disclosure, the interactive question library is established according to the filtered interactive sentence pair after the Qa-Qb interactive sentence pair is filtered.

  In step S402, an intention explanation guidance sentence is provided according to the initial query.

  In some embodiments of the present disclosure, the corresponding intent description guidance text is provided to the user according to a query input by the user, such as the first query.

  In some embodiments of the present disclosure, after establishing the interactive question library, the search engine searches the interactive question library according to the input of the query Q by the user and Qa-Qb interactive statements associated with the query Q. To pair, Qbs is sequenced according to a machine learning model to ensure the association between Qb and Q, Qbs is clustered according to a clustering algorithm to obtain several types, and these Sequence types. In this way, the guidance sentence for the intention explanation is acquired.

  In some embodiments of the present disclosure, the reverse index of the interactive question library is established according to Qa, the reverse index is searched according to query Q, and the Qa-Qb interactive sentence pair associated with query Q is obtained. Is done. The Qa-Qb interactive sentence pairs are then filtered according to the search rules. In some embodiments of the present disclosure, if the semantic similarity between Qa and Q is less than a preset threshold, i.e., if the association between Qa and Q is not large, Qa- Qb interactive sentence pairs are filtered. In some embodiments of the present disclosure, Qa-Qb interactive sentence pairs are filtered if the Q and Qb interrogators are the same. In some embodiments of the present disclosure, if the semantic similarity between Q and Qb is very large, that is, if Q and Qb belong to the same question, Pairs are filtered.

  In some embodiments of the present disclosure, the search engine sequences Qbs after obtaining a Qa-Qb interactive sentence pair associated with Q. In some embodiments of the present disclosure, the search engine obtains feature quantities from the knowledge base “Q-Qa-Qb-other information on community-based Q & A resources” and sequences the Qbs according to the feature quantities. . The feature amount includes the semantic similarity between Qa and Q, the semantic similarity between Qb and Q, the distance between Qb and Qa in the interaction in resources, and the relative between Qb and Q May include at least one of the characteristics, the ratio of the number of characters of Qa to the number of characters of Qb, and the average probability that one Qb will occur among all Qbs.

In some embodiments of the present disclosure, the search engine calculates the score of Qbs by using the machine learning random forest model represented by Equation 1.
Here, P _i represents the voting value of the i-th tree, N represents represents the number of trees, each tree in the forest can be voted yes (1) or no (0) to the Qb .

In some embodiments of the present disclosure, the search engine may sequence the Qbs according to the Qbs score. Then, the search engine may cluster Qbs having the same meaning (for example, “how old is a baby”, “how old is a baby”) according to the semantic similarity of Qbs by the clustering algorithm. The clustering algorithm may include at least one of a K-means clustering algorithm and a hierarchical clustering algorithm. After clustering Qbs, the clustered types of scores can be calculated according to Equation 2.
Here, c represents a type, | c | represents the number of Qb in the type, and λ represents a weight value for probability evaluation of the type (0 <λ <1). The clustered types may then be sequenced according to the score. In some embodiments of the present disclosure, the optimal Qbs is obtained among the three types having the maximum score (ie, guidance texts of three intention explanations). In the present embodiment, the three intention explanation guidance texts are provided as examples only. In some embodiments of the present disclosure, one or more intent guidance text may be obtained without particularly limiting the present disclosure.

  In step S403, the second query updated according to the guidance text of the intention explanation is acquired, and the search result is acquired according to the second query.

  In some embodiments of the present disclosure, after obtaining the intent description guidance text, the search engine updates the first query to obtain the second query according to the intent explanation guidance text, and the first The search result may be acquired according to the query 2. In some embodiments of the present disclosure, a search engine displays an intent description guidance text to a user through a client, and the client enters a second query that is updated according to the intent description guidance text in the search engine input box. Thereafter, the search engine may perform a search according to the second query and obtain a search result.

  In step S404, the search result is returned to the client.

  In some embodiments of the present disclosure, after obtaining the search results, the search engine may display the search results to the user by returning the search results to the client. The user may acquire resources required by the user according to the displayed search result.

  FIG. 6A shows the effect of acquiring the intention explanation guidance text, and FIG. 6B shows the search result acquisition effect based on the second query updated according to the intention explanation guidance text. In some embodiments of the present disclosure, as shown in FIG. 6 (a) and FIG. 6 (b), the user enters the query “Exit Visa” to provide “you to which country” provided by the search engine. Get a guidance statement that includes "Would you like to go" and "Why do you want a visa for travel, study abroad, or others?" The user then enters and searches for the updated query “visa for travel to the UK”. Thus, the search results returned by the search engine will fully satisfy the user's request.

  According to the information search method according to the embodiment of the present disclosure, the guidance sentence for the intention explanation is provided by the search engine, the first query is updated according to the guidance sentence for the intention explanation, and the search result is like the second query. Provided according to the updated query, so that the user's search intention is clear. In addition, the accuracy of the search engine for identifying the user's search intent is improved and the user's requirements are met, thus improving the user experience.

  An embodiment of another aspect of the present disclosure further provides an information retrieval device.

  FIG. 7 is a block diagram of an information retrieval device according to an embodiment of the present disclosure. As shown in FIG. 7, the information search device includes a first acquisition module 100, a second acquisition module 200, a response module 300, and an establishment module 400. In some embodiments of the present disclosure, the first acquisition module 100 includes a first acquisition unit 110, a second acquisition unit 120, and a third acquisition unit 130. In some embodiments of the present disclosure, the third acquisition unit 130 includes a clustering subunit 131 and an acquisition subunit 132. In some embodiments of the present disclosure, the establishment module 400 includes a first acquisition unit 410, a second acquisition unit 420, an establishment unit 430, and a filtering unit 440.

  In some embodiments of the present disclosure, the first acquisition module 100 is configured to receive a first query and acquire a guidance sentence for an intention description according to the first query. The first query may include terms and sentences.

  In some embodiments of the present disclosure, the first acquisition module 100 includes a first acquisition unit 110, a second acquisition unit 120, and a third acquisition unit 130.

  In some embodiments of the present disclosure, the first obtaining unit 110 obtains the feature amount information of the first query, and obtains a first sentence related to the feature amount information according to the feature amount information. It is configured.

  In some embodiments of the present disclosure, the first obtaining unit 110 obtains feature information of the first query, and an inverted index generated according to the feature information of all the first sentences in the sentence library. And the feature amount information matching is executed according to the reverse index, whereby the first sentence related to the feature amount information is acquired.

  In some embodiments of the present disclosure, the first acquisition unit 110 may select the first if the semantic similarity between the first sentence and the first query is initially greater than a preset threshold. Get the statement. In some embodiments of the present disclosure, the first sentence is filtered if the semantic similarity between the first sentence and the first query is less than a preset threshold initially. That is, the first sentence is unrelated to the first query.

  In some embodiments of the present disclosure, the second obtaining unit 120 is configured to search the sentence library according to the first sentence and obtain a second sentence corresponding to the first sentence.

  In some embodiments of the present disclosure, after the first acquisition unit 110 acquires the first sentence, the second acquisition unit 120 searches the sentence library according to the first sentence and Get the corresponding second sentence. The first sentence related to the feature amount information may correspond to a plurality of second sentences. Therefore, the second acquisition unit 120 corresponds to the first sentence by searching the sentence library. You may acquire a 2nd sentence.

  In some embodiments of the present disclosure, the second acquisition unit 120 acquires the second sentence if the first query does not include a preset type keyword of the second sentence. In some embodiments of the present disclosure, the second sentence, if the first query includes a preset type keyword of the second sentence, such as the interrogative “what”, Filtered.

  In some embodiments of the present disclosure, the second acquisition unit 120 may determine whether the second similarity unit 120 has a semantic similarity between the first query and the second sentence that is less than a second preset threshold. Get the second sentence. In some embodiments of the present disclosure, the second sentence is when the semantic similarity between the first query and the second sentence is greater than a second preset threshold, ie, the first If the query and the second sentence belong to the same type of sentence, they are filtered.

  In some embodiments of the present disclosure, the third obtaining unit 130 is configured to obtain the intent explanation guidance sentence according to the second sentence. The third acquisition unit 130 may include clustered subunits 131 and acquisition subunits 132. In some embodiments, the second obtaining unit 120 is configured to search the sentence library according to the first sentence and obtain a plurality of second sentences corresponding to the first sentence. . The clustering subunit 131 is configured to sequence and cluster the second sentence to obtain a clustered second sentence.

In some embodiments of the present disclosure, the clustered subunit 131 sequences the second sentence according to Equation 1.
Here, Qb represents the second sentence, P _i represents the vote value of the i-th tree, and N represents the number of trees.
Formula 1 is a random forest model formula for calculating the score of Qbs (ie, the second sentence). Next, the clustered subunit 131 clusters the second sentence according to the semantic similarity. For example, the sentence “How old is a baby” and the other sentence “How old is a baby” have the same meaning and may be clustered. In some embodiments of the present disclosure, clustering may be performed by at least one of a K-means clustering algorithm (hard clustering algorithm) and a hierarchical clustering algorithm.

  In some embodiments of the present disclosure, the acquisition subunit 132 is configured to sequence the clustered second sentence and to obtain the intent description guidance sentence.

In some embodiments of the present disclosure, the acquisition subunit 132 may calculate the score of the second sentence (ie, type) clustered according to Equation 2.
Here, c represents a type, | c | represents the number of Qb in the type, and λ represents a weight value for probability evaluation of the type (0 <λ <1). In some embodiments of the present disclosure, after the score of the clustered second sentence (ie, type) is calculated, the clustered second sentence (ie, type) is sequenced according to the score. The three intent explanation guidance sentences with the highest scores are obtained and returned to the client for display to the user, so that the user updates the first query according to the intent explanation guidance sentences. It becomes possible. In the present embodiment, the three intention explanation guidance texts are provided as examples only. In some embodiments of the present disclosure, one or more intent guidance text may be obtained without particularly limiting the present disclosure.

  In some embodiments of the present disclosure, the second acquisition module 200 is configured to acquire a second query that is updated according to the intent description guidance text, and to acquire search results according to the second query. ing.

  In some embodiments of the present disclosure, the second obtaining module 200 may obtain a second query that is updated according to the intent explanation guidance text, and obtain the search result according to the second query. Good. For example, the search engine displays a guidance sentence for the intention explanation to the user through the client, and the user inputs the revised query (for example, the second query) in the input box provided by the search engine, and then the search engine updates. The search may be performed according to the query to be obtained, and the search result may be acquired.

  In some embodiments of the present disclosure, the response module 300 is configured to return search results to the client.

  In some embodiments of the present disclosure, after the second acquisition module 200 acquires the search results, the response module 300 may return the search results to the client and display the search results to the user. The user may acquire necessary resources based on the search result.

  In some embodiments of the present disclosure, the establishment module 400 is configured to establish a sentence library.

  In some embodiments of the present disclosure, the establishment module 400 may establish a sentence library before the intent explanation guidance text is obtained, and the search engine may use the sentence to obtain the intent explanation guidance text. You can search the library. In some embodiments of the present disclosure, a sentence library may be pre-established, and thus the step of establishing a sentence library may be avoided.

  In some embodiments of the present disclosure, the establishment module 400 includes a first acquisition unit 410, a second acquisition unit 420, an establishment unit 430, and a filtering unit 440.

  In some embodiments of the present disclosure, the first acquisition unit 410 is configured to acquire the first sentence according to the first information.

  In some embodiments of the present disclosure, the first acquisition unit 410 divides the first information, and the first information is converted into a sentence of a preset type according to a preset feature amount. Whether it belongs or not may be detected. When the first information belongs to a preset type of sentence, the first acquisition unit 410 acquires the first sentence according to the first information. In some embodiments of the present disclosure, the first information includes sentences in the community user's knowledge exchange platform, and the preset types of sentences include various questions such as question questions, affirmative sentences, rhetorical questions, etc. It is possible to include a sentence, and a preset feature amount may include a question dictionary or a feature amount of a question sentence.

  In some embodiments of the present disclosure, the second obtaining unit 420 is configured to obtain the second sentence according to second information related to the first information.

  In some embodiments of the present disclosure, the second acquisition unit 420 divides the second information, and the second information is converted into a sentence of a preset type according to a preset feature amount. Whether it belongs or not may be detected. If the second information belongs to a preset type of sentence, the second acquisition unit 420 may acquire the second sentence according to the second information related to the first information.

  In some embodiments of the present disclosure, the establishment unit 430 is configured to establish at least one sentence pair in the sentence library according to the first sentence and the second sentence.

  In some embodiments of the present disclosure, the establishing unit 430 may determine that at least one sentence according to the first and second sentences to establish or extend a sentence library after the first and second sentences are obtained. And a sentence pair may be stored in the sentence library.

  In some embodiments of the present disclosure, the search engine may search for resources in the community user's knowledge exchange platform. The resource includes a plurality of questions (first information) provided by the user and answers (second information) provided by other users to these questions. After determining that both the question and the answer belong to a preset type of sentence, the establishing unit 430 determines that at least one question (first sentence) and a corresponding answer (second sentence) A sentence pair consisting of may be acquired. A sentence library is thus constructed. For example, one user asks Baidu Knows what to do if the baby has a fever of 38.5 degrees Celsius, and another user says, “It depends on how old the baby is.” I will provide a. The establishment unit 430 may obtain a sentence pair consisting of this question and this answer, and store the sentence pair in the sentence library. Thus, the sentence library is established or expanded.

  In some embodiments of the present disclosure, the filtering unit 440 is configured to filter sentence pairs in a sentence library.

  In some embodiments of the present disclosure, the filtering unit 440 may filter the sentence pairs in the sentence library after the establishment unit 430 has established at least one sentence pair in the sentence library. In some embodiments of the present disclosure, the filtering unit 440 filters the second sentence in the sentence library according to preset filtering rules. For example, the first sentence “A place that is interesting to travel on October 1” and the second sentence “Where is an interesting place on October 1” have similar meanings and belong to the same question. Should be filtered.

  According to the information search device according to the embodiment of the present disclosure, the guidance sentence of the intention explanation provided by the search engine is acquired, the first query is updated according to the guidance sentence of the intention explanation, and the search result is the second query. As a result, the search intention of the user becomes clear. Furthermore, the accuracy of the search engine for identifying the user's search intention is improved, the user's requirements are met, and the user experience is improved.

  An embodiment of the third aspect of the present disclosure provides a computer readable storage medium. The computer readable storage medium includes a computer program for executing the information retrieval method according to the embodiment of the first aspect of the present disclosure, which is executed on a computer.

  One or more modules, portions or components for storing executable code that implements a particular logic function or procedure as described in any flowchart or in any other manner herein. Should be understood to include Furthermore, advantageous embodiments of the present disclosure include other implementations whose execution order is different from that shown or described, and that the execution functions in substantially simultaneous or reverse order according to the related functions. It is included. This should be understood by one skilled in the art to which embodiments of the present disclosure belong.

  The specific sequence table of executable instructions to implement the logic and / or steps described elsewhere herein or shown in the flowcharts, eg, logic functions, is specifically instruction execution. System, device or apparatus (computer based system, system including processor, or instruction execution system, device, other system capable of obtaining instructions from device, executing instructions, etc.) or instruction execution system Any computer-readable medium used in combination with the device and apparatus. As its specification, “computer readable medium” is suitable to include a storage program, a communication program, a propagation program, or a transfer program used by an instruction execution system, device or apparatus, or in combination with these. Any device may be used. More specific examples of computer readable media include, but are not limited to, electrical connections (electronic devices) having one or more wires, portable computer enclosures (magnetic devices), random access. Memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM or flash memory), fiber optic devices and portable compact disk read only memory (CDROM). In addition, the computer readable medium may be paper or other suitable medium on which the program can be printed. Because, for example, paper or other suitable media is optically scanned in other suitable ways when it is necessary to obtain the program in an electrical manner and then edited, decoded, or Because it can be processed and the program can then be stored in computer memory.

  It is understood that each part of the present disclosure can be realized by hardware, software, firmware, or a combination thereof. In the above embodiments, the steps or methods may be implemented by software or firmware stored in a memory, and may be executed by a suitable instruction execution system. For example, when implemented by hardware, the same is true in other embodiments, but these steps or methods may be implemented in a distributed logic circuit having a logic gate circuit for implementing a logic function of a data signal, suitable combinational logic. It may be realized by one or a combination of known techniques in this technical field such as an application specific integrated circuit having a gate circuit, a programmable gate array (PGA), and a field programmable gate array (FPGA).

  Those skilled in the art will understand that all or part of the above-described exemplary implementation methods of the present disclosure may be achieved by commanding the associated hardware by a program. The program may be stored on a computer readable storage medium and includes one or a combination of steps of the disclosed method embodiments executed on a computer.

  In addition, each functional cell of embodiments of the present disclosure may be integrated into a processing module, or these cells may be separate physical entities, or two or more cells may be processing modules. May be integrated. The integration module may be realized in the form of hardware or in the form of a software function module. If the integration module is implemented in the form of a software function module and sold or used as a stand-alone product, the integration module may be stored on a computer readable storage medium.

  The storage medium may be a read-only memory, a magnetic disk, a CD, or the like.

  References throughout this specification to “embodiments”, “some embodiments”, “examples”, “examples” or “some examples” are specific statements made in connection with the embodiments or examples. Is included in at least one embodiment or example of the present disclosure. Statements appearing throughout this specification are not necessarily referring to the same embodiment or example of the present disclosure. Furthermore, the particular features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

While illustrative embodiments have been shown and described, those of ordinary skill in the art will not be able to interpret the above embodiments as limiting the present disclosure, and alterations, alternatives, and modifications are the intent of the present disclosure. It will be appreciated that this can be done in embodiments without departing from the principles and scope.

Claims (15)

Receiving a first query and obtaining a guidance sentence of an intention explanation according to the first query;
Receiving a second query updated in accordance with the intent description guidance text and obtaining a search result in accordance with the second query;
Returning the search results to the client;
An information search method comprising: Further comprising establishing a sentence library prior to obtaining the intent explanation guidance sentence according to the first query,
In establishing the sentence library, in particular,
Obtaining a first sentence according to the first information;
Obtaining a second sentence according to second information related to the first information;
Establishing at least one sentence pair in the sentence library according to the first sentence and the second sentence;
The information search method according to claim 1, further comprising: Further comprising filtering the sentence pairs in the sentence library;
Filtering the sentence pairs in the sentence library, in particular,
Filtering the second sentence in the sentence library according to preset filtering rules;
The information search method according to claim 2, further comprising: Obtaining a guidance sentence for an intention explanation according to the first query includes:
Obtaining feature quantity information of the first query, and obtaining a first sentence related to the feature quantity information according to the feature quantity information;
Searching the sentence library according to the first sentence and obtaining a second sentence corresponding to the first sentence;
Obtaining a guidance sentence for the intention explanation according to the second sentence;
And / or
Obtaining a second sentence according to second information related to the first information,
Obtaining a plurality of second sentences according to the second information related to the first information;
And having
Obtaining the intent explanation guidance sentence according to the second sentence, in particular,
Sequencing and clustering the second sentence to obtain a clustered second sentence;
Sequencing the clustered second sentence to obtain the intent description guidance sentence;
The information search method according to claim 2 or 3. Acquiring the first sentence according to the first information is
Dividing the first information;
Detecting whether the first information belongs to a preset type of sentence according to a preset language model;
Obtaining the first sentence in accordance with the first information when it is determined that the first information belongs to a preset type of sentence;
And / or
Obtaining a second sentence according to second information related to the first information,
Dividing the second information;
Detecting whether the second information belongs to a sentence of a preset type according to a preset feature amount;
When the second information is determined to belong to a preset type of sentence, obtaining the second sentence according to the second information;
The information search method according to any one of claims 2 to 4, further comprising: Sequencing and clustering the second sentence to obtain a clustered second sentence comprises:
If Qb represents the second sentence, P _i represents the vote value of the i th tree, and N represents the number of trees,
Sequencing the second sentence according to and clustering the second sentence according to semantic similarity,
The information search method according to claim 4 or 5. Acquiring a first sentence related to the feature amount information according to the feature amount information,
Obtaining the first sentence when it is determined that the semantic similarity between the first sentence and the first query is greater than a first preset threshold, or
Obtaining a second sentence corresponding to the first sentence;
Obtaining the second sentence when it is determined that the first query does not include a preset type keyword of the second sentence, or
Obtaining the second sentence when determining that the semantic similarity between the first query and the second sentence is less than a second preset threshold,
The information search method according to any one of claims 4 to 6. A first obtaining module configured to receive a first query and obtain a guidance sentence of an intention explanation according to the first query;
A second acquisition module configured to receive a second query updated according to the intent explanation guidance text and to acquire a search result according to the second query;
A reply module configured to reply the search results;
An information retrieval device comprising: In addition, an establishment module configured to establish a sentence library is provided,
In particular, the establishment module comprises:
A first acquisition unit configured to acquire a first sentence according to first information;
A second acquisition unit configured to acquire a second sentence according to second information related to the first information;
An establishing unit configured to establish at least one sentence pair in the sentence library according to the first sentence and the second sentence;
The information search device according to claim 8. The establishment module is
A filtering unit configured to filter the sentence pairs in the sentence library;
In particular, the filtering unit is configured to filter the second sentence in the sentence library according to preset filtering rules. The first acquisition module is
A first acquisition unit configured to acquire feature amount information of the first query and to acquire a first sentence related to the feature amount information according to the feature amount information;
A second acquisition unit configured to search the sentence library according to the first sentence and to obtain a second sentence corresponding to the first sentence;
A third acquisition unit configured to acquire a guidance sentence of the intention explanation according to the second sentence;
And / or the second acquisition unit is configured to acquire a plurality of second sentences according to second information related to the first information,
The third acquisition unit is
A clustering subunit configured to sequence and cluster the second sentence to obtain a clustered second sentence;
An acquisition subunit configured to sequence the clustered second sentence to obtain the intent description guidance sentence;
The information search device according to claim 9 or 10. The first acquisition unit divides the first information, detects whether the first information belongs to a sentence of a preset type according to a preset language model, and the first information unit Is configured to acquire the first sentence according to the first information when it is determined that the information belongs to the preset type of sentence, and / or the second acquisition unit includes the Dividing the second information, detecting whether the second information belongs to a sentence of a preset type according to a preset feature amount, and the second information is of a preset type Configured to obtain the second sentence according to the second information when determined to belong to a sentence;
The information search device according to any one of claims 9 to 11. The clustering subunit includes: Qb represents the second sentence, P _i represents the vote value of the i-th tree, and N represents the number of trees.
Configured to sequence the second sentence according to and cluster the second sentence according to semantic similarity,
The information search device according to claim 11 or 12. The first acquisition unit acquires the first sentence when it is determined that the semantic similarity between the first sentence and the first query is initially greater than a preset threshold. Configured to, or
The second acquisition unit is configured to acquire the second sentence when it is determined that the first query does not include a preset type keyword of the second sentence; Or
The second acquisition unit determines the second sentence when the semantic similarity between the first query and the second sentence is determined to be smaller than a second preset threshold. 14. An information retrieval device according to any one of claims 11 to 13, configured to obtain Receiving a first query and obtaining a guidance sentence of an intention explanation according to the first query;
Receiving a second query updated in accordance with the intent explanation guidance text and obtaining a search result in accordance with the second query;
A computer-readable storage medium comprising a computer program for executing on the computer the step of returning the search result to the client.