CN106611058A - Test question searching method and device - Google Patents

Abstract

The invention provides a method and a device for searching test questions, comprising the following steps: acquiring an original image of a target test question, and carrying out image recognition on the original image of the target test question; carrying out full text retrieval in a question bank based on the result of image identification to obtain a test question searching result; when the number of the obtained test question search results is more than two, calculating first scores corresponding to the obtained test question search results according to a scoring mechanism of full-text retrieval; calculating second scores corresponding to the obtained test question search results respectively according to a similarity calculation method; and according to a preset weighted linear scheme, carrying out weighted calculation on the first score and the second score, determining a final score, and sequencing and outputting the test question search results from high to low according to the final score. By the scheme of the invention, the accuracy of the test question search is improved.

Description

Method and device for searching test questions

technical field

The invention relates to the field of communication technology, in particular to a test question search method and device.

Background technique

At present, the education industry is also in line with the Internet, and many online education products have appeared, including question-searching products with functions such as taking pictures and answering questions. Question-searching products are designed for student users to obtain images containing questions and perform image recognition on the images when they encounter difficulties in homework. Based on the results of image recognition, they can search for questions and answer analysis that users need in the background question bank.

However, because the existing image recognition technology will be affected by the ambient light and lighting of the shooting, it often leads to the inability to accurately search for the test questions and answer analysis required by the user when using the search question product, and cannot satisfy the user's search question Demand affects the user experience of this type of product.

Contents of the invention

The invention provides a test question search method and device, aiming at improving the accuracy of test question search.

According to the first aspect of the embodiments of the present invention, a method for searching for test questions is provided, and the method for searching for test questions includes:

Obtaining the original image of the target test question, and performing image recognition on the original image of the target test question;

Based on the results of image recognition, perform a full-text search in the question bank to obtain the search results of the test questions;

When the number of obtained test question search results is more than two, according to the scoring mechanism of the full-text search, calculate the first score corresponding to each obtained test question search result;

According to the similarity algorithm, calculate the second score corresponding to the obtained search results of each test question;

According to the preset weighted linear scheme, the first score and the second score are weighted and calculated to determine the final score, and according to the final score, the search results of the test questions are sorted from high to low and output.

According to the second aspect of the embodiments of the present invention, a test question search device is provided, and the test question search device includes:

A target test item acquisition unit, configured to acquire the original image of the target test item, and perform image recognition on the original image of the target test item;

A preliminary retrieval unit, configured to perform a full-text search in the question bank based on the image recognition results obtained by the target test question acquisition unit to obtain test question search results;

The first score calculation unit is used to calculate the first scores corresponding to each of the obtained test question search results according to the scoring mechanism of the full-text search when the number of test question search results obtained by the preliminary retrieval unit is more than two;

The second score calculation unit is used to calculate the second scores corresponding to the search results of each test question obtained by the preliminary retrieval unit according to the similarity algorithm;

A search result determining unit, configured to perform weighted calculations on the first score obtained by the first score calculation unit and the second score obtained by the second score calculation unit according to a preset weighted linear scheme, to determine a final score, and according to The final score is to sort and output the test question search results from high to low.

As can be seen from the above, in the solution of the present invention, first obtain the original image of the target test question, and perform image recognition on the original image of the target test question, then perform a full-text search based on the image recognition result in the question bank to obtain the test question search result, when When the test question search results contain more than two results, according to the scoring mechanism of the full-text search, the first score corresponding to each test question search result is calculated, and the second score corresponding to each test question search result is calculated according to the similarity algorithm. A weighted linear scheme is set up to perform weighted calculation on the first score and the second score to determine the final score, and according to the final score, the search results of the test questions are sorted from high to low and output. The results of the test question search can be further screened, and finally more accurate, higher matching test questions and corresponding analysis can be screened out. Compared with the prior art, the inaccuracy of image recognition affects the final search results of test questions, resulting in the inability to search for matching test questions. The solution of the present invention improves the accuracy of test question search and better satisfies the needs of users. search questions, improving the user experience.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only These are some embodiments of the present invention. For those skilled in the art, other drawings can also be obtained according to these drawings without any creative effort.

Fig. 1 is the implementation flowchart of the test question search method provided by Embodiment 1 of the present invention;

Fig. 2 is the specific implementation flowchart of step S102 of the test question search method provided by Embodiment 1 of the present invention;

FIG. 3 is a structural block diagram of a test question search device provided in Embodiment 2 of the present invention.

detailed description

In order to make the purpose, features and advantages of the present invention more obvious and understandable, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described The embodiments are only some of the embodiments of the present invention, but not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

In the embodiment of the present invention, first obtain the original image of the target test question, and perform image recognition on the original image of the target test question, and then perform a full-text search in the question bank based on the image recognition result to obtain the test question search result, when the test question search result When there are more than two results, according to the scoring mechanism of full-text retrieval, calculate the first score corresponding to each test question search result, and calculate the second score corresponding to each test question search result according to the similarity algorithm, and finally according to the preset weighted linear The solution is to perform weighted calculation on the first score and the second score, determine the final score, and sort and output the search results of the test questions from high to low according to the final score. The results of the test question search can be further screened, and finally more accurate, higher matching test questions and corresponding analysis can be screened out.

The realization of the present invention is described in detail below in conjunction with specific embodiment:

Embodiment one

Fig. 1 shows the implementation process of the test question search method provided by Embodiment 1 of the present invention, which is described in detail as follows:

In step S101, the original image of the target test question is acquired, and image recognition is performed on the original image of the target test question.

In the embodiment of the present invention, the original image containing the target test question may be acquired by means of camera shooting, for example, after the camera function is activated, the camera may take a picture of the target test question. Alternatively, in step S101, the original image containing the target test question may also be actively or passively obtained from other devices; or, in step S101, the original image containing the target test question may also be obtained from an image library, which is not limited here. After the original image of the target test question is obtained, the text recognition is performed on the original image of the target test question, so as to realize the convenient and rapid acquisition of the original image containing the target test question in various ways. Specifically, in step S101, optical character recognition (OCR, Optical Character Recognition) technology may be used to perform text recognition on the acquired original image containing the target test question. Of course, step S104 may also use other text recognition technologies to perform text recognition on the acquired original image containing the target test question, which is not limited here.

In step S102, a full-text search is performed in the question bank based on the image recognition result to obtain test question search results.

In the embodiment of the present invention, after receiving the original image containing the target test question and performing image recognition on the original image, the test question search device will perform a full-text search in the question bank based on the result of image recognition to obtain the target test question. Related search results. Wherein, the question bank includes a local question bank, and also includes a question bank on the Internet. The local question bank can be the built-in question bank in the test question search device, or it can be the question bank downloaded to the test question search device by the user via the Internet, so that the user can download the required question bank in a networked environment, and can access and download it in an offline environment. The question bank is not limited here. Since the full-text search will retrieve all the results obtained from the image recognition in step S101, the obtained search results will have a high degree of relevance and matching.

Specifically, when performing a full-text search in the question bank based on the image recognition result, the Lucene framework can be used to perform a full-text search on the image recognition result. The development language used by Lucene is Java, which is an open source full-text search engine toolkit. It is not a complete full-text search engine, but a full-text search engine architecture, which provides a complete query engine, index engine and partial text analysis engine. Lucene provides developers with an easy-to-use toolkit to facilitate the establishment of a complete full-text search engine in the test question search device, so that users can realize the full-text search function based on this. Of course, in step S102, other retrieval frameworks or retrieval programs may also be used to perform full-text retrieval of the image recognition results, such as Galago, Xapian, or Zebra, etc., which are not limited here.

In step S103, when the number of obtained test question search results is more than two, according to the scoring mechanism of the full-text search, calculate the first scores corresponding to each of the obtained entity search results.

In the embodiment of the present invention, the number of test question search results finally obtained in step S102 cannot be determined. When the target test question searched by the user is relatively new, in step S102 there may be few or no search results for the test question retrieved. Depending on the number of search results obtained for test questions, there are the following application scenarios:

In an application scenario, step S102 does not obtain any search results of test questions matching the target test question. At this time, step S103 may directly return the information that there is no search result, so as to inform the user that a question matching the target test question cannot be found in the existing question bank.

In another application scenario, step S102 only obtains one test question search result. At this time, step S103 directly returns the acquired search result of the above-mentioned test question to the user for the user to check.

Optionally, in the above two application scenarios, it is also possible to display on the screen which question banks the test question search device has searched in step S102, and recommend other question banks that the user can try to search for, and the user can choose Download the question bank to search again or search online again to get more test question search results.

In the third application scenario, step S102 obtains more than two test question search results. At this point, since there are multiple test question search results, step S103 may first perform the first scoring screening on them. The full-text retrieval technology provides a scoring mechanism that can score the search results. The higher the score of the search result, the higher its relevance to the target test questions. In step S103, each search result obtained in step S102 is scored once, and the obtained score is the first score corresponding to each search result, which is recorded as S1. The above-mentioned first score S1 will be temporarily stored for subsequent further calculation using the above-mentioned first score S1.

Specifically, when the Lucene framework is used to perform full-text search on the results of image recognition in step S102, for the above-mentioned third application scenario, step S103 may be to perform Lucene scores on the obtained search results of each test question to obtain the above-mentioned The first score corresponding to the test question search results respectively. When performing full-text search based on the Lucene framework, it can score each search result of the test question. Among them, the maximum score of the Lucene score corresponding to any test question search result is 1.0.

In step S104, according to the similarity algorithm, the second scores corresponding to the acquired search results of each test question are calculated.

In the embodiment of the present invention, there may be three application scenarios in step S103, and step S104 is for the third application scenario, that is, a step that will be executed only when the number of search questions obtained in step S102 is more than two . In step S104, according to the similarity algorithm, each search result obtained in step S102 will be compared with the target test questions for similarity, and scoring will be performed again, and the obtained score will be the second score corresponding to each search result. It is denoted as S2. The above-mentioned second score S2 will be temporarily stored for further calculation using the above-mentioned second score S2.

Specifically, the above similarity algorithm may be the longest common subsequence algorithm. Then, at this time, step S104 is embodied as, according to the longest common subsequence algorithm, scoring each obtained search result of the test question respectively, and obtaining the second score corresponding to each search result of the test question. The longest common subsequence, English abbreviation is LCS (longest CommonSubsequence), it can describe the similarity between two paragraphs of texts, so in step S104, use the longest common subsequence algorithm, can effectively search for each test item search result and The similarity between the target test questions is calculated, so as to further analyze the search results of the test questions. Optionally, the test question search device may perform secondary ranking on the test question search results from high to low according to the scores of the second score S2.

In step S105, according to the preset weighted linear scheme, the above-mentioned first score and the second score are weighted and calculated to determine the final score, and according to the above-mentioned final score, the search results of the above two or more test questions are sorted from high to low and output.

In the embodiment of the present invention, when in the above-mentioned third application scenario, that is, when the number of test question search results obtained in step S102 is more than two, the test question search device in step S105 will search for each test question search result, Obtain the corresponding first score S1 and second score S2, and perform weighted calculations on the first score S1 and second score S2 corresponding to each test question search result according to the preset weighted linear scheme, and obtain the final score corresponding to the test question search result Score, record it as S3, and sort the corresponding test question search results from high to low according to the final score S3, and output the sorted results for users to review.

Further, the above preset weighted linear scheme may be: S3=n1*S1+n2*S2. Wherein, n1>0, n2>0, n1+n2=1. Specifically, n1=0.3 and n2=0.7 can be set. In this case, the final score S3 and the sorting of the corresponding search results of the test questions are more in line with the search questions of most users.

It can be seen from the above that in the embodiment of the present invention, the search results of the test questions can be scored twice, and the two scores can be weighted to obtain the final score. Since the final score is a weighted process of the previous two scores, the The accuracy of the search results of the final test questions will be improved, which can well solve the problem that when users take pictures of the target test questions and search for them when the light is not strong or are disturbed by other external environments, the search results of the test questions will be affected, resulting in Inaccurate on this question. It improves the accuracy of the user's search questions, improves the user's operating experience, and better meets the user's search questions.

Fig. 2 shows the specific implementation flowchart of step S102 of the test question search method provided by the embodiment of the present invention, which is described in detail as follows:

In step S201, a full-text search is performed in the question bank according to the image recognition result to obtain all relevant test question search results.

In step S202, when the number of all the above-mentioned relevant test question search results is more than two, perform multi-dimensional statistics on all the above-mentioned relevant test question search results according to the preset data model, and determine the test question category containing the largest number of test question search results .

In this embodiment of the present invention, the number of all related test question search results obtained in step S201 is uncertain. Only when the number of all relevant test question search results obtained in step S201 is more than two, step S202 will perform multi-dimensional statistics on all the above-mentioned relevant test question search results according to the preset data filtering model, so as to filter out the most Search results for test questions in the best category. Specifically, the above-mentioned dimensions include but are not limited to more than one of the following: time, region, subject, and grade.

Among them, the time is the proposition year corresponding to the test question. For student users, especially those in middle and high school, the questions they face are often real exam questions or simulated exam questions from previous years, and the year information of these questions is often kept in the question bank. At this time, in step S202, the test question search device can count the time distribution of each test question result in the test question search results related to the target test question, and filter out the year that contains the most test question search results, and obtain all the results of the test question under the year. Exam search results.

Correspondingly, the region is the proposition area corresponding to the test question. For middle and high school student user groups, region is also an important factor when they search for test questions, because middle and high school test questions often have strong regional characteristics. At this time, in step S202, when the user searches for a test question, the test question search device can take the region as a dimension of statistics, count the geographical distribution of the results of each test question, and filter out the region that contains the most test question search results, and obtain the belonging region. Search results for all test questions under the region.

At the same time, subject is also an important dimension. Each test question has its own subject, but because in real life, knowledge is crossed, so some knowledge points may appear in multiple subjects. Taking science as an example, subjects such as physics, chemistry, biology, and mathematics will all intersect with each other. In particular, mathematics, as a basic subject of science, will be involved in physics, chemistry, and biology. For example, when a user searches for an application problem in mathematics as the target test question, if the application problem contains a physical background, and the topic is to calculate the displacement, velocity, distance or other physical quantities of an object, if the above target test question In the full-text search in the question bank, it is very likely that not only the test questions under the subject of mathematics can be retrieved, but also the questions containing physics or other subjects can be retrieved. At this time, in step S202, the subject is set as one of the statistical dimensions. After performing subject statistics on all the searched test question results and determining the number of test question search results belonging to different subjects, it is possible to effectively filter out the results that are related to the target test question. Search results for test questions in the same subject.

As for grades, in real life, for a certain knowledge point, the number of questions is most concentrated in the grade where this knowledge point has been learned, while other grades will not focus on this knowledge point. Therefore, the test question search device is also equipped with statistics on the grades of the questions, which can more accurately filter out test question search results that are in the same grade as the target test question, avoiding the occurrence of test question search results that exceed the outline or other situations that do not meet user expectations.

Optionally, the user can choose which dimension or dimensions to use for statistics.

Optionally, after performing multi-dimensional statistics, the user can choose the test question category that the user is interested in or needs from among the different test question categories contained in different dimensions, instead of using the test question search device to automatically select the test questions that contain the largest number of search results category of test questions.

In step S203 , under the test question category containing the largest number of test question search results, keep the test question search results with high relevance.

In the embodiment of the present invention, since step S202 can obtain a test question category with the largest number of test question search results in a certain dimension, the number of test question search results under this test question category is uncertain, and we only need to keep the same The test question search results with high relevance to the target test question can be filtered for the test question search results that are related to the target test question but not highly relevant.

Specifically, when the number of test question search results under the above test question category is more than N, the test question search results under the above test question category are sorted by relevance, and the top N test question search results with high relevance are retained. The above-mentioned test question category is the test question category determined in step S202 that contains the largest number of test question search results. When sorting the above test item categories, you can use any of the following algorithms to calculate the correlation: TF-IDF (Term Frequency-Inverse Document Frequency) or Okapi BM25 (Best Match 25). According to the score calculated by the above correlation, sort the search results of the test questions from high to low, and filter based on the above correlation score, only keep the top N test search results with high correlation scores after sorting, and do not Considering the test question category, step S201 is eliminated to obtain all other test question search results.

If the number of test question search results under the above test question category is not more than N, all the test question search results under the above test question category will be retained. At this time, since the number of test question search results under the above test question categories does not exceed N, no matter how high or low the correlation between the test question search results and the target test question is, they will be retained, and there is no need for further screening based on the correlation under the above test question categories. However, search results for test questions under other test question categories will all be excluded.

Wherein, the aforementioned N is a preset natural number greater than or equal to 2. In step S203, N may be initially set by the test question search device, or may be set by the user. Optionally, the test question search device will initially set N to be 100.

It can be seen from the above that in this embodiment, before scoring the results of the test questions, the results of the test questions can be first roughly screened, so as to avoid the subsequent first scoring and second scoring of all the test question search results. scoring and final scoring. It greatly saves the occupation of system resources by the test question search method.

It should be noted that the test question search device mentioned in the embodiment of the present invention can specifically be integrated in a mobile terminal (such as a smart phone, a tablet computer, a learning machine, etc.) in the form of software (such as an App) and/or hardware. )middle.

Those of ordinary skill in the art can understand that all or part of the steps in the method of the above-mentioned embodiments can be completed by instructing related hardware through a program, and the corresponding program can be stored in a computer-readable storage medium. The above-mentioned storage medium, Such as ROM/RAM, disk or CD, etc.

Embodiment two

FIG. 3 shows a specific structural block diagram of the test question search device provided by the second embodiment of the present invention. For the convenience of description, only the parts related to the embodiment of the present invention are shown. This test question searching device 3 comprises:

A target test question acquisition unit 31, configured to acquire the original image of the target test question, and perform image recognition on the original image of the target test question;

The preliminary retrieval unit 32 is used to perform a full-text search in the question bank based on the image recognition results obtained by the above-mentioned target test question acquisition unit 31, and obtain the test question search results;

The first score calculation unit 33 is used to calculate the first scores corresponding to the obtained test question search results respectively according to the scoring mechanism of the full-text search when the quantity of the test question search results obtained by the above-mentioned preliminary retrieval unit 32 is more than two;

The second score calculation unit 34 is used to calculate the second score corresponding to each test question search result obtained by the above-mentioned preliminary retrieval unit 32 according to the similarity algorithm;

The search result determination unit 35 is configured to perform weighted calculations on the first score obtained by the first score calculation unit 33 and the second score obtained by the second score calculation unit 34 according to a preset weighted linear scheme to determine a final score, and According to the above final scores, the search results of the above test questions are sorted from high to low and output.

Specifically, the above-mentioned preliminary retrieval unit 32 is used to perform full-text retrieval on the results of image recognition using the Lucene framework;

Specifically, the above-mentioned first score determination unit 33 is configured to perform Lucene scoring on each test question search result obtained by the above-mentioned preliminary retrieval unit 32 to obtain the first score corresponding to each test question search result.

Specifically, the above-mentioned second score determination unit 34 is configured to, when the similarity algorithm used is the longest common subsequence algorithm, according to the longest common subsequence algorithm, search results of each test question obtained by the above-mentioned preliminary retrieval unit 32 respectively Scoring is performed to obtain the second scoring corresponding to the search results of the above test questions.

Optionally, the above-mentioned preliminary retrieval unit 32 also includes:

The search result acquisition subunit is used to perform a full-text search in the question bank according to the image recognition result obtained by the above-mentioned target test question acquisition unit 31, and obtain all relevant test question search results;

The multi-dimensional statistical subunit is used to analyze all relevant test questions obtained by the above search result acquisition subunit according to the preset data model when the number of all relevant test question search results obtained by the above search result acquisition subunit is more than two Perform multi-dimensional statistics on the search results of test questions to determine the category of test questions that contains the largest number of search results for test questions;

The result screening subunit is used for retaining highly relevant test question search results under the test question category determined by the above-mentioned multidimensional statistics subunit that contains the largest number of test question search results.

Specifically, the above-mentioned result screening subunit is used for, if the number of test question search results under the test question category determined by the above-mentioned multi-dimensional statistics subunit is more than N, then sort the test question search results under the above test question category by relevance, and keep the relevant The top N test question search results with high degree; if the number of test question search results under the test question category determined by the above-mentioned multi-dimensional statistics sub-unit is not more than N, all the test question search results under the above test question category will be kept; Let it be a natural number greater than or equal to 2.

It should be noted that the test question search device in the embodiment of the present invention can specifically be integrated in a mobile terminal (such as a smart phone, a tablet computer, a learning machine, etc.) in the form of software (such as an App) and/or hardware. .

It should be understood that the test question search device in the embodiment of the present invention can be used to realize all the technical solutions in the above-mentioned method embodiments, and the functions of each functional module can be specifically realized according to the methods in the above-mentioned method embodiments, and the specific implementation process can refer to Relevant descriptions in the foregoing embodiments will not be repeated here.

It can be seen from the above that in the embodiment of the present invention, the test question search device can first perform a rough preliminary screening on the test question results before scoring the test question results, so as to avoid subsequent first-time screening of all test question search results. Scoring, Second Scoring and Final Scoring. It greatly saves the occupation of system resources by the test question search device.

It should be noted that, in the several embodiments provided in this application, it should be understood that the disclosed devices and methods may be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of the above units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or can be Integrate into another system, or some features may be ignored, or not implemented. In another point, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces, and the indirect coupling or communication connection of devices or units may be in electrical, mechanical or other forms.

For the foregoing method embodiments, for the sake of simplicity of description, they are expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, Certain steps may be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification belong to preferred embodiments, and the actions and modules involved are not necessarily required by the present invention.

In the foregoing embodiments, the descriptions of each embodiment have their own emphases, and for parts not described in detail in a certain embodiment, reference may be made to relevant descriptions of other embodiments.

The above is just a preferred embodiment provided by the present invention. For those of ordinary skill in the art, according to the idea of the embodiment of the present invention, there will be changes in the specific implementation and application range. In summary, this The content of the description should not be construed as limiting the present invention.

Claims (10)

1. a kind of examination question searching method, it is characterised in that include：

The original image of target examination question is obtained, and the original image to the target examination question carries out image recognition；

Result based on image recognition carries out full-text search in exam pool, obtains examination question Search Results；

When the quantity of the examination question Search Results for obtaining is two or more, according to the scoring of full-text search, calculate what is obtained Corresponding first scoring of each examination question Search Results difference；

According to similarity algorithm, corresponding second scoring of each examination question Search Results difference for obtaining is calculated；

According to default weighted linear scheme, the described first scoring and the second scoring are weighted, it is determined that final scoring, and According to the final scoring, from high to low examination question Search Results are ranked up and are exported.

2. examination question searching method as claimed in claim 1, it is characterised in that the result based on image recognition is in exam pool Full-text search is carried out, examination question Search Results are obtained, including：

Full-text search is carried out in exam pool according to the result of image recognition, all related examination question Search Results are obtained；

When the quantity of all related examination question Search Results is two or more, according to default data model to the institute The examination question Search Results for having correlation carry out various dimensions statistics, it is determined that comprising the most examination question classification of examination question Search Results quantity；

Under the examination question classification most comprising examination question Search Results quantity, retain the high examination question Search Results of degree of association.

3. examination question searching method as claimed in claim 2, it is characterised in that it is described described comprising examination question Search Results quantity Under most examination question classifications, retain the high examination question Search Results of degree of association, including：

If the examination question Search Results quantity under the examination question classification is more than N number of, to the examination question search knot under the examination question classification Fruit carries out relevancy ranking, retains the high top n examination question Search Results of degree of association；

If the examination question Search Results quantity under the examination question classification is not more than N number of, retain all examinations under the examination question classification Topic Search Results；

N be it is default be more than or equal to 2 natural number.

4. the examination question searching method as described in any one of claim 1-3, it is characterised in that the result based on image recognition Full-text search is carried out in exam pool, including：

Full-text search is carried out to the result of image recognition using Lucene frameworks；

It is described when obtain examination question Search Results quantity be two or more when, according to the scoring of full-text search, calculating is obtained Corresponding first scoring of each examination question Search Results difference for taking, including：

Each examination question Search Results to obtaining carry out respectively Lucene scorings, obtain described each examination question Search Results right respectively The first scoring answered.

5. the examination question searching method as described in any one of claim 1-3, it is characterised in that the similarity algorithm is most long public affairs Common subsequence algorithm；It is described that corresponding second scoring of each examination question Search Results difference for obtaining is calculated based on similarity algorithm, Including：

According to longest common subsequence algorithm, each examination question Search Results to obtaining score respectively, obtain it is described each Corresponding second scoring of examination question Search Results difference.

6. a kind of examination question searcher, it is characterised in that the examination question searcher includes：

Target examination question acquiring unit, for obtaining the original image of target examination question, and the original image to the target examination question enters Row image recognition；

Preliminary search unit, the result for being obtained image recognition based on the target examination question acquiring unit is carried out entirely in exam pool Text retrieval, obtains examination question Search Results；

First score calculation unit, the quantity of the examination question Search Results for getting when the preliminary search unit be two with When upper, according to the scoring of full-text search, corresponding first scoring of each examination question Search Results difference for obtaining was calculated；

Second score calculation unit, for according to similarity algorithm, calculating each examination question that the preliminary search unit gets Corresponding second scoring of Search Results difference；

Search result determination unit, for according to default weighted linear scheme, the first score calculation unit is obtained the The second scoring that one scoring and the second score calculation unit are obtained is weighted, it is determined that final scoring, and according to institute Final scoring is stated, from high to low the examination question Search Results is ranked up and is exported.

7. a kind of examination question searcher as claimed in claim 6, it is characterised in that the preliminary search unit, including：

Search Results obtain subelement, for the image recognition result that obtained according to the target examination question acquiring unit in exam pool Full-text search is carried out, all related examination question Search Results are obtained；

Various dimensions count subelement, for obtaining all related examination question search knot that subelement gets when the Search Results When the quantity of fruit is two or more, all phases that subelement gets are obtained to the Search Results according to default data model The examination question Search Results of pass carry out various dimensions statistics, it is determined that comprising the most examination question classification of examination question Search Results quantity；

As a result subelement is screened, for counting the most comprising examination question Search Results quantity of subelement determination in the various dimensions Under examination question classification, retain the high examination question Search Results of degree of association.

8. a kind of examination question searcher as claimed in claim 7, it is characterised in that the result screens subelement, concrete to use In when the examination question Search Results quantity under the various dimensions count the examination question classification that subelement determines is more than N number of, to the examination Examination question Search Results under topic classification carry out relevancy ranking, retain the high top n examination question Search Results of degree of association；

When the examination question Search Results quantity under the various dimensions count the examination question classification that subelement determines is not more than N number of, retain All examination question Search Results under the examination question classification；

N be it is default be more than or equal to 2 natural number.

9. the examination question searcher as described in any one of claim 6-8, it is characterised in that the preliminary search unit is specifically used In carrying out full-text search to the result of image recognition using Lucene frameworks；

The first score calculation unit is specifically for each examination question Search Results got to the preliminary search unit point Lucene scorings are not carried out, corresponding first scoring of described each examination question Search Results difference is obtained.

10. the examination question searcher as described in any one of claim 6-8, it is characterised in that the second score calculation unit Specifically for when the similarity algorithm for using is longest common subsequence algorithm, according to longest common subsequence algorithm, to institute State each examination question Search Results that preliminary search unit gets to be scored respectively, obtain described each examination question Search Results point Not corresponding second scoring.