CN102375839A - Method and device for acquiring target data set from candidate data set, and translation machine - Google Patents

Abstract

The present invention relates to a method and apparatus for obtaining a target data set from a candidate data set and a translation machine. Wherein, features are extracted from target data samples; target data is extracted from the candidate data set by using the features to form a target data set. According to the embodiment of the present invention, the target data can be extracted from the candidate data set according to the provided samples.

Description

Method and device for obtaining target data set from candidate data set and translation machine

technical field

The present application relates to data extraction, in particular, to a method and device for obtaining a target data set from a candidate data set. In addition, the present application also relates to a translation machine.

Background technique

Traditionally, to obtain specific target data from candidate data sets based on specific target data samples is usually manually selected by judging the similarity between the data in the candidate data set and the target data samples, or even randomly selecting some from the candidate data class. data as target data. Obviously, this traditional way and method cannot provide high-quality target data.

Contents of the invention

In the following, a brief overview of the present invention is given in order to provide a basic understanding of some aspects of the invention. It should be understood that this summary is not an exhaustive overview of the invention. It is not intended to identify key or critical parts of the invention nor to delineate the scope of the invention. Its purpose is only to present some concepts in a simplified form as a prelude to the more detailed description that is discussed later.

According to an embodiment of the present application, features are extracted from target data samples, and target data is extracted from the candidate data set by using the features, thereby forming a target data set.

In this way, the target data set is formed by subsetting the candidate data set based on the target data samples for the specific purpose. The resulting target data set is more rapidly generated. In addition, the formed target data set is more in line with the requirements of subsequent processing.

Description of drawings

The present invention can be better understood by referring to the following description given in conjunction with the accompanying drawings, wherein the same or similar reference numerals are used throughout to designate the same or similar parts. The accompanying drawings, together with the following detailed description, are incorporated in and form a part of this specification, and serve to further illustrate preferred embodiments of the invention and explain the principles and advantages of the invention. In the attached picture:

FIG. 1 shows a flowchart of a method for obtaining a target data set from a candidate data set according to an embodiment of the present invention,

Fig. 2 shows a flow chart of a method for obtaining a target data set from a candidate data set according to another embodiment of the present invention,

Fig. 3 shows a flow chart of a method for obtaining a target data set from a candidate data set according to another embodiment of the present invention,

Fig. 4 shows a flow chart of a method for obtaining a target data set from a candidate data set according to another embodiment of the present invention,

Fig. 5 shows a schematic structural diagram of a device for obtaining a target data set from a candidate data set according to an embodiment of the present invention,

Fig. 6 shows a schematic structural diagram of an extraction unit of an apparatus for obtaining a target data set from a candidate data set according to an embodiment of the present invention,

Fig. 7 shows a schematic structural diagram of an extraction unit of an apparatus for obtaining a target data set from a candidate data set according to another embodiment of the present invention,

Fig. 8 shows a schematic structural diagram of an extraction unit of an apparatus for obtaining a target data set from a candidate data set according to another embodiment of the present invention, and

Fig. 9 shows a schematic block diagram illustrating a computer that may be used to implement embodiments according to the present invention.

Detailed ways

Exemplary embodiments of the present invention will be described below with reference to the accompanying drawings. In the interest of clarity and conciseness, not all features of an actual implementation are described in this specification. However, it should be understood that many implementation-specific decisions can be made during the development of any such actual implementation in order to achieve the developer's specific goals, and that these decisions may vary from implementation to implementation .

Here, it should also be noted that, in order to avoid obscuring the present invention due to unnecessary details, only the device structure closely related to the solution according to the present invention is shown in the drawings, and the relationship with the present invention is omitted. Little other details.

first embodiment

Fig. 1 shows a flowchart of a method for obtaining a target data set from a candidate data set according to an embodiment of the present application. In order to obtain the target data set from the candidate data set, features are extracted from the target data samples in S110. The target data sample may include one or more data, where the data includes data elements. Data can include: strings, sentences, or collections of pictures. Correspondingly, the data elements may be characters, words or pictures. Apparently, the feature may be any feature, and as a non-limiting example, the extracted feature may consist of at least a part of the data elements. For example, if the target data sample is a sentence, its data elements are words constituting the sentence, and the extracted feature is at least one word in the sentence. When multiple features are extracted from the target data sample, the weight of each feature is determined based on the frequency of occurrence of the feature in the target data sample. Select high-weight features as the features of the target data sample. In other words, the more frequently a feature occurs in the target data sample, the higher its weight.

In S120, the feature is used to extract target data from the candidate data set to form a target data set. A candidate dataset may be a dataset covering the target dataset and may include pictures, text, or corpus, among others. The target data set to be formed is a data set specific to the target data sample. By means of S120, according to the characteristics of the target data samples extracted in S110, candidate data with the characteristics are found in the candidate data set to form the target data set. Here, the candidate data can also include data elements, and the data elements can also be characters, words or pictures.

Through this method, for example, relevant information (here, target data samples) can be extracted from a large information base or information collection (here, candidate data sets) according to the information of the user's interested field (here, target data samples). ), thus forming user-customized information (here, the target data set). For example, a user is interested in information in the computer field, so he can use an article related to the computer field as a sample, search for information related to this field from a large number of information on the Internet, and select information closely related to the computer field Provided to this user as the target dataset.

second embodiment

As can be seen from FIG. 2 , this embodiment is a modification of the embodiment shown in FIG. 1 . It should be pointed out that, for the sake of brevity, the description of the part with the same action and function as in FIG. 1 is omitted.

In S110, features are extracted from target data samples. S110 in this embodiment is the same as S110 in the embodiment shown in FIG. 1 , so details are not repeated here.

In S130, the feature is used to query the candidate data set. The candidate data of the candidate data set is queried by using the feature of the target data sample extracted in S110 as an information retrieval keyword. Here, keywords may be characters, words or pictures. In other words, the extracted feature is used to compare the candidate data in the candidate data set to find the candidate data with the feature.

In S140, the target data is obtained according to the similarity between the queried candidate data and the target data sample. Here, the similarity can be determined according to how many features are included in the queried candidate data. That is to say, the more features the query candidate data contains, the more similar the candidate data is to the target sample, and vice versa. Of course, the frequency of features appearing in candidate data can also be used as the basis for similarity evaluation. For example, among multiple candidate data with the feature, the higher the frequency of occurrence of the feature in the candidate data, the higher the similarity, and vice versa. Of course, multiple evaluation methods can also be used to comprehensively evaluate the similarity. In addition, similarity can also be determined with the help of similarity scores obtained by information retrieval methods.

third embodiment

As can be seen from FIG. 3 , this embodiment is an improvement over the embodiment shown in FIG. 1 . It should be pointed out that, for the sake of brevity, the description of the part with the same role and function as in FIG. 1 is omitted.

In S110, features are extracted from target data samples. S110 in this embodiment is the same as S110 in the embodiment shown in FIG. 1 , so details are not repeated here.

In S150, cluster the candidate data in the candidate data set by using the features of the target data samples extracted in S110.

In S160, an appropriate class is selected as the target data according to the similarity between the class generated by the clustering and the target data sample. Here, a suitable class should be understood as a class with a high similarity between the target data sample and the class generated by clustering.

Fourth embodiment

As can be seen from FIG. 4, the embodiment shown in FIG. 4 is a combination of the second embodiment and the third embodiment.

In S110, features are extracted from target data samples.

In S130, the feature is used to query the candidate data set, and in S170, the similarity between the queried candidate data and the target data sample is judged.

In S150, cluster the candidate data in the candidate data set by using the features of the target data sample extracted in S110, and judge the similarity between the clustered cluster and the target data sample in S180. For the sake of brevity, the detailed description of the parts in FIG. 4 that have the same role and function as those in FIG. 2 and FIG. 3 is omitted. For the specific content, please refer to the description of FIG. 2 and FIG. 3 , which will not be repeated here.

In S190, the similarity between the cluster and the target data sample is compared with the similarity between the queried candidate data and the target data sample, and appropriate candidate data is selected as the target data according to the comparison result. Compare the similarities between the candidate data extracted in Figure 2 and Figure 3 and the target data samples, and select the candidate data with high similarity as the target data. Therefore, the advantages of the above two methods are combined to provide more accurate target data.

fifth embodiment

In this embodiment, the method for obtaining a target data set from a candidate data set according to an embodiment of the present application is exemplarily applied to a target data sample in text form. At this point, features can be extracted from the target data samples by means of interval n-gram phrases, n-gram phrases or vocabulary. Here, the candidate data set may correspondingly include a plurality of texts, wherein each text may be regarded as a set of sentences composed of a plurality of sentences. Therefore, when the target data sample is a single sentence, it can be considered that the target data only includes text with one sentence. Here, only text is taken as an example for illustration, but the method according to the embodiment of the present application is also applicable to the target data sample of a single sentence.

In an embodiment using spaced n-grams or n-grams for querying, a window of size S is applied to the target data sample and the target data sample is thus represented by [w ₁ , w ₂ , . . . , w _S ]. An n-gram phrase generated from a target data sample is denoted as P(w ₁ , w ₂ , ..., w _S ), where w _i is the i-th word in the window and |P|≤n, n is a natural number. The words in the phrase can be continuous or discontinuous, and it is an n-gram phrase in the continuous situation, and it is an interval n-gram phrase in the discontinuous situation. For example, the sentence "promote the stable development of the national economy" is represented by an interval n-gram phrase, if n=3, then "promote", "promote the country", "promote the national economy", "promote the country", "promote national stability", "promoting the economy", "promoting the economy", and "promoting economic stability". The candidate data set is queried by means of the n-grams extracted from the target data sample, and the candidate data is determined according to the comprehensive evaluation, thereby forming the target data.

In embodiments where a vocabulary is used for querying, the vocabulary is formed from target data samples. All words from the target data sample are sorted in descending order according to their frequency of occurrence in the target data sample. Therefore, the candidate data sets can be queried in groups with the highest frequency words, or the candidate data sets can be queried in groups with medium frequency words, and the candidate data can be determined according to the similarity, thereby forming the target data. Here, the determination of the similarity can be determined according to how many words with the highest frequency are contained, for example.

Similarly, the candidate data sets can also be clustered by means of interval n-gram phrases, n-gram phrases or vocabulary formed by the target data samples. For example, in an embodiment in which interval n-gram phrases are used to cluster candidate data, a feature set F in the form of n-gram phrases is formed from target data samples. Transform all the data in the candidate dataset into feature vectors V _s < f ₁ : w ₁ , f ₂ : w ₂ , ..., f _m : w _m >, where f _i is the n-gram found in F; w _i is the corresponding weight, and i is a natural number. Preferably, the weight of the feature is determined based on the frequency of occurrence of the feature in the target data sample. Form all vectors into an eigenmatrix. Perform clustering and select the top N classes from the clustering results according to the similarity to obtain the target data, where N is a natural number and can be selected based on experience.

The embodiment of using vocabulary to cluster candidate data is similar to the above embodiment of using interval n-grams to cluster candidate data, and will not be repeated here.

In addition, in the embodiment in which the method for obtaining a target data set from a candidate data set according to an embodiment of the present application is applied to the field of translation, the candidate data set is a translation corpus, and the translation corpus includes mutually corresponding corpus in at least two languages . Here, the expression "corresponding to each other" should be understood as mutual translation between two languages. The term "corpus" should be understood here as terms, sentences or articles. A target dataset is a subset drawn from the translation corpus for a specific purpose. The target data sample is a text in at least one of the at least two languages prepared for the specific purpose. For example, the text in at least one language may be an article or a sentence written in one of the at least two languages.

Exemplarily, the candidate data set is an English-Chinese corpus in the field of control, where the field of control covers electromechanical control, electrical control, mechanical control, and so on. The target data samples only involve English corpus in the field of electrical control. According to the method of the embodiment of the present application, feature can be extracted from target data samples (English corpus or Chinese corpus in electrical control field here); utilize described feature from candidate data set (English-Chinese corpus in control field here) The target data is extracted to form a target data set (here, an English-Chinese corpus in the field of electrical control). Here, it is also possible to query the English-Chinese corpus in the control field by forming the above-mentioned n-grams, interval n-grams or vocabulary from the English corpus (or Chinese corpus) in the field of electrical control, or to aggregate the English-Chinese corpus in the control field. class, or combine these two methods, according to the similarity between the candidate data (here, the English corpus or the Chinese corpus in the English-Chinese corpus in the control domain) and the target data sample (correspondingly, English or Chinese corpus) to determine the candidate data, Thus, the target data set is formed, that is, the English-Chinese bilingual corpus in the field of electrical control is formed. This enables the formation of a purpose-specific corpus.

Of course, those skilled in the art should understand that the candidate data sets also include multiple languages, such as English-German, English-French, and so on. Target data samples and candidate data sets can also cover a wider range than limited to technical fields, such as general dictionaries and so on.

Similarly, the method for obtaining a target data set from a candidate data set according to an embodiment of the present application can also be applied to a professional dictionary to form a target data set explaining idioms, idioms, and professional terms. Here, unlike applications in the translation domain, candidate datasets include term subsets and explanatory corpora. Other processing is the same as the application in the field of translation, so it will not be repeated here.

It should also be noted that the method of the embodiment of the present application is described here only by taking the translated and interpreted corpus as an example. However, those skilled in the art should understand that the embodiments of the present application are not limited to the above-mentioned word processing field, but can be applied to any data set that is most relevant to the target data set from the candidate data set according to the characteristics of the target data set. set of spheres. For example, if the target number of samples is an image, extracting features from the image extracts images with the extracted features from a candidate data set, ie, a gallery, where there may be multiple extracted images.

Seventh embodiment

Fig. 5 shows a schematic structural diagram of an apparatus for acquiring a target data set from a candidate data set according to an embodiment of the present application. The device has: a feature extraction unit 510 for extracting features from target data samples; an extraction unit 520 for using the features to extract target data from the candidate data set to form a target data set. The feature extraction unit 510 provides the extracted features to the extraction unit 520 so that it extracts target data from the candidate data set to form a target data set. Likewise, the target data sample may include one or more data, where the data includes data elements. Data can include: strings, sentences, or collections of pictures. Correspondingly, the data elements may be characters, words or pictures. Obviously, said feature can be any feature. As a non-limiting example, the extracted features may consist of at least some of the data elements. For example, if the target data sample is a sentence, its data elements are words constituting the sentence, and the extracted feature is at least one word in the sentence. When multiple features are extracted from the target data sample, the weight of each feature is determined based on the frequency of occurrence of the feature in the target data sample. Select high-weight features as the features of the target data sample. In other words, the more frequently a feature occurs in the target data sample, the higher its weight. The specific details of the working mode of the device correspond to the flow charts of the methods described in FIGS. 1 to 4 . I won't repeat them here.

In another embodiment, the only difference from the above implementation is that the extraction unit 520 further includes: a query unit 521 and a generation unit 522 , as shown in FIG. 6 . That is to say, the extraction unit 520 in this embodiment extracts candidate data and generates a target data set by means of an information retrieval method. In this embodiment, the query unit 521 uses features to query candidate data sets. The generation unit 522 obtains the target data according to the similarity between the candidate data queried by the query unit 521 and the target data sample. For specific working methods of the query unit 521 and the generation unit 522, please refer to the above-mentioned embodiments described for the method according to the embodiments of the present application. I won't repeat them here.

In another embodiment, the only difference from the above implementation is that the extracting unit 520 includes: a clustering unit 523 and a generating unit 522 , as shown in FIG. 7 . That is to say, the extraction unit 520 in this embodiment extracts candidate data by means of a clustering method and generates a target data set. In this embodiment, the clustering unit 523 uses the features to cluster the data in the candidate data set. The generating unit 522 selects an appropriate class as the target data according to the similarity between the cluster generated class and the target data sample. For specific working methods of the clustering unit 523 and the generating unit 522, please refer to the above-mentioned embodiments described for the method according to the embodiments of the present application. I won't repeat them here.

Furthermore, in another embodiment, the extracting unit 520 includes a query unit 521, a clustering unit 523, a comparing unit 521 and a generating unit 522, as shown in FIG. 8 . The extraction unit 520 in this embodiment extracts candidate data by means of information retrieval and clustering methods, and compares the similarity between the candidate data obtained by these two methods and the target data sample to generate a target data set. In other words, it is a combination of the above-mentioned information retrieval and clustering methods. In this embodiment, the clustering unit 523 uses features to cluster the candidate data in the candidate data set. The comparison unit 524 compares the clusters generated according to the clustering and the similarity between the queried candidate data and the samples of the target data. The generation unit 522 is configured to select suitable candidate data as target data according to the comparison result.

In one embodiment, the device according to the embodiments of the present application is applied in the field of translation. In this case, the candidate data set is a translation corpus, the translation corpus includes mutually corresponding corpus of at least two languages, the target data set is a subset extracted from the translation corpus for a specific purpose, and the target data sample is A text in at least one of the at least two languages prepared for the specific purpose. The details of the specific operation of the device are the same as those described in the fifth embodiment. I won't repeat them here.

The embodiment of the present application also proposes a translation machine, which has a corpus database including corpus corresponding to each other in at least two languages, and the corpus database includes the target data set obtained according to the method of the embodiment of the present application.

In addition, the embodiment of the present application also proposes a translation machine, which has the device according to the embodiment of the present application.

The feature extraction unit and extraction unit of the device for obtaining the target data set from the candidate data set can be configured by means of software, firmware, hardware or a combination thereof. Specific means or manners that can be used for configuration are well known to those skilled in the art, and will not be repeated here. In the case of realization by software or firmware, the program constituting the software is installed from a storage medium or a network to a computer (such as a general-purpose computer 900 shown in FIG. 9 ) having a dedicated hardware configuration. When the computer is installed with various programs, Capable of performing various functions, etc.

In FIG. 9 , a central processing unit (CPU) 901 executes various processes according to programs stored in a read only memory (ROM) 902 or programs loaded from a storage section 908 to a random access memory (RAM) 903 . In the RAM 903, data required when the CPU 901 executes various processing and the like is also stored as necessary. The CPU 901, ROM 902, and RAM 903 are connected to each other via a bus 904. An input/output interface 905 is also connected to the bus 904 .

The following components are connected to the input/output interface 905: an input section 906 (including a keyboard, a mouse, etc.), an output section 907 (including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.), A storage section 908 (including a hard disk, etc.), a communication section 909 (including a network interface card such as a LAN card, a modem, etc.). The communication section 909 performs communication processing via a network such as the Internet. A driver 910 may also be connected to the input/output interface 905 as needed. A removable medium 911 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 910 as necessary, so that a computer program read therefrom is installed into the storage section 908 as necessary.

In the case of realizing the above-described series of processes by software, the programs constituting the software are installed from a network such as the Internet or a storage medium such as the removable medium 911 .

Those skilled in the art should understand that such a storage medium is not limited to the removable medium 911 shown in FIG. 9 in which the program is stored and distributed separately from the device to provide the program to the user. Examples of the removable media 911 include magnetic disks (including floppy disks (registered trademark)), optical disks (including compact disk read only memory (CD-ROM) and digital versatile disks (DVD)), magneto-optical disks (including )) and semiconductor memory. Alternatively, the storage medium may be a ROM 902, a hard disk contained in the storage section 908, etc., in which the programs are stored and distributed to users together with devices containing them.

The invention also proposes a program product storing machine-readable instruction codes. When the instruction code is read and executed by a machine, the above-mentioned method according to the embodiment of the present invention can be executed.

Correspondingly, a storage medium for carrying the program product storing the above-mentioned machine-readable instruction codes is also included in the disclosure of the present invention. The storage medium includes, but is not limited to, a floppy disk, an optical disk, a magneto-optical disk, a memory card, a memory stick, and the like.

Finally, it should also be noted that the term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements, but also Other elements not expressly listed, or inherent to the process, method, article, or apparatus are also included. Furthermore, without further limitations, an element defined by the phrase "comprising a..." does not preclude the presence of additional identical elements in the process, method, article, or apparatus comprising said element .

Although the embodiments of the present invention have been described in detail above with reference to the accompanying drawings, it should be understood that the above-described embodiments are only used to illustrate the present invention, rather than to limit the present invention. Various modifications and changes can be made to the above-described embodiments by those skilled in the art without departing from the spirit and scope of the present invention. Accordingly, the scope of the present invention is limited only by the appended claims and their equivalents.

It is not difficult to see from the above description that according to the embodiments of the present invention, the following solutions are provided:

Additional Note 1. A method for obtaining a target dataset from a candidate dataset, comprising:

Extract features from target data samples;

Using the features to extract target data from the candidate data set to form a target data set.

Additional Note 2. The method according to Additional Note 1, wherein,

Extracting target data from the candidate data set includes: using the features to query the candidate data set, and obtaining the target data according to the similarity between the queried candidate data and target data samples.

Additional Note 3. The method according to Additional Note 1, wherein,

Extracting target data from the candidate data set includes: using the features to cluster the candidate data in the candidate data set, and selecting an appropriate class as the target data according to the similarity between the classes generated by the clustering and the target data samples .

Additional Note 4. The method according to Additional Note 2, wherein,

Extracting target data from the candidate data set includes: using the features to cluster the data in the candidate data set, and combining the similarity between the class generated according to the clustering and the target data sample with the candidate data and the target data queried Comparing the similarity of samples, selecting suitable candidate data as the target data according to the comparison result.

Supplementary Note 5. The method according to one of Supplementary Notes 1 to 4, wherein,

Each item of said data comprises data elements, at least some of which constitute said characteristic.

Supplementary Note 6. The method according to one of Supplementary Notes 1 to 4, wherein,

The weight of the feature is determined based on the frequency of occurrence of the feature in the target data sample.

Supplement 7. The method according to any one of Supplements 1 to 4, wherein the data elements are characters, words or pictures, and correspondingly, the data includes: character strings, sentences or picture sets.

Supplementary Note 8. The method according to any one of Supplementary Notes 1 to 4, wherein the candidate data set is a translation corpus, the translation corpus includes at least two languages corresponding to each other, and the target data set is for a specific purpose A subset extracted from the translation corpus, the target data sample being text in at least one of the at least two languages prepared for the specific purpose.

Supplement 9. The method according to Supplement 8, wherein the feature is an n-gram, an interval n-gram or a vocabulary.

Supplementary note 10. A device for obtaining a target data set from a candidate data set, comprising:

A feature extraction unit is used to extract features from the target data sample;

An extraction unit, configured to use the features to extract target data from the candidate data set to form a target data set.

Supplementary Note 11. The device according to Supplementary Note 10, wherein,

The extraction unit includes a query unit and a generation unit, the query unit uses the features to query the candidate data set, and the generation unit obtains the target data.

Supplementary Note 12. The device according to Supplementary Note 10, wherein,

The extraction unit includes a clustering unit and a generation unit, the clustering unit uses the features to cluster the data in the candidate data set, and the generation unit selects the appropriate class according to the similarity between the class generated by the cluster and the target data sample class as the target data.

Supplementary Note 13. The device according to Supplementary Note 11, wherein,

The extraction unit further includes a clustering unit and a comparison unit, the clustering unit utilizes the features to cluster the candidate data in the candidate data set, and the comparison unit combines the classes generated according to the clustering and the queried candidate data with For similarity comparison of samples of target data, the generating unit is configured to select suitable candidate data as the target data according to the comparison result.

Supplementary Note 14. The device according to any one of Supplementary Notes 10 to 13, wherein,

Each item of said data comprises data elements, at least some of which constitute said characteristic.

Supplementary Note 15. The device according to any one of Supplementary Notes 10 to 13, wherein,

The feature extraction unit is configured to determine the weight of the feature based on the frequency of occurrence of the feature in the target data sample.

Supplement 16. The device according to any one of Supplements 10 to 13, wherein the data element is a character, a word or a picture, and correspondingly, the data includes: a character string, a sentence or a collection of pictures.

Supplementary Note 17. The device according to any one of Supplementary Notes 10 to 13, wherein the candidate data set is a translation corpus, the translation corpus includes at least two languages corresponding to each other, and the target data set is for a specific purpose A subset extracted from the translation corpus, the target data sample being text in at least one of the at least two languages prepared for the specific purpose.

Supplementary Note 18. The device according to Supplementary Note 17, wherein the feature is an n-gram, an interval n-gram or a vocabulary.

Supplement 19. A translation machine having a corpus database including corpus corresponding to each other in at least two languages, the corpus database including the target data set obtained according to the method described in any one of Supplements 1 to 9.

Supplementary Note 20. A translation machine having the device according to any one of Supplementary Notes 10 to 18.

Claims (10)

1. A method for obtaining a target dataset from a candidate dataset, comprising: Extract features from target data samples; Using the features to extract target data from the candidate data set to form a target data set. 2. The method of claim 1, wherein, Extracting target data from the candidate data set includes: using the features to query the candidate data set, and obtaining the target data according to the similarity between the queried candidate data and target data samples. 3. The method of claim 1, wherein, Extracting target data from the candidate data set includes: using the features to cluster the candidate data in the candidate data set, and selecting an appropriate class as the target data according to the similarity between the classes generated by the clustering and the target data samples . 4. An apparatus for obtaining a target data set from a candidate data set, comprising: A feature extraction unit is used to extract features from the target data sample; An extraction unit, configured to use the features to extract target data from the candidate data set to form a target data set. 5. The apparatus of claim 4, wherein, The extraction unit includes a query unit and a generation unit, the query unit uses the features to query the candidate data set, and the generation unit obtains the target data. 6. The apparatus of claim 4, wherein, The extraction unit includes a clustering unit and a generation unit, the clustering unit uses the features to cluster the data in the candidate data set, and the generation unit selects the appropriate class according to the similarity between the class generated by the cluster and the target data sample class as the target data. 7. The apparatus of claim 5, wherein, The extraction unit further includes a clustering unit and a comparison unit, the clustering unit utilizes the features to cluster the candidate data in the candidate data set, and the comparison unit combines the classes generated according to the clustering and the queried candidate data with For similarity comparison of samples of target data, the generating unit is configured to select suitable candidate data as the target data according to the comparison result. 8. Apparatus according to any one of claims 4 to 7, wherein Each item of said data comprises data elements, at least some of which constitute said characteristic. 9. A translation machine having a corpus database comprising corpus corresponding to each other in at least two languages, the corpus database comprising the target data set obtained according to the method according to one of claims 1 to 3. 10. A translation machine with a device according to one of claims 4 to 8.

Images