KR100745293B1 - Method for generating word tree of electronic dictionary and word searching system - Google Patents

Abstract

The present invention relates to a method of constructing a word tree for an electronic dictionary and an apparatus for searching for a word based on a word tree structure.

According to the word tree construction method and word tree structure-based electronic dictionary search apparatus of the present invention, a search word and word ID are extracted from raw data, and a word tree is generated based on the extracted search word and word ID. By storing dictionary information based on the generated word tree structure, a query can be quickly searched based on the stored word tree structure.

Electronic dictionary, word, search, tree

Description

METHOOD FOR GENERATING WORD TREE OF ELECTRONIC DICTIONARY AND WORD SEARCHING SYSTEM}

1 is a block diagram of a word tree construction system of an electronic dictionary according to the present invention;

2 is a diagram showing an example of a raw data configuration in the present invention;

3 is a diagram showing an example of search word and word ID extraction in the present invention;

4 is a diagram showing an example of word tree generation in the present invention;

5 is a flowchart of a method for generating a word tree in the present invention.

Figure 6 is a flowchart of the word tree storage method in the present invention

7 illustrates a tree node format in the present invention.

8 illustrates an example of an initial form of a format for storing a word tree in the present invention.

9 illustrates an example of a final form of a format in which a word tree is stored in the present invention.

<Explanation of symbols for the main parts of the drawings>

10: Raw Data 20: Search Word and Word ID Extraction Unit

30: word tree generation unit 40: word tree storage unit

50: search engine 60: user interface unit

The present invention relates to a method of constructing a word tree for an electronic dictionary and an apparatus for searching for a word based on a word tree structure.

Electronic dictionaries are not only used as portable electronic dictionaries (products) equipped with their own functions, but also electronic dictionaries are used in portable terminals such as mobile phones and PDAs. In order to implement an electronic dictionary function in a portable terminal, if the English-Korean electronic dictionary is taken as an example, the original dictionary data including raw words, phonetic symbols, meanings, idioms, synonyms, examples, etc. It should be processed and reconfigured as appropriate for your implementation.

In other words, it is necessary to secure an optimal search result, an optimal search speed, and efficient operation by constructing a dictionary database by appropriately reconstructing raw data to provide an electronic dictionary function. Of particular importance is the speed and reliability of the search. Currently, at least 200,000 words or more are not included in the English-Korean electronic dictionary, but a word tree is generally used to search a large number of words at a high speed and give a search result.

Therefore, there is a need for an efficient database configuration for fast and efficient word search based on a word tree.

An object of the present invention is to provide a method and apparatus for generating, storing and retrieving a word tree for ensuring an efficient and fast retrieval speed in implementing an electronic dictionary function based on a word tree. .

Also, the present invention constructs a word tree by extracting only a search word and a word ID from a raw data, and searches for a word using the constructed word tree, obtains a word ID of the corresponding word, and refers to an interpreter using the ID. The purpose of the present invention is to provide a method of constructing an electronic dictionary word tree.

In addition, the present invention forms an AVL (Adelson-Velskii and Landis) binary search word tree by extracting only the search word and the word ID from the raw data, and forms a suitable word tree for high-speed search. The present invention provides a method and an apparatus for constructing an electronic dictionary word tree for inducing fast word searches.

In addition, the present invention provides a method for constructing an electronic dictionary word tree that can simplify the word tree structure and provide a basis for a high-speed search by using the electronic dictionary function in a portable terminal to efficiently use limited memory and process resources. The purpose of the present invention is to provide a word retrieval apparatus based on a word tree structure.

According to an aspect of the present invention, there is provided a method of constructing an electronic dictionary word tree, the method comprising: extracting a search word and a word ID from raw data; Generating a word tree based on the extracted search word and word ID; Storing dictionary information based on the generated word tree structure; Characterized in that comprises a.

In addition, the electronic dictionary search method of the present invention for achieving the above object, the step of extracting a search word and word ID from the raw (raw data); Generating a word tree based on the extracted search word and word ID; Storing dictionary information based on the generated word tree structure; Searching for a query word based on the stored word tree structure; Characterized in that comprises a.

In addition, the electronic dictionary search apparatus of the present invention for achieving the above object, the database storage means of the word tree structure based on the search word and the word ID extracted from the raw data (raw data), query input and output of the search results And a search engine for searching for and outputting the corresponding word based on the word tree structure of the database storage means with respect to the input query word.

Hereinafter, with reference to the accompanying drawings will be described an embodiment of the electronic dictionary word tree configuration method according to the present invention.

Figure 1 shows the configuration of the electronic dictionary word tree configuration system of the present invention.

The electronic dictionary word tree construction system according to the present invention includes a raw data 10 of dictionary information, a search word and a word ID extracting unit 20 for extracting a search word and a word ID from the raw data; And a word tree generator 30 for generating a word tree using the extracted information, and a word tree storage 40 for storing the generated word tree in an appropriate format. In addition, a search engine 50 and a user interface 60 for searching for words queried by a user from a base to tree structure-based dictionary data stored in the word tree storage 40 are included.

Raw data 10 of the dictionary information may be, for example, data provided from an electronic dictionary manufacturer. This data includes information such as word ID, search word, headword, interpretation, and so on. The search word and word ID extractor 20 extracts the search word and the word ID from the raw data 10 of the dictionary information. The extracted search word and the word ID are transmitted to the word tree generator 30. The word tree generation unit 30 generates a word tree based on the search word and the word ID, and the word tree information is transferred to the storage unit 40 and stored in an appropriate format proposed by the present invention. Fast word search is possible based on the word tree information stored therein. This is done by searching for the query based on the tree structure based database of the word tree storage 40 in the search engine 50 when the query to search is input through the user interface 60. The word search method by the search engine 50 will be described later in detail.

2 shows an example of raw data. Looking at Figure 2 it can be seen the structure of the raw data (raw data) provided for the electronic dictionary implementation. Raw data includes word ID, search word, headword, and interpretation. Generally, a dictionary contains a large amount of information, and at least hundreds of thousands of words exist in the English-Korean dictionary. Therefore, in order to systematically manage a word and all information of the word, an ID is assigned to each word for management and update. Considering this point, raw data of the dictionary includes components such as word ID, search word, headword, and interpretation, as shown in FIG. 2. In addition, information such as a part-of-speech, derivatives, and example sentences may additionally be followed. However, the examples provided in the present invention only include the most basic forms as examples.

When the search word and the word ID are extracted using the raw data information shown in FIG. 2, the search word and the word ID extracting unit 20 described above obtain information as shown in FIG. That is, it is possible to obtain an extraction result consisting of only a search word and a word ID from the raw data 10 of the dictionary. The extracted search word and word ID are used as the most necessary information for constructing the word tree in the present invention. The word tree generator 30 generates a word tree based on the extracted search word and word ID.

4 is a diagram illustrating a word tree structure generated by the word tree generator 30. The word tree structure shown in Fig. 4 is an AVL tree which will be described later, and also has a structure that satisfies a binary search condition. Here, each node of the tree is generated using a word as a key, and each node also has a corresponding word ID.

FIG. 5 illustrates a method of constructing a word tree having the structure shown in FIG. 4 by using the search word and the word ID.

The first step S10 is to read a word and an ID. That is, the word tree generating unit 30 reads the search word and the word ID extracted from the raw data and the word ID extracting unit 20 from the raw data.

The second step S20 is a step of generating a node. That is, a step of generating a tree node to be the basis of the word tree structure.

The third step S30 is to insert the generated node into a tree. To this end, the present invention applies AVL BST (Adelson-Velskii and Landis Binary Search Tree).

In this case, the AVL tree refers to a tree structure in which the heights of subtrees whose nodes are rooted in the tree structure are balanced even with reference to any node at a specific position in the tree structure. That is, since the height difference between the left and right subtrees of the root node has a maximum difference of 1, this is also called a balance tree. A BST is a binary search tree where each node in the tree has one key value, and each node's key value is always greater than the key values of the nodes in the left subtree rooted at that node. It is a tree that is always smaller than the key values of the nodes of the right subtree as the root. If the key values are the same, they are generally inserted into the left subtree.

The word tree proposed by the present invention is an AVL binary tree. The key value of each node is a search word, and the word ID is also stored. That is, the word ID is stored together with the search word as a key value of each node. In other words, when a word tree is created, a tree is generated based on the search word, but the search word and the word ID value are also stored in each node.

After generating nodes for words and word IDs and inserting them into the tree, it is determined whether the current tree structure is an AVL tree. This corresponds to the fourth step S40.

As a result of the determination in the fourth step S40, if the current tree structure is not the AVL tree, the node is rotated so as to become the AVL tree. That is, in step S50, when the current tree structure is not the AVL tree, node rotation is performed on the current tree structure so that the current tree structure becomes an AVL tree. In this process, after the node is created and the tree is inserted for the read word, if the word tree does not satisfy the AVL tree condition, the word tree satisfies the AVL tree by rotating the node. The process proceeds to step S60 and corresponds to a process of generating a word tree while repeatedly performing the next word to the last word in the same manner as described above.

As described above, FIG. 4 shows an example of a word tree generated through the word tree generation process. Here, each node of the tree is generated using a word as a key, and each node also has a corresponding word ID. The tree of Fig. 4 satisfies the AVL tree, and also satisfies the condition of the binary search tree.

As described above, when the word tree is created, the generated word tree is saved as a file. 6 is a flowchart showing a word tree storing method.

The first step S100 is to obtain a root node pointer of the word tree. That is, the word tree generation unit 30 obtains a pointer of the root node of the word tree already generated. Next, in a second step (S200), the space to store the location of the root node is initialized. This is a step to initialize to an arbitrary value because the current step does not know where the root node will be stored. In FIG. 8, the location of the root node is initialized and stored as 'PP'. In this way, the position values of the root nodes are initialized and stored, and then the information of each node is stored in the order of visited nodes while the tree is traversed in order using the pointer of the root node (in order traversal) (step S300). . Here, the information value of the node refers to a search word and word ID stored in the node and a position value (address) using pointers to left and right child nodes belonging to the node.

As such, when a node visits information values of each node in order traversal and stores node information, the search word and word ID of the node are known, but it is not known at which address the child node is stored. In other words, the location address of each node is not known until all nodes have been visited and stored. Therefore, in the step of storing the node information for the first time, the positions of the left and right subtrees are simply initialized. After that, the location of all nodes is updated (step S400), and in FIG. 8, since the location information is not known as 'LLRR' for each node, it can be seen that it is simply initialized.

Through such a series of processes, each node has a total of five information values, including word count, search word, word ID, position of left child and position of right child. Here, the number of word characters is sufficient because only one byte can be expressed in English words because there are no words having more than 128 characters. In the case of English, the search word occupies one byte for each Hangul character, and thus a storage space corresponding to the number of characters × 1 byte is required. In the case of the word ID, it may be expressed differently according to the total range of ID values recorded by the provider providing the dictionary. In general, in the case of an English dictionary, if the total number of words is about 200,000 words to about 300,000 words, in this case, it is safe to express by 4 bytes. In the case of the left child position and the right child position, it can be expressed as four bytes. 8 and 9 apply the example of FIG. 4 described above, so only 2 bytes are allocated and represented.

7 shows the format of a tree node including the number of word characters, search word, word ID, outer child position, and right child position information.

FIG. 8 shows an example of a tree stored in the step of initializing and storing the position of the root node in the word tree storage unit and visiting and storing the nodes in order traversal.

Look at this. At the very beginning it says 'PP4base100LLRR'. Here, 'PP' is initialized to an arbitrary value because the location of the root node is not known. The next '4' means that the length of the word stored in the current node is '4'. 'Base' is the search word of the current node, and '100' is the word ID value of the current node. Next, 'LLRR' refers to the location information of the left child node and the right child node. Since the location address is not known yet, it is initialized to an arbitrary value.

After initial storage as shown in FIG. 8, location updates of all nodes are performed (S400). That is, as shown in FIG. 8, the initial position values such as 'PP', 'LL', and 'RR' are updated in the result of the step S300.

9 shows the result of updating the position of each node in the form of hexadecimal code.

If you simply describe the beginning of '2800 04 62617365 64000000 0000 00 00', the first '2800' means that the root node is at '28' as a result of updating 'PP'. If you go to '28', you can see that it points to the front node of the word 'computer'. Then '04' means the word length of the current node is '4', and '62617365' is the ASCII code value of 'base'. That is, 0x62 = b, 0x61 = a, 0x73 = s, and 0x65 = e.

Next, '64000000' refers to the ID value of the current word. The ID value of the current word 'base' is '100'. That is, 0x00000064 = 100. In this example, the word ID is represented by 4 bytes. Next, '0000' refers to the location information of the left subtree. Referring to FIG. 4, since there is no left subtree in the 'base' node, it indicates null. Therefore, the left subtree location information is updated with the value of '0'. Similarly, the next '0000' refers to the position information of the right subtree. In FIG. 4, since the right subtree of the 'base' node does not exist, it is updated to a value of '0'. In this case, two bytes are used to represent child position information. If a word tree composed of such a format is saved as a file, the saved file can be opened again and the word search can be performed at any time.

For example, suppose that the word 'girl' is input as the user's query word when the word search DB is configured as shown in FIG. First, the word search DB file configured as shown in FIG. 9 is opened. Then read the first two bytes. At this time, since the value is '2800', move the file pointer to '0x28'. That is, as shown in Fig. 7, the format of the tree node is already defined, so that one byte is read in the moved situation. This information will mean the word length of the search term stored in the currently visited node.

Move to '0x28' and read 1 byte to get '08'. This means '0x08', so the word length is '8'. Therefore, reading the next 8 bytes gives the current node's search term. Here you will read '636F6D7075746572'. Each byte value is an ASCII code value of 'computer'. Therefore, the current node is the root node where the 'computer' search word is stored.

Next, the query is compared with the search term at the current visited node. In other words, when comparing 'girl' and 'computer', 'girl' is larger than 'computer' and must be moved to the right child node. Therefore, to move from the current node to the right child node, you need to know the address where the right child node is stored. Therefore, at the time of the current reading, the word ID moves by the number of bytes stored, jumps over two bytes for storing the left child node position, and reads two bytes at that point to be the position value of the right child node.

Therefore, we read '6100' here. Since this value means '0x61', move the file pointer to this address. Then, reading 1 byte in the same way brings the value '05', and then reads 5 bytes, where 'media', which is '6D65646961', is a stored search term.

When comparing the query words 'girl' and 'media', the query is small, so it must be moved to the left child node. Therefore, if 2 bytes are read after reading 4 bytes for the word ID, the left child node returns the stored address value '4600'. Get In other words, it moves back to '0x46', and if it reads 1 byte again, it gets '05', so it reads 5 bytes.

Then you can see the 'heart' which is '6865617274'. If you compare 'heart' with the query 'girl' again, the query 'girl' is smaller and you must move back to the left child node. Therefore, at this point, reading 4 bytes beyond 4 bytes for word ID will move the child node to '0x39'.

If the file pointer is moved back to the point and 1 byte is read, the value '04' is read, and thus 4 bytes are read again, which indicates that the word 'girl' is stored. Compared with the query girl, it is determined to be the same, so the word ID of the currently visited node is read. In other words, reading 4 bytes from the current file pointer yields '68000000', which means '0x68', so we can finally see that the word ID is '104 = (0x68)'. By using the word ID of the query word thus found, it is possible to obtain the interpretation part of the query word by checking the part matching the word ID in the dictionary DB.

The present invention relates to a technique of constructing a word tree for implementing an electronic dictionary function. By obtaining only a search word and a word ID from raw data of a dictionary and constructing a word tree, a desired word from many hundreds of thousands of words is obtained. Allow for quick search. In addition, when constructing a word tree, it has AVL tree property, so that the word tree is balanced so that the structure is suitable for the word search on average, and the nature of the binary search tree is also the same, which leads to quick results of the search. have.

For example, if the total number of words is N, an average number of comparisons of log ₂ N is required. According to the present invention, a 200,000 word English dictionary performs only about 18 word comparisons, which enables a fairly fast search. In addition, the generated word tree is stored in a file according to a specific storage format proposed by the present invention, thereby providing a function and a foundation for opening a file and performing a word search at any time.

Claims (14)

In a method for organizing a file for searching a word based on a tree structure in an electronic dictionary, Extracting a search word and a word ID from a raw data; A word tree is generated based on the extracted search word and word ID, and a key value of a node is the extracted search word and a key value of each node is a key of nodes of left and right subtrees whose node is a root. Generating a word tree by storing the extracted word ID together with a search word in a corresponding node, having a structure always placed in a constant case with values; Storing dictionary information based on the generated word tree structure; Electronic dictionary word tree configuration method comprising a. The method of claim 1, wherein the word tree is constructed based on an AVL tree structure. The method of claim 1, wherein the word tree is constructed based on a binary search tree structure. The method of claim 1, wherein the generating of the word tree comprises: generating a node based on the extracted search word and the word ID, inserting the generated node into a tree, and verifying whether the tree structure is an AVL tree structure. And converting the electronic dictionary word tree. The method of claim 4, wherein if the tree structure is not an AVL tree structure, node rotation is performed to convert the tree structure into an AVL tree structure. The method of claim 1, wherein the storing of the word tree comprises: initializing and storing the position of the root node using a root node pointer of the word tree, and ordering the visited nodes in order of each node. And storing the information values of the nodes and updating the position values of all the stored nodes. The method of claim 6, wherein the node information value includes a search word, a word ID, and a position value of left and right child nodes in a tree structure. The method of claim 1, wherein the tree node comprises a word number, a search word, a word ID, a position value of a left child node on a tree structure, and a position value of a right child node. . In the method of searching for words based on the tree structure in the electronic dictionary, Extracting a search word and a word ID from a raw data; A word tree is generated based on the extracted search word and word ID, and a key value of a node is the extracted search word and a key value of each node is a key of nodes of left and right subtrees whose node is a root. Generating a word tree by storing the extracted word ID together with a search word in a corresponding node, having a structure always placed in a constant case with values; Storing dictionary information based on the generated word tree structure; Searching for a query word based on the stored word tree structure; Electronic dictionary search method comprising a. The electronic dictionary search method according to claim 9, wherein in the searching step, a corresponding word is searched for based on a node of the tree structure. 10. The electronic dictionary search method of claim 9, wherein the searching step searches for a corresponding word by comparing a search word, a word ID, and a query word of the node while visiting a node of the tree structure. . In the device for searching a word based on the tree structure in the electronic dictionary, In the word tree structure, the key value of a node is a search word extracted from raw data, and the key value of each node is always consistent with the key values of the nodes of the left and right subtrees rooted at the node. Database storage means of a word tree structure having a structure placed in a large and small relationship, and storing the extracted word ID together with a search word in a corresponding node; Interface means for inputting a query and outputting a search result; A search engine that searches for a corresponding word in a word tree structure of the database storage means by searching a corresponding word and finds interpretation information based on the ID of the searched word and outputs a search result for the query word; A search engine for searching for and outputting the word based on the search result; Electronic dictionary search device based on the word tree structure, characterized in that comprises a. The electronic dictionary search apparatus according to claim 12, wherein the word tree structure stored in the database storage means is an AVL binary tree structure. 13. The word according to claim 12, wherein the node in the word tree structure stored in the database storage means includes the number of word characters, a search word, a word ID, and position information of left and right child nodes on the tree structure. Electronic dictionary search device based on tree structure.

Images