KR101739766B1 - Apparatus and Method for similarity based keyword mapping in Semantic Search - Google Patents

Abstract

According to the present invention, there is provided a similarity-based keyword mapping method in semantic search, comprising: performing a first term mapping for searching a knowledge base object mapped to a search keyword from a knowledge base; , And performing a secondary term mapping with the searched similar term.

Description

[0001] The present invention relates to a similarity-based keyword mapping apparatus and method for semantic search,

The present invention relates to a knowledge search technique, and more particularly, to an apparatus and a method for mapping a search keyword input from a user into a knowledge entity.

With the diversification of personal smart devices such as smart phones and smart TVs, users can perform information searches using these various devices. In the information retrieval using the smart device, different types of requirements are presented as compared with the conventional PC based information retrieval.

First, since smartphones and smart TVs do not provide a user-friendly input device such as a keyboard provided in a conventional PC, it is necessary to provide search results with high accuracy even with a few simple keywords.

Second, smartphones and smart TVs do not provide the same high resolution as conventional PCs, so it is difficult to display numerous search results. In addition, it is difficult for a user to search a large number of search results directly to find desired search results.

Therefore, another type of information retrieval technology such as semantic retrieval based on accuracy is required rather than keyword based retrieval based on recall rate.

Web search engine, which is a typical form of information retrieval, is based on keyword based retrieval. The keyword-based retrieval method collects documents on the web, indexes frequently occurring words in the collected documents, and finds the location of a document containing a large number of words matching the keyword entered by the user.

However, this type of information retrieval system finds only a document containing the same keyword as the keyword inputted by the user without considering the semantic of the keyword inputted by the user. Therefore, in this method, although the recall rate for retrieving a document including a keyword is high, the precision rate for retrieving a document having exactly the same meaning as the keyword inputted by the user is inevitably lowered.

Therefore, in order to solve such a problem, a semantic search technique which is a semantic-based search has appeared. This interprets the meaning based on the knowledge base and generates the search result, so that the user can provide an accurate search result desired by the user.

This semantic search is performed through terminology mapping (Term Mapping) and query graph construction (Query Graph Construction) to provide accurate search results. The term mapping is a process of searching an object mapped to a search keyword input by a user from a knowledge base. Query graph generation is a process of searching a graph connecting knowledge base objects acquired through the term mapping process.

However, if any one of the term mapping fails, accurate search results desired by the user can not be generated. That is, the search keywords based on user input errors such as spacing, typos, and foreign words, and the search keywords in the form of synonyms not registered in the knowledge base can not be analyzed semantically.

In the semantic search, search keywords that are not registered in the knowledge base due to user input errors such as spacing, typo, and foreign language marking are not correctly mapped to knowledge base entities, thereby degrading the quality of the search results.

The present invention relates to an apparatus and a method for mapping a keyword based on a similarity to a search keyword that is not registered in a knowledge base, and a search keyword based on a user input error such as spacing, typo, and foreign word.

According to the present invention, there is provided a similarity-based keyword mapping method in semantic search, comprising: performing a first term mapping for searching a knowledge base object mapped to a search keyword from a knowledge base; And performing a secondary term mapping with the retrieved similar term.

The present invention relates to a similarity-based keyword mapping apparatus in semantic search, comprising: a term mapping unit for searching a knowledge base object mapped to a search keyword from a knowledge base; And the term mapping unit re-executes the term mapping with a similar term that is returned.

According to the present invention, it is possible to prevent deterioration in search result quality due to spacing errors, typographical errors, and foreign language marking errors.

Also, it is possible to accommodate various search keywords of a plurality of users by expanding the keyword mapping through searching for synonyms.

In addition, since the near-word / thesaurus table for typing, foreign-word representation, and thesaurus processing is dependent on a knowledge base, it is possible to easily construct a close-proximity / thesaurus table using a knowledge base already established.

In addition, semantic search can be commercialized by applying functions such as spacing, typing, foreign word marking, and thesaurus processing in the semantic search domain.

1 is a structural diagram of a general semantic search keyword mapping apparatus.
2 is a block diagram of a similarity-based keyword mapping apparatus in a semantic search according to an embodiment of the present invention.
3 is a flowchart illustrating a method of mapping similarity-based keywords in semantic search according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the like elements throughout.

In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

The terms used throughout the specification are defined in consideration of the functions in the embodiments of the present invention and can be sufficiently modified according to the intentions and customs of the user or the operator. It should be based on the contents of.

1 is a structural diagram of a general semantic search keyword mapping apparatus.

Referring to FIG. 1, when a user search keyword is input from a search client 10, a term mapper 110 of a semantic analyzer 100 searches a knowledge base object . Then, the Query Graph Constructor 120 constructs a query graph composed of knowledge objects, which are the result of the term mapping, and provides the result as a semantic analysis result.

However, this semantic search assumes that the user correctly knows the object label included in the knowledge base and inputs the search keyword matching the object label without error. However, the general user can not accurately know the label of the object registered in the knowledge base structure and knowledge base. Therefore, the user can input an incorrect search keyword including spacing, foreign word marking, and typing, and input a search keyword that is not registered in the knowledge base. In order to provide accurate search results in response to various search keywords input by a plurality of users, a keyword mapping method capable of processing typo, foreign words, and thesaurus is needed.

For example, when a user inputs a search keyword 'action movie with a prize', the term mapping unit 110 searches the knowledge base 130 for an object mapped to the 'action movie'. However, if the label of the object registered in the knowledge base 130 is registered as 'Ryu Seungbeom' or 'Action Movie', the entity mapped to the search keyword can not be searched.

As another example, if a user inputs a search keyword such as 'horror movie' or 'scary movie' to search for 'horror movie', the search result can not be provided due to the failure of keyword mapping. This is because the object labeled 'Fear' is registered in the knowledge base, but the object labeled 'Horror' or 'Scary' having similar meaning is not registered.

Accordingly, the present invention provides a keyword mapping apparatus and method capable of processing the misspelled words, foreign words, and synonyms.

2 is a block diagram of a similarity-based keyword mapping apparatus in a semantic search according to an embodiment of the present invention.

Referring to FIG. 2, the automatic mapping unit 210 and the similarity-based mapping unit 230 are further included in comparison with the keyword mapping apparatus shown in FIG. In addition, the dictionary DB 215, the proximity table 240 and the synonym table 250 are further included in the form.

The automatic spacing unit 210 automatically spaces the search keyword inputted by the user. At this time, the spacing is determined based on the morphological analysis. In addition to the basic spacing, spacing of complex nouns can be adjusted by adding the corresponding word to the morpheme dictionary DB 215 in advance. For example, when an 'unfair transaction' is input as a search term, it is basically required to place a word 'unfair transaction'. However, in order to make the word 'unfair transaction' Unfair trading 'can be added. The automatic spacing unit 210 is driven first when a user search keyword is input.

If the term mapping unit 221 fails to terminate the mapping, the similarity-based mapping unit 230 searches for similar words similar to the user's search keyword and returns it to the term mapping unit 221. The similarity-based mapping unit 230 is composed of a near-field mapping unit 231 and a synonym mapping unit 232 in detail.

The proximity mapping unit 231 searches for a word closest to the search keyword entered by the user in order to deal with typographical and foreign word error. In one embodiment, the close proximity mapping unit 231 extracts all terms having the same length as the search keyword from the proximity table 240, and returns a term having the smallest edit distance among the extracted terms do.

Here, the edit distance is defined as a distance between a search keyword and all extracted terms when an edit operation such as insert, delete, or replace occurs, ) Is defined as 1, and when a minimum edit operation occurs, it is defined as an edit distance of two strings. As the edit distance increases, the similarity of the two strings decreases.

The thesaurus mapping unit 232 searches for terms having the most similar meaning to the search keyword entered by the user. If there is a term matching the value in the synonym table 250 shown in Fig. 2, the thesaurus mapping unit 232 detects and returns the key of the term.

The proximity table 240 and the synonym table 250 must be constructed in advance in order to search for synonyms and synonyms of user-entered keywords through the similarity-based mapping unit 230. [

The proximity word table 240 is a table used for error processing of typographical and foreign language notation, and is used when a term (nearest word) closest to a search keyword inputted by a user is found. The proximity table 240 basically registers the labels of all the objects (Class, ObjectProperty, DatatypeProperty, Instance) registered in the knowledge base 223. Also, the synonyms constructed in the synonym table 250 are also registered in the proximity table 240, and the reason for registering the synonyms in the proximity table 240 is that typos and foreign word marking errors may also exist in the synonym input.

If there is a synonym with a label of Class, ObjectProperty, or DatatypeProperty existing in knowledge base 223, the synonym table 250 specifies a label as a key and one or more synonyms of the label as a value Registered. As shown in FIG. 2, a plurality of values may be registered in one key.

The proximity table 240 and the synonym table 250 should be constructed in an updatable structure as the schema or instance of the knowledge base 223 is added.

The term mapping unit 221 performs term mapping on a word by word basis of the search keyword in which the spacing is completed. At this time, if the term mapping is not successful, the similarity-based mapping unit 230 requests the similarity-based term search of terms in which the mapping fails.

According to one embodiment, when the proximity word searched for by the proximity mapping unit 231 is returned, the term mapping unit 221 replaces the keyword that failed mapping to the searched adjacent word through the near-term search, performs the secondary mapping do. However, if the secondary mapping is unsuccessful, the synonym search unit 221 requests a synonym search. The term mapping unit 221 replaces the unmapped keyword through the synonym search and performs the tertiary mapping again. If the mapping of the terms is not completed, the term mapping unit 221 means that there is no knowledge base object mapped to the user search keyword, and the mapping is terminated even if the mapping is not completed. Here, it is disclosed that the first-order proximity search is performed first and the second-order thesaurus search is performed, but this is an embodiment of the present invention, but the present invention is not limited thereto. That is, the search order may be reversed or may be performed at once.

The query graph construction unit 222 constructs a query graph composed of knowledge objects that are the result of term mapping, and provides the result as a semantic analysis result. Since this is well known, a detailed description will be omitted here.

3 is a flowchart illustrating a method of mapping similarity-based keywords in semantic search according to an embodiment of the present invention. Here, for convenience of explanation, an example in which the user inputs a search keyword 'Ryu Seung-hwan's produced horror movie' will be described together.

Referring to FIG. 3, when the user search keyword is input, the automatic spacing unit 210 performs automatic spacing in S310. For example, 'Ryu Seung-hwan directing horror film' is performed with the search keyword 'Ryu Seung-hwan directing horror movie'.

Then, the term mapping unit 221 performs term mapping on a word-by-word basis in step S320, and determines whether mapping is successful for all the terms in step S320. For example, if the first term is mapped to the keyword 'Ryu Seung-hwan's production horror movie' with the spacing completed, 'Ryu Seung-hwan' and 'Horror' will fail to map the term. This is because 'Ryu Seung-hwan' and 'Horror' do not exist in the label registered as the knowledge base object due to the error of the user's typing in the foreign word.

As a result of the determination in step S330, if the mapping is successful for all the terms, the process proceeds to step S420. If the mapping is unsuccessful for all terms as in the above example, the similarity-based mapping unit 230 searches the similarity- .

More specifically, the close proximity mapping unit 231 performs proximity search first in S340. For example, in the keyword 'Ryu Seung-hwan directed horror film', a close word is searched for 'Ryu Seung-hwan', which is a term that failed mapping. Specifically, the proximity mapping unit 231 detects all terms having a word length of 3 from the proximity table 240, calculates an edit distance between all the detected terms and 'Ryu Seung-hwan' You can find 'Ryu Seung-wan' in terms of minimal editing distance. Likewise. The proximity mapping unit 231 detects all terms having a word length of '2' with respect to 'horror', and detects the terms 'horror' and 'horror' You will find the term. That is, in the case of 'horror', since the edit distance value is '0', it can be understood that it is not a typographical or foreign language marking error.

Then, the term mapping unit 221 replaces the keyword that failed mapping to the searched keyword through the proximity search in step S350, and performs the secondary mapping in step S360. For example, the term mapping is re-executed as the 'Ryu Seung-wan directed horror movie' which is replaced with the searched neighboring words.

The term mapping unit 221 determines whether the mapping is successful in step S370, and proceeds to step S420 if the mapping is successful.

On the other hand, if the mapping is not successful in step S370, the thesaurus searching unit 221 performs thesaurus searching in step S380. For example, if the second term mapping is performed with the replaced search keyword, the 'horror' fails to term map, so that the 'horror' synonym of the second term mapping fails. A term that exactly matches the value of the synonym table 250 is searched for and a value matching the 'horror' is searched for a 'horror' key.

The term mapping unit 221 replaces the unmapped keyword through the thesaurus search in S390, and performs the tertiary mapping again in S400.

If the mapping of the terms is not completed as a result of the determination in S410, the term mapping unit 221 means that there is no knowledge base object mapped to the user search keyword. Even if the mapping is not completed, Lt; / RTI >

However, if the term mapping is successful as a result of the determination in S410, the process proceeds to S420.

In step S420, the query graph construction unit 222 constructs a query graph composed of knowledge objects that are the result of term mapping, and provides the result as a semantic analysis result. Since this is well known, a detailed description will be omitted here.

Here, it is disclosed that the first-order proximity search is performed first and the second-order thesaurus search is performed, but this is an embodiment of the present invention, but the present invention is not limited thereto. That is, the search order may be reversed or may be performed at once.

Claims (14)

A similarity-based keyword mapping method in which a similarity-based keyword mapping apparatus searches an object mapped to a user search keyword inputted from a search client,
Performing a primary term mapping for searching an object mapped to the user search keyword from a knowledge base;
Searching for a term similar to the term in which the primary term mapping fails based on the result of performing the primary term mapping,
And performing a secondary term mapping with the searched similar term,
The searching step
Detecting all terms having the same length as the term that failed the primary term mapping from the pre-established proximity table,
Calculating an edit distance between each of the detected terms and a term that fails the primary term mapping;
And selecting a term having the smallest edit distance among all the detected terms as the similar term
A method for mapping a keyword based on similarity in a semantic search.
The method according to claim 1,
Further comprising performing automatic spacing when the user search keyword is input,
The step of performing the primary term mapping
Wherein the mapping is performed using the automatically spaced user search keyword.
3. The method of claim 2, wherein performing the automatic spacing comprises:
Wherein the spacing is determined based on the morpheme analysis, and the compound nouns are adjusted to be spaced according to a specific compound noun word added in advance in the morpheme dictionary.
2. The method of claim 1, wherein performing the primary and secondary term mapping comprises:
Wherein the similarity-based keyword mapping is performed on a word-by-word basis.
delete 2. The method of claim 1,
Retrieving a value (Value) that matches the term that failed the primary term mapping from a pre-built synonym table;
And detecting a key including the retrieved value,
Wherein a label is designated as a key and one or more synonyms of the label are registered as a value when there is a synonym with a label of an entity existing in the knowledge base. Mapping method.
The method according to claim 1,
Performing the term mapping with the searched word or the thesaurus in the searching step in the step of performing the secondary term mapping,
If the mapping fails, re-executing a thesaurus search or a near-term search for terms that have failed mapping,
And performing third-term mapping using the searched thesaurus or the nearest word.
A term mapping unit for searching an object mapped to a user search keyword from a knowledge base;
And a similarity-based mapping unit for searching for and returning a term similar to the term in which the mapping is unsuccessful, upon request of the term mapping unit,
The term mapping unit re-executes the term mapping with similar terms returned,
Further comprising a proximity table in which all labels registered in the knowledge base are registered,
The similarity-based mapping unit
Detecting all the terms having the same length as the term that failed the mapping from the proximity table, calculating an editing distance between each of the detected terms and the term that failed the mapping, and selecting, from among all the detected terms, Includes a near-term search section for selecting terms similar to the above-mentioned terms
Wherein the similarity-based keyword-mapping device is used in semantic search.
delete 9. The method of claim 8,
A thesaurus table in which a label is designated as a key and at least one synonym of the label is registered as a value when there is a synonym with a label of an entity existing in the knowledge base,
The similarity-based mapping unit
Searching for a value matching the term that failed the mapping from the synonym table and detecting a key including the searched value. ≪ RTI ID = 0.0 > 8. < / RTI > delete delete delete delete

Images