JP6808851B2 - Topic structuring method, search result provision method, computer program and topic structuring system - Google Patents

Description

The following description relates to a technique for automatically extracting and structuring subtopics suitable for queries.

When the search system provides search results for a query entered by a user, it provides various functions to support additional search of the user in addition to documents matched with the search conditions. Typical examples that support additional user search include related search terms, related tags, and automatic search term completion. These are based on co-occurrence, in which words are shown together as a pair, and are based on grasping search terms and tags that frequently appear as queries.

For example, Korean Published Patent No. 10-2012-909608 (publication date: August 31, 2012) provides a search term recommendation system that selects recommended search terms based on the location information of the user terminal and provides them to the user. The search term recommendation method is disclosed.

On the other hand, in the case of a query intended for shopping, the shopping search system may provide a function to support the search for products by utilizing hierarchical information such as product brand, color, and price.

Korean Published Patent No. 10-2012-090686

It is necessary to provide query topics with a clear subject and to structure the topics for efficient information retrieval.

When a specific subject is given, only the subtopics suitable for the subject are extracted and the hierarchical information automatically constructed for the specific subject is utilized to the extent desired by the user (subdivision). ) Provide a method of appropriately structuring and presenting.

A computer-implemented topic structuring method that extracts subtopics related to the subject by subject, generates a topic tree for the subtopic using the hierarchical information of the subject, and searches. Given a query for, provides a topic structuring method that includes the step of providing the subtopic as a related search term for the query in a hierarchical manner, based on the topic tree of the subject to which the query belongs.

According to one aspect, the extraction step may extract the subtopic by analyzing the words associated with the core object that determines the subject.

According to another aspect, the subtopic may further include filtering by at least one of document appearance frequency and search frequency.

According to another aspect, the subtopic may be further clustered by synonyms or substring relationships to select a representative for each cluster.

According to another aspect, in the generation step, the topic tree may be generated by labeling the subtopic with each class name of the hierarchical information.

According to another aspect, the generation step is a step of extracting similar words from word embedding data for the subtopic, and the similar words are clustered by synonyms or substrings (substrings). It may include a step of labeling and a step of labeling the clustered words by mapping them to each class in linguistic taxonomy.

Further, according to another aspect, the step of performing rebalancing on the topic tree by reducing at least one of the width and depth of the topic tree may be further included.

According to yet another aspect, the provided step is at least one of a subject score indicating the relevance of the query to the subtopic, a number of documents corresponding to the subtopic, and the presence or absence of a correctness topic for the query. It may include a step of filtering the subtopic according to one condition.

It is a search result providing method realized by a computer, and when a query for a search is given, a related search for the query is performed based on the hierarchical information of the subject to which the query belongs at the stage of providing the search result corresponding to the query. The stage of providing subtopics related to the subject as words in a hierarchical form having a plurality of depths, and when at least one search term of the subtopics is selected, the selected search term. Provided is a search result providing method including a step of providing a search result corresponding to the query including.

A computer program recorded on a computer-readable recording medium to be combined with a computer system to perform a topic structuring method, wherein the topic structuring method extracts subtopics associated with the subject by subject. Given a stage, a stage to generate a topic tree for the subtopic using the hierarchical information of the subject, and a query for searching, the association to the query is based on the topic tree of the subject to which the query belongs. Provided is a computer program including a step of providing the subtopic as a search term in a hierarchical manner.

A computer-implemented topic-structured system that includes at least one processor implemented to execute computer-readable instructions, the at least one processor extracting subtopics associated with the subject by subject. When a query is given to the subtopic, a generator that generates a topic tree based on the hierarchical information of the subject, and a query for searching, the query is based on the topic tree of the subject to which the query belongs. Provided is a topic structuring system including a providing unit that provides the subtopics in a hierarchical manner as a related search term for.

According to the embodiment of the present invention, when a specific subject is given, only the subtopics suitable for the subject are extracted and the hierarchical information suitable for the specific subject is automatically constructed, and then the user desires the subject. By appropriately structuring and presenting according to the degree (subdivision), the user can efficiently grasp the detailed attributes / sub-subjects that are suitable for the subject but related to the relevant query, which is practical. It can contribute to supporting additional searches.

It is a figure which showed the example of the network environment in one Embodiment of this invention. It is a block diagram for demonstrating the internal structure of an electronic device and a server in one Embodiment of this invention. It is a figure which showed the example of the process of hierarchizing a patterned query in one Embodiment of this invention. It is a figure which showed the example of the component which the processor of the server can include in one Embodiment of this invention. It is a flowchart which showed the example of the method which a server can execute in one Embodiment of this invention. It is a figure which showed an example of the process of filtering and grouping subtopic candidates with respect to the query of "Guam" and "potato" in one Embodiment of this invention. It is a flowchart for demonstrating an example of the process of constructing hierarchical information in one Embodiment of this invention. It is a figure which showed an example of the hierarchical information constructed by using clustering and linguistic taxonomy for the subject of travel. It is a figure which showed the example of the process of converting the constructed topic network into a tree (2-depth tree) having a depth of 2 in one Embodiment of this invention. It is a flowchart for demonstrating an example of the tree rebalancing process in one Embodiment of this invention. It is a figure which showed the example of the tree rebalancing process in one Embodiment of this invention. It is a figure which showed the example of the tree rebalancing process in one Embodiment of this invention. It is a figure which showed the example of the search result screen which reflected the 2-depth topic structure in one Embodiment of this invention. It is a figure which showed the example of the search result screen which reflected the 2-depth topic structure in one Embodiment of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

An embodiment of the present invention relates to a technique for automatically extracting and structuring subtopics suitable for a query.

Embodiments, including those specifically disclosed herein, can provide a well-defined query topic and structure the topic for efficient information retrieval, thereby accuracy. Achieve considerable advantages in terms of efficiency, scalability, cost savings, etc.

FIG. 1 is a diagram showing an example of a network environment according to an embodiment of the present invention. The network environment of FIG. 1 shows an example including a plurality of electronic devices 110, 120, 130, 140, a plurality of servers 150, 160, and a network 170. Such FIG. 1 is merely an example for explaining the invention, and the number of electronic devices and the number of servers are not limited as in FIG.

The plurality of electronic devices 110, 120, 130, 140 may be fixed terminals or mobile terminals realized by a computer device. Examples of a plurality of electronic devices 110, 120, 130, 140 include smartphones, mobile phones, tablets, navigation systems, PCs (personal computers), notebook pancons, digital broadcasting terminals, PDAs (Personal Digital Assistants), and PMPs (Portables). Multimedia Player) and the like. As an example, the first electronic device 110 may use a wireless or wired communication method to communicate with other electronic devices 120, 130, 140 and / or servers 150, 160 via a network 170.

The communication method is not limited, and not only the communication method utilizing the communication network (for example, mobile communication network, wired Internet, wireless Internet, broadcasting network) that can be included in the network 170, but also the short distance between devices. Wireless communication may be included. For example, the network 170 includes a PAN (personal area network), a LAN (local area network), a CAN (campus area network), a MAN (metropolitan area network), a WAN (wide area network), etc. It may include any one or more of the networks. Further, network 170 may include any one or more of network topologies, including bus networks, star networks, ring networks, mesh networks, star-bus networks, tree or hierarchical networks, and the like. It is not limited to these.

Each of the servers 150, 160 may be implemented by one or more computer devices that communicate with electronic devices 110, 120, 130, 140 via a network 170 to provide instructions, codes, files, contents, services, and the like.

As an example, the server 160 may provide a file for installing an application to a first electronic device 110 connected via a network 170. In this case, the first electronic device 110 may install the application by using the file provided by the server 160. Further, a service provided by the server 150 by connecting to the server 150 under the control of the operating system (OS) included in the first electronic device 110 and at least one program (for example, a browser or the installed application). And content may be provided. For example, when the first electronic device 110 sends a service request message to the server 150 via the network 170 according to the control of the application, the server 150 may send a code corresponding to the service request message to the first electronic device 110. , The first electronic device 110 may provide the content to the user by forming and displaying a screen based on the code according to the control of the application.

FIG. 2 is a block diagram for explaining the internal configurations of the electronic device and the server according to the embodiment of the present invention. In FIG. 2, the internal configuration of the first electronic device 110 will be described as an example for one electronic device, and the internal configuration of the server 150 will be described as an example for one server. Other electronic devices 120, 130, 140 and server 160 may also have the same or similar internal configuration.

The first electronic device 110 and the server 150 may include memories 211 and 221, processors 212 and 222, communication modules 213 and 223, and input / output interfaces 214 and 224. The memories 211 and 221 are computer-readable recording media and include a RAM (random access memory), a ROM (read only memory), and a permanent mass storage device such as a disk drive. Good. Further, the memory 211 and 221 may record an operating system and at least one program code (for example, a code for an application installed and driven in the first electric device 110). Such software components may be loaded from a computer-readable recording medium separate from the memories 211 and 221. Such other computer-readable recording media may include computer-readable recording media such as floppy® drives, disks, tapes, DVD / CD-ROM drives, memory cards, and the like. In other embodiments, software components may be loaded into memory 211 and 221 through communication modules 213 and 223 that are not computer readable recording media. For example, at least one program is a program installed by a file provided by a file distribution system (as an example, the server 160 described above) that distributes a developer or application installation file via a network 170 (as an example, described above). It may be loaded into memory 211,221 based on the application).

Processors 212 and 222 may be configured to process instructions in a computer program by performing basic arithmetic, logic, and input / output operations. Instructions may be provided to processors 212 and 222 by memory 211, 221 or communication modules 213 and 223. For example, processors 212 and 222 may be configured to execute instructions received according to program code stored in a recording device such as memory 211 and 221.

The communication modules 213 and 223 may provide a function for the first electronic device 110 and the server 150 to communicate with each other via the network 170, or another electronic device (for example, the second electronic device 120). Alternatively, it may provide a function for communicating with another server (for example, server 160). As an example, a request (as an example, a search request) generated by the processor 212 of the first electronic device 110 according to a program code stored in a recording device such as a memory 211 is passed through the network 170 under the control of the communication module 213. May be transmitted to the server 150. On the contrary, the control signals, instructions, contents, files, etc. provided under the control of the processor 222 of the server 150 pass through the communication module 223 and the network 170, and pass through the communication module 213 of the first electronic device 110 to the first electronic device. It may be received by the device 110. For example, the control signals and instructions of the server 150 received through the communication module 213 may be transmitted to the processor 212 and the memory 211, and the contents and files are stored in a storage medium that can be further included in the first electronic device 110. May be done.

The input / output interface 214 may be a means for an interface with the input / output device 215. For example, an input device may include a device such as a keyboard or mouse, and an output device may include a device such as a display for displaying a communication session of an application. As another example, the input / output interface 214 may be a means for an interface with a device such as a touch screen in which functions for input and output are integrated into one. As a more specific example, the processor 212 of the first electronic device 110 is configured by using the data provided by the server 150 and the second electronic device 120 in processing the instructions of the computer program loaded in the memory 211. The service screen or content may be displayed on the display through the input / output interface 214. Similarly, the input / output interface 224 may output information configured by using the data provided by the server 150 when the processor 222 of the server 150 processes the instruction of the computer program loaded in the memory 221.

Also, in other embodiments, the first electronic device 110 and the server 150 may include more components than the components of FIG. However, most prior art components need not be clearly illustrated. For example, the first electronic device 110 may be realized to include at least a part of the above-mentioned input / output device 215, a transceiver, a GPS (Global Positioning System) module, a camera, various sensors, a database, and the like. It may further include other components such as. As a more specific example, when the first electronic device 110 is a smartphone, an acceleration sensor, a gyro sensor, a camera, various physical buttons, a button using a touch panel, and an input / output port generally included in the smartphone are included. Various components such as a vibrating device for vibration may be further included in the first electronic device 110.

The following describes a topic structuring method for automatic subtopic extraction and structuring of thematic queries, and a specific embodiment of the topic structuring system.

Topic structuring (grouping and hierarchization) is required to enable efficient information retrieval by providing the maximum variety of search results for a single query.

When the search system provides search results for queries entered by the user, it provides a variety of features that support additional user searches, as well as documents that match the search criteria. Typical examples include related search terms, association tags, and automatic search term completion. Such a function is considered to be useful in the following points.

First, it has the function of recommending a query when the user does not know the clear query name related to the information he / she wants to know.

Second, it allows the user to indirectly understand the detailed attributes / sub-subjects associated with the query in question.

However, the relationship of "relationship" has the following problems.

First, the relationship "association" is ambiguous, and the specific relationship with the query in question (eg, upper / lower concept, synonym, or sibling concept) is unknown. Therefore, if the number of search terms and tags provided increases, it becomes difficult to organize them structurally, and from the viewpoint of user usability, the number of search terms and tags provided to users must be reduced. It ends up.

Secondly, when the query has ambiguity, the related search terms and tags are provided in an unorganized state according to each meaning, which makes it difficult to support additional search.

On the other hand, in the case of a query intended for shopping, there is hierarchical information provided to support the search for the product desired by the user. Since the hierarchy such as product brand, color, price, etc. is extremely systematic, it supports quick and efficient search, but such information is manually entered by each sales company etc. Therefore, there is a problem that the extensibility is limited and it tends to be applied only to shopping queries.

The present invention solves the above-mentioned limitations while supporting the additional search of the user and performing the function of enabling the user to efficiently grasp the detailed attributes / sub-subjects related to the query. We propose automatic subtopic extraction and structuring technology that can be used.

The core contents of the topic structuring system according to the embodiment of the present invention are as follows.

(1) The topic structuring system patterns the main query by subject into "main object + subtopic". At this time, the main object means the core object that determines the subject, and the subtopic includes at least one of the sub-object and the attribute. The sub-object means an object that embodies the subject, and the attribute means a word that indicates the attribute of the subject such as suffix or prefix.

(2) The topic structuring system hierarchies patterned queries based on the semantic relationships between sub-objects and attributes. FIG. 3 is a diagram showing an example of a process of layering a patterned query in one embodiment of the present invention. As shown in FIG. 3, a query of a specific subject, that is, a main object (MainObj) and a subtopic (Subtopj, Suffix) may be hierarchized based on the main object (MainObj).

(3) The topic structuring system may provide a layered query and subtopics to the user together with the search result (document). At this time, the topic structuring system appropriately structures and presents the subtopics according to the degree (subdivision) desired by the user, thereby displaying detailed subtopics that are suitable for the subject but related to the query. It can be grasped efficiently by the user and can contribute to the support of substantial additional search.

FIG. 4 is a diagram showing an example of components that can be included in the processor of the server in one embodiment of the present invention, and FIG. 5 is a diagram that can be executed by the server in one embodiment of the present invention. It is a flowchart which showed the example of the method.

As shown in FIG. 4, the processor 222 of the server 150 may include an extraction unit 410, a purification unit 420, a generation unit 430, an adjustment unit 440, and a provision unit 450 as components. Such processor 222 and the components of processor 222 may control the server 150 to perform steps 510 to 550 included in the method of FIG. At this time, the processor 222 and the components of the processor 222 may be realized to execute an instruction (instruction) by the code of the operating system included in the memory 221 and the code of at least one program. Further, the components of the processor 222 may be representations of different functions executed by the processor 222 according to control instructions provided by the operating system or at least one program. For example, the extraction unit 410 may be used as a functional expression in which the processor 222 extracts the main object and the subtopic according to the control instructions described above.

A closer look at the components of processor 222 is as follows.

The MainObj + SoftwareExtractor (extraction module) of the extraction unit 410 extracts the main objects and attributes by subject. Natural Language Understanding techniques may be used to extract the main objects and attributes.

The SubObj Extractor (extraction module) of the extraction unit 410 extracts sub-objects when there is a main object + (attribute) for each subject. In order to extract the sub-object, various statistical information (for example, the number of clicks, the number of likes, the number of comments, the number of creators, etc.), dictionary information, and the like may be used.

The Ranker (Suffix Ranker, SubObj Ranker) module of the purification unit 420 determines the ranking of sub-objects and / or attributes with respect to the main object in order of importance. At this time, various information such as the number of clicks from the user, the number of likes, the number of comments, and the number of creators may be utilized to determine the importance.

The Post-Processor (Post-Processor Ranker, Post-Processor Ranker) module of the purification unit 420 receives the sub-object and / or attribute whose ranking is determined, and uses this based on synonyms and substring (substring) relationships. Select the representative value of each cluster by clustering.

The TopicGraphToTree module of the generator 430 collects clustered sub-objects and / or attributes, searches for the relationship strength of how much they are shown together in a query or document, generates this in a network (graph) structure, and then a tree ( Search / Cluster infrastructure tree) will be converted. The specific search / cluster infrastructure tree will be described below.

The Tree Constructor module of the generation unit 430 integrates the dictionary infrastructure tree and the search / cluster infrastructure tree structure to form a final topic tree (for example, a 2-depth tree structure).

The Topic learner module of the adjustment unit 440 additionally filters the topic tree according to the filtering conditions (for example, the number of documents, the goodness of fit of the subject, the correctness, etc.).

The New Object Assigner module of the coordinator 440 extracts new items related to the main object and assigns them to the tree structure. If there are more items than can be allocated to the tree structure, the process is started from the beginning to construct a new tree.

The Document Finding API module of the provider 450 constructs a query based on the final topic tree and extracts conforming documents. At this time, a filtering function may be included.

The Auto-Tagger module of the provider 450 configures topic tags based on the final topic tree and tags to conforming documents.

The processor 222 including the components described above may perform steps 510 to 550 as included in the method of FIG.

In FIG. 5, at step 510, the extraction unit 410 may extract a main object, which is a core object for determining the subject for each subject, and a subtopic that embodies the subject. At this time, the extraction unit 410 analyzes the words that frequently appear together with the main object in the document, or analyzes the words that are frequently used for the search together with the main object in the search system, thereby performing the sub-object and /. Alternatively, attribute candidates may be extracted.

At step 520, purification unit 420 may perform grouping based on the relationships between words after filtering subtopics by frequency of occurrence or search frequency in the document. In the candidate filtering process, at least a part of sub-objects and / or attribute candidates may be filtered and used. As a specific example, Purification Unit 420 may filter sub-objects and / or attributes by at least one of the frequency of occurrence in the document and the frequency of user searches. In addition, the frequency of appearance may be limited to data in a specific period and filtered. At this time, the filtering method may differ depending on the characteristics of the subject. For example, in the case of a subject with high demonstrability (timeliness), data for a certain period of recent years (for example, a period from the present to one week ago) May be filtered. Further, the purification unit 420 may group sub-objects and / or attribute candidates selected by the candidate filtering process in consideration of synonyms, sub-string relationships, and the like, and may select representatives from each group after grouping. There may be various methods for selecting a representative, and as one embodiment, the one with the highest search frequency may be selected as the representative. In other words, the purification unit 420 ranks the subtopics in order of importance (for example, frequency of appearance in the document, frequency of search, etc.), and then clusters the ranked subtopics by synonyms or substring relationships, so that each cluster A representative may be selected. The above-mentioned subtopic extraction and purification processes are performed in the candidate selection, grouping / representative selection process. FIG. 6 is a diagram showing an example of a process of filtering and grouping candidates and selecting a representative topic for the queries “Guam” and “potato” in one embodiment of the present invention. In FIG. 6, WTRIP and FOOD are classification codes (category classification codes) indicating the subject of the query, and the numbers next to the words indicate the frequency of searches related to the query. is there.

Again, in FIG. 5, at step 530, the generation unit 430 may generate a topic tree for the grouped subtopics using the hierarchical information of the subject. The generation unit 430 may generate a topic tree by labeling each grouped subtopic so as to match each class name in the hierarchy by utilizing the hierarchy information. Depending on the subject, there are cases where hierarchical information exists and cases where hierarchical information does not exist. For example, dictionary information constructed in a database containing various types of contents is one of the information that can be effectively used for hierarchical information. At this time, the generation unit 430 may generate a topic tree based on the hierarchical information that exists in advance, such as dictionary information. For example, the subject of a dish or recipe has a wealth of hierarchical information based on a culinary encyclopedia. On the other hand, the subject of travel and shopping does not have hierarchical information, and various subtopics are generated or changed depending on the time. When the hierarchical information does not exist, the generation unit 430 may construct the hierarchical information based on the word embedding-based clustering technique and the language taxonomy (taxonomy) and utilize it for generating the topic tree. The present invention has an advantage that topics can be automatically hierarchized even when hierarchical information does not exist.

FIG. 7 is a flowchart for explaining an example of a process of constructing hierarchical information using a word embedding-based clustering technique and language taxonomy in one embodiment of the present invention. Referring to FIG. 7, the generation unit 430 extracts similar words from the word embedding data for the subtopic (S701), clusters the extracted words based on synonyms and substring relationships (S702), and then clusters. The words given may be labeled based on linguistic taxonomy (S703). FIG. 8 is a diagram showing an example of hierarchical information constructed by using clustering and linguistic taxonomy for a travel subject. In the word embedding-based clustering process S702, word embedding data is learned as a thematic document (for example, a blog bulletin board), and after grasping the word vector value of a subtopic that requires clustering from such learning data, the word Clustering is performed based on the vector value. At this time, various methods such as hierarchical clustering, K-means algorithm, density clustering, and the like may be utilized for clustering. Further, in the language taxonomy utilization process S703, the clustered results may be labeled by mapping each class on the language taxonomy. At this time, since linguistic taxonomy is general-purpose, there are many unnecessary classes when comparing with hierarchical information specialized in a subject. Therefore, it is necessary to delete unnecessary classes, which will be described in the rebalancing process of the adjustment unit 440 described later.

In the topic hierarchy stage, clustered subtopics are collected, the relationship strength is searched for how much they are shown together in a query or document, generated in a network (graph) structure, and then converted into a topic tree (cluster infrastructure tree). The topic tree built on the dictionary base and the topic tree built on the clustering base may be integrated to form the final tree structure.

Again, in step 540, in step 540, the coordinating section 440 may perform rebalancing on the topic tree constructed in the topic hierarchy step of the generating section 430 according to the purpose of the user or the system. Further, the coordinating unit 440 may appropriately perform pruning on the topic tree in consideration of the subject goodness of fit, the search intention, the amount of search results, and the like.

Table 1 shows the definition of the topic network according to the embodiment of the present invention.

(Table 1)

生成部４３０は、検索頻度とクラスタリングを利用して構築した情報を利用してトピックツリーを生成する。先ず、各単語をノード（ｎｏｄｅ）とし、各単語同士の関連関係をエッジ（ｅｄｇｅ）で示してトピックネットワーク（Ｇ）を構築する。本発明の一実施形態に係るトピックネットワーク（Ｇ）におけるノード（Ｖ）およびエッジ（Ｅ）は、表１のように定義されてよい。このとき、生成部４３０は、検索頻度を考慮してトピックネットワークをトピックツリーに変更する。図９は、本発明の一実施形態によって構築されたトピックネットワークを深さが２であるツリー（２−デプスツリー）に変換する過程の例を示した図である。ネットワークをツリーに変更する方法としては多様なアルゴリズムが存在してよく、例えば、加重値グラフにおける最小全域木（ｍｉｎｉｍｕｍ ｓｐａｎｎｉｎｇ ｔｒｅｅ）構築アルゴリズムなどが応用されてよい。この後、調整部４４０は、言語分類学を基盤としたツリーと検索頻度／クラスタリングを基盤としたツリーとを合わせ、ユーザやシステムの目的に応じてリバランシングを実行してよい。図１０は、本発明の一実施形態における、ツリーリバランシング過程の一例を説明するためのフローチャートである。図１０を参照すると、調整部４４０は、クラスタリング基盤ツリーでリーフノードに該当するクラスタを辞書基盤ツリーの該当のクラスに挿入してよい（Ｓ１００１）。トピックツリーの幅（ｂｒｅａｄｔｈ）と深さ（ｄｅｐｔｈ）はクエリ別にそれぞれ異なり、ユーザが活用するものとしては概して深さおよび幅の両方の値が大きい傾向にあるため、これを減らす作業が必要となる（Ｓ１００２〜Ｓ１００３）。幅と深さを減らす方法は、言語分類学を基盤としたツリーと検索頻度／クラスタリングを基盤としたツリーとを結合する過程からなる。トピックツリーの幅値と深さ値は、システムの必要事項に応じて異なるように設定してよく、本発明の実施形態では深さが２であると仮定する（２ ｄｅｐｔｈ ｔｒｅｅ化）。この後、調整部４４０は、主題適合度、検索意図、検索結果の量などを考慮してトピックツリーのプルーニングを実行してよい（Ｓ１００４）。図１１は、ツリーリバランシング過程中に幅を減らすためのいくつかの方法を示すものであって、ボトムアップ（ｂｏｔｔｏｍ−ｕｐ）ノード移動、および／またはトップダウン（ｔｏｐ−ｄｏｗｎ）ノード移動によってトピックツリーの幅を減らしてよい。また、図１２は、ツリーリバランシング過程中に深さを減らすためのいくつかの方法を示すものであって、一部のノードを子ノードに交換する方式によってトピックツリーの深さを減らしてよい。

The generation unit 430 generates a topic tree by using the information constructed by utilizing the search frequency and clustering. First, each word is set as a node, and the relationship between each word is indicated by an edge to construct a topic network (G). Nodes (V) and edges (E) in the topic network (G) according to one embodiment of the present invention may be defined as shown in Table 1. At this time, the generation unit 430 changes the topic network to the topic tree in consideration of the search frequency. FIG. 9 is a diagram showing an example of a process of converting a topic network constructed by one embodiment of the present invention into a tree having a depth of 2 (2-depth tree). Various algorithms may exist as a method of changing the network into a tree, and for example, a minimum spanning tree construction algorithm in a weighted graph may be applied. After that, the coordinating unit 440 may combine the tree based on language taxonomy and the tree based on search frequency / clustering and execute rebalancing according to the purpose of the user or the system. FIG. 10 is a flowchart for explaining an example of the tree rebalancing process in one embodiment of the present invention. Referring to FIG. 10, the coordinating unit 440 may insert the cluster corresponding to the leaf node in the clustering infrastructure tree into the corresponding class in the dictionary infrastructure tree (S1001). The width and depth of the topic tree are different for each query, and users generally tend to have large values for both depth and width, so it is necessary to reduce this. (S1002 to S1003). The method of reducing width and depth consists of combining a tree based on linguistic taxonomy with a tree based on search frequency / clustering. The width value and the depth value of the topic tree may be set differently according to the requirements of the system, and it is assumed that the depth is 2 in the embodiment of the present invention (2 depth tree). After that, the adjustment unit 440 may execute the pruning of the topic tree in consideration of the subject goodness of fit, the search intention, the amount of search results, and the like (S1004). FIG. 11 shows some ways to reduce the width during the tree rebalancing process, by topic by bottom-up node movement and / or top-down node movement. You may reduce the width of the tree. In addition, FIG. 12 shows some methods for reducing the depth during the tree rebalancing process, and the depth of the topic tree may be reduced by exchanging some nodes for child nodes. ..

Again in FIG. 5, at step 550, the provider 450 may utilize the topic tree of the subject to which the query belongs and provide subtopics with search results corresponding to the query. At this time, the providing unit 450 may provide the subtopic as a related search term for the query together with the search result after filtering it according to various conditions. As an example, the provider 450 may filter subtopics by the goodness of fit of the query. When a query for searching is given, it will check the subject to which the query belongs, but at this time, if the query belongs to multiple subjects, if you want to filter topics that do not match the given subject There is. For this purpose, the subject score of "query + subtopic", which is a score indicating the relationship between the query and the subtopic, may be utilized. As a method for grasping the subject score, a text classification algorithm (for example, SVM (support vector machine), kNN (k-Nearest Neighbor), CNN (Convolutional Neural Network), etc. may be used). As another example, the provider 450 may filter the subtopics by using the number of documents corresponding to the subtopics. If the number of documents included in the search results is as small as a certain number or less, the usefulness decreases, so the relevant subtopic may be excluded. As another example, the provider 450 may filter the subtopics according to the presence or absence of correctness topics for the query. If providing correct information is a more appropriate subtopic than providing more documents as a search result (for example, when correct information is required, such as the weather in Guam), then for the query It may be included as a related search term.

The provider 450 may display detailed subtopics (sub-objects and / or attributes) related to the query as related search terms in a hierarchical manner for the query entered by the user for the search. The topic tree for each subject may be updated in units of a fixed cycle, and the update cycle of the topic tree may be determined in consideration of the characteristics of the subject according to the subject.

13 and 14 are diagrams showing an example of a search result screen in which a topic structure of 2 depths is reflected in one embodiment of the present invention. For example, in the process of providing the search result corresponding to the input query for the query input by the user, the query of depth1 and the query of depth2 are used as the related search terms of the input query based on the hierarchical information of the subject corresponding to the input query. May be provided. Referring to FIG. 13, when the user inputs the query "Guam" in the search window 1301, the related search of the input query "Guam" is performed based on the hierarchical information of the corresponding subject "Guam" together with the search result corresponding to the input query. Query 1310 of depth1 and query 1320 of depth2 may be provided as terms. At this time, each of the queries provided as related search terms is configured in a form that can be selected by the user, and the query selected by the user is automatically additionally input to the search window 1301. As shown in FIG. 13, when the user selects "gourmet shop" from the query 1310 of depth1 provided as the related search term of the initial query "Guam", "gourmet shop" is added to the search window 1301. The input and the search result 1302 of depth1 using "Guam gourmet restaurant" as a query may be displayed. Next, as shown in FIG. 14, when the user selects the query "handmade hamburger" from the query 1320 of depth2, "handmade hamburger" is additionally input to the search window 1301 and "Guam gourmet shop handmade hamburger". The search result 1402 of depth2 with the query "" may be displayed.

As described above, according to the embodiment of the present invention, the search results (documents) are provided together with the topic of the hierarchical structure to support efficient additional search, and the maximum variety of search results are provided as one query. Can be done. Furthermore, the hierarchical topic structure can be used for search ranking. In other words, documents containing sub-objects and attributes are likely to be relatively high quality documents and can be used to boost such documents during search rankings.

The devices described above may be implemented by hardware components, software components, and / or combinations of hardware components and software components. For example, the devices and components described in the embodiments include a processor, a controller, an ALU (arithmetic logic unit), a digital signal processor, a microcomputer, an FPGA (field program gate array), a PLU (programmable log unit), a microprocessor, and the like. Alternatively, it may be implemented using one or more general purpose computers or special purpose computers, such as various devices capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on said OS. The processor may also respond to software execution, access data, and store, manipulate, process, and generate data. For convenience of understanding, one processing device may be described as being used, but one of ordinary skill in the art may indicate that the processing device may include multiple processing elements and / or multiple types of processing elements. You can understand. For example, the processing unit may include multiple processors or one processor and one controller. Other processing configurations, such as parallel processors, are also possible.

The software may include computer programs, codes, instructions, or a combination of one or more of these, configuring the processing equipment to perform the desired operation, or instructing the processing equipment independently or collectively. You may do it. The software and / or data is embodied in any type of machine, component, physical device, computer storage medium or device to be interpreted based on the processing device or to provide instructions or data to the processing device. Good. The software is distributed on a computer system connected by a network and may be stored or executed in a distributed state. The software and data may be stored on a recording medium readable by one or more computers.

The method according to the embodiment may be implemented in the form of program instructions that can be executed by various computer means and recorded on a computer-readable medium. At this time, the medium may be a continuous recording of a computer-executable program, or a temporary recording for execution or download. Further, the medium may be various recording means or storage means in the form of a combination of a single piece of hardware or a plurality of pieces of hardware, and is not limited to a medium directly connected to a certain computer system, and is distributed on a network. It may exist. Examples of media include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, opto-magnetic media such as Floptic disks, and ROMs, RAMs. , Flash memory, etc., and may be configured to record program instructions. In addition, other examples of media include recording media or storage media managed by application stores that distribute applications, sites that supply or distribute various other software, servers, and the like.

As described above, the embodiments have been described based on the limited embodiments and drawings, but those skilled in the art will be able to make various modifications and modifications from the above description. For example, the techniques described may be performed in a different order than the methods described, and / or components such as the systems, structures, devices, circuits described may be in a form different from the methods described. Appropriate results can be achieved even if they are combined or combined, or confronted or replaced by other components or equivalents.

Therefore, even if the embodiment is different, if it is equivalent to the claims, it belongs to the attached claims.

222: Processor 410: Extraction unit 420: Purification unit 430: Generation unit 440: Adjustment unit

Claims (11)

It is a topic structuring method realized by a computer.
The stage of extracting subtopics related to the subject by subject,
When a stage for generating a topic tree using the hierarchical information of the subject and a query for searching are given to the subtopic, a related search term for the query is given based on the topic tree of the subject to which the query belongs. the sub-topic viewing including the step of providing a hierarchical form as,
The generation stage is
The stage of constructing a topic network in which each word corresponding to the subtopic is used as a node and the relationship between words is shown as an edge, and
The stage of generating the topic tree by changing the topic network to a tree structure in consideration of the search frequency
How to structure topics , including . The extraction step is
The topic structuring method according to claim 1, wherein the subtopic is extracted by analyzing words related to the core object that determines the subject. The topic structuring method according to claim 1, further comprising filtering the subtopic by at least one of a document appearance frequency and a search frequency. The topic structuring method according to claim 1, further comprising a step of clustering the subtopics by synonyms or substring relationships to select a representative of each cluster. The generation stage is
The topic structuring method according to claim 1, wherein the subtopic is labeled with each class name of the hierarchical information to generate the topic tree. The generation stage is
The stage of extracting similar words from the word embedding data for the subtopic,
The topic structuring method according to claim 1, comprising a step of clustering the similar words by synonyms or substring relationships, and a step of labeling the clustered words by mapping them to each class in linguistic taxonomy. The topic structuring method of claim 1, further comprising reducing at least one of the width and depth of the topic tree to perform rebalancing against the topic tree. The stage to be provided is
A claim including a step of filtering the subtopic according to at least one condition of the subject score indicating the relevance of the query and the subtopic, the number of documents corresponding to the subtopic, and the correctness topic for the query. The topic structuring method according to Item 1. It is a search result providing method realized by a computer.
When a query for search is given, the stage of providing search results corresponding to the query,
Based on the hierarchical information of the subject to which the query belongs, a step of providing a subtopic related to the subject in a hierarchical form having a plurality of depths as a related search term for the query, and at least one search term of the subtopic. If There is selected, viewing including the step of providing a search result corresponding to the query that contains the selected search word,
The stage of providing subtopics related to the subject in a hierarchical manner with multiple depths is
For the subtopic, the stage of generating a topic tree using the hierarchical information of the subject, and
The stage where the subtopic is provided hierarchically by the topic tree
Including
The generation stage is
The stage of constructing a topic network in which each word corresponding to the subtopic is used as a node and the relationship between words is shown as an edge, and
The stage of generating the topic tree by changing the topic network to a tree structure in consideration of the search frequency
How to provide search results , including . A computer program recorded on a computer-readable recording medium to combine with a computer system to perform topic structuring methods.
The topic structuring method is
The stage of extracting subtopics related to the subject by subject,
When a stage for generating a topic tree using the hierarchical information of the subject and a query for searching are given to the subtopic, a related search term for the query is given based on the topic tree of the subject to which the query belongs. the sub-topic viewing including the step of providing a hierarchical form as,
The generation stage is
The stage of constructing a topic network in which each word corresponding to the subtopic is used as a node and the relationship between words is shown as an edge, and
The stage of generating the topic tree by changing the topic network to a tree structure in consideration of the search frequency
Including computer programs. A topic-structured system realized by a computer
Includes at least one processor implemented to execute computer-readable instructions
The at least one processor
Extraction unit that extracts subtopics related to the subject by subject,
When a generation unit that generates a topic tree using the hierarchical information of the subject and a query for searching are given to the subtopic, a related search for the query is performed based on the topic tree of the subject to which the query belongs. It has a provider that provides the subtopic as a word in a hierarchical manner .
The generator
After constructing a topic network in which each word corresponding to the subtopic is used as a node and the relationship between words is shown as an edge, the topic tree is changed to a tree structure in consideration of the search frequency to form the topic tree. Generate,
Topic structuring system.

Images