KR102150560B1 - Apparatus and method for analyzing target using topic - Google Patents

Abstract

The present invention proposes a target analysis device and method for analyzing targets related to an event through a community. An apparatus according to the present invention includes a message processing unit for collecting and preprocessing text messages related to an event; A keyword detection unit for detecting topics as keywords from documents related to text messages based on weights assigned to words included in the text messages as having meanings; And a target derivation unit for deriving targets by age and year by performing social network analysis on topics.

Description

Target analysis device and method using topic {Apparatus and method for analyzing target using topic}

The present invention relates to an apparatus and method for analyzing a target. More specifically, it relates to an apparatus and method for analyzing a target using a topic.

Stress is a non-specific biological reaction that occurs in the body to various injuries and stimuli inflicted on the living body, and refers to the feelings of anxiety and threats that humans feel when they are in a situation that is difficult to handle psychologically or physically.

However, if such stress is accumulated in the body without being resolved, fundamental damage to human health may occur, and this may lead to exposure to various diseases such as heart disease, obesity, diabetes, cancer, etc., or death such as suicide.

Korean Patent Publication No. 2017-0115037 (Publication date: 2017.10.16.)

The present invention has been devised to solve the above problem, and detects topics based on weights in text messages collected through a community, and performs social network analysis (SNA) on these topics. It is an object of the present invention to propose a target analysis apparatus and method for deriving targets such as stress factors by executing.

However, the object of the present invention is not limited to the above-mentioned matters, and other objects not mentioned will be clearly understood by those skilled in the art from the following description.

The present invention, as conceived to achieve the above object, is a message processing unit for collecting and pre-processing text messages related to events; A keyword detection unit that has meaning and detects topics as keywords from documents related to the text messages based on weights assigned to terms included in the text messages; And a target derivation unit for deriving targets for each age by executing a social network analysis on the topics.

In addition, the present invention comprises the steps of collecting and preprocessing text messages related to events; Detecting topics as keywords from documents related to the text messages based on weights assigned to terms included in the text messages as having meaning; And deriving targets for each age group by performing a social network analysis on the topics.

Further, the present invention proposes a computer program stored in a computer-readable medium for executing a target analysis method in a computer.

The present invention can obtain the following effects through configurations for achieving the above object.

First, it becomes possible to effectively derive targets by age and year.

Second, it becomes possible to propose a plan for solving social problems by analyzing the correlation between targets.

1 is a conceptual diagram schematically showing a target analysis device according to an embodiment of the present invention.
2 is a block diagram schematically showing an internal configuration of a data collection and preprocessor constituting a target analysis device.
3 is a reference diagram for explaining functions of a data collection and preprocessing unit provided in a target analysis device.
4 to 7 are reference diagrams for explaining a function of a topic extraction unit provided in the target analysis device.
8 to 12 are reference diagrams for explaining the function of a topic-based network analysis unit provided in the target analysis device.
13 is a conceptual diagram schematically showing the internal configuration of a target analysis device according to a preferred embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. First, in adding reference numerals to elements of each drawing, it should be noted that the same elements are to have the same numerals as possible even if they are indicated on different drawings. In addition, in describing the present invention, if it is determined that a detailed description of a related known configuration or function may obscure the subject matter of the present invention, a detailed description thereof will be omitted. In addition, a preferred embodiment of the present invention will be described below, but the technical idea of the present invention is not limited or limited thereto, and may be modified and variously implemented by those skilled in the art.

The present invention relates to an apparatus for analyzing a target related to an event using a topic, and in one embodiment, related to an apparatus for analyzing a stressor related to a life event according to age. will be. In the present invention, a social network analysis technique (SNA), for example, a topic-based social network analysis technique (topic-based SNA), is used to analyze a target. Hereinafter, the present invention will be described in detail with reference to the drawings.

1 is a conceptual diagram schematically showing a target analysis device according to an embodiment of the present invention.

Referring to FIG. 1, the target analysis apparatus 100 includes a data collection and preprocessing unit 110, a topic extraction unit 120, and a topic-based network analysis unit 130. Includes.

The data collection and preprocessor 110 collects text messages and performs a function of preprocessing these text messages. For this purpose, the data collection and preprocessing unit 110 includes a message collection unit 111, a non-standard language removal unit 112, a text message classification unit 113, and a vector space modeling (VSM) execution unit ( 114) may be included. 2 is a block diagram schematically showing an internal configuration of a data collection and preprocessor constituting a target analysis device.

The message collection unit 111 performs a function of collecting text messages from online communities such as SNS (Social Network Services), cafes, and blogs. In the present invention, for example, in order to analyze the stress factors of women related to life events, the message collection unit 111 may collect text messages from an online community frequently used by women. The message collection unit 111 stores the collected messages in a memory provided in advance.

The non-standard language removal unit 112 includes meaningless characters, unimportant words, prefixes, and suffixes in each text message collected by the message collection unit 111. ) And other non-standard languages (removing non-standard language). The data collection and preprocessor 110 may obtain a list of words suitable to represent the contents of each message through this function of the non-standard language removal unit 112.

When the non-standard language is removed from the text message by the non-standard language removal unit 112, the text message classifying unit 113 divides the sentences included in the text message into terms or morphemes. Functions. The text message classifier 113 may use a POS (Part-Of-Speech) tagging technique to perform a tokenization function.

The POS tagging technique is a technique that identifies the parts of speech of words included in a sentence and attaches tags. The text message classifier 113 may classify sentences into tokenized terms such as nouns, verbs, and adjectives using a POS tagging technique. In addition, the text message classifier 113 may output tokenized words in the form of a tuple using a POS tagging technique.

The VSM execution unit 114 converts words tokenized by the text message classifier 113 into normalized weighted terms using a vector space modeling (VSM) technique. It performs the function to make. In the above, the vector space modeling technique refers to an information retrieval technique in which words and messages are expressed as vectors.

When converting the tokenized words into normalized and weighted words, the VSM execution unit 114 displays tokenized words and text messages containing the words in a matrix form as shown in FIG. 3. I can. 3 is a reference diagram for explaining the function of the VSM execution unit 114.

3 is a schematic presentation of a VSM. As shown in FIG. 3, the VSM execution unit 114 includes m tokenized words (m-terms) and n text messages (n-documents) in a matrix (X _t,d ) arranged in rows and columns, respectively. ) Can be expressed. Through this, the VSM execution unit 114 uses m tokenized words (Term 1, Term 2, Term 3, …, Term) in n text messages (Document 1, Document 2, Document 3, …, Document N). It becomes possible to calculate the frequency of M).

The VSM execution unit 114 performs a function of assigning a weight to each word based on the frequency of words representing each text message using a vector space modeling technique. The VSM execution unit 114 may assign a weight to each word based on a method of calculating a Term Frequencies (TF) value and a method of calculating an Inverse Document Frequencies (IDF) value. The method of calculating the TF value refers to a method of calculating the total frequency of words appearing in a specific text message, and the method of calculating the IDF value refers to a method of calculating the total frequency of specific words appearing in the entire text message set. The VSM execution unit 114 can obtain an effect of preventing a high weight from being assigned to words that are common in many text messages through a method of calculating a TF value and a method of calculating an IDF value.

The VSM execution unit 114 performs a function of normalizing each text message when a weight is given to each word. Since all text messages in the matrix X _t,d of FIG. 3 must have the same significance, the VSM execution unit 114 performs a normalization function for each text message in consideration of this point in the present invention.

The VSM execution unit 114 may calculate a normalization value of each word to which a weight is assigned through Equation 1 below for normalization of the text message.

In the above, w _id denotes a weighted term value of a word for each term-document combination. Also, n means the number of documents.

The topic extraction unit 120 performs a function of performing a topic analysis to extract keywords from text messages preprocessed by the data collection and preprocessing unit 110. That is, the topic extracting unit 120 extracts messages related to the target by analyzing the meaning contained in the text messages, and extracts words having a weight greater than or equal to a reference value from the extracted messages as keywords.

For example, in order to analyze women's stress related to life events, the topic extraction unit 120 identifies patterns and meanings of semantic construction hidden in the text from women's online activities. Keywords can be extracted according to the recurring pattern of co-occurring words to understand the stress of

The topic extraction unit 120 executes a Latent Semantic Analysis (LSA) algorithm in order to extract main topics based on the words when weights are given to words by the VSM execution unit 114. Functions. That is, the topic extraction unit 120 classifies the text image in a text image-word space to reveal the semantic structures of the corpus, and based on this, the topics contained in the text image (that is, potential It performs a function of extracting latent meanings.

In addition, the topic extraction unit 120 also performs a function of removing a portion of the text image that does not include meaningful elements in order to easily identify and classify text images based on topics.

A generalized model for the content of a text image can be expressed as follows.

X = ΛF + ε

In the above, X denotes a corpus of text images, and ΛF denotes a linear combination of topics in large-capacity text data. In addition, ε means an error term.

The latent semantic analysis (LSA) algorithm performed by the topic extraction unit 120 includes execution of Singular Value Decomposition (SVD) and dimensionality reduction.

The singular value decomposition (SVD) process mathematically decomposes the matrix X _t,d by the product of three matrices such as U _t,k , Σ _k,k , and V _k,d as shown in FIG. 4. 4 is a first reference diagram for explaining a function of a topic extraction unit provided in a target analysis device.

4 is a diagram schematically showing a singular value decomposition (SVD) procedure. In FIG. 4, U _t,k denotes a term-by-factor matrix including eigenvectors of the XX ^T matrix. Σ _k,k denotes a term-by-term covariance matrix representing potential semantic elements of words. These Σ _k,k can be expressed as a diagonal matrix of singular values ranked in decreasing order. V _k,d denotes a factor-by-document matrix showing the eigenvectors of X ^T X. In the above, X ^T X denotes a document-by-document covariance matrix showing the loading of each document for each element. Document loadings indicate the degree of relevance of the document to the identified topics.

Singular value decomposition (SVD) is expressed according to mathematical notation as follows.

In the above, U means eigenvectors of a word, and V means eigenvectors of a document. In addition, Σ denotes a diagonal matrix of singular values including square roots of eigenvectors, and T denotes transposition.

The topic extracting unit 120 may modify term frequencies including only very important words using TSVD (Truncated SVD). The TSVD process is very similar to Principal Component Analysis (PCA) and involves orthogonal transformation, which transforms correlated factors into a set of principal components.

The TSVD model is related to the dimensionality reduction of X _t,d by recognizing that topics are organized in a low-dimensional space in a large corpus of text data. The threshold of eigenvalues is generally set to 1 or more. TSVD effectively solves the problem of synonymy and polysemy of words and has a significant influence on inquiry performance. If TSVD is expressed in terms of mathematical notation, it is as follows.

Here, ^{^} X _t,d denotes a term-by-document matrix related to words and documents, and U _t,k denotes a truncated version of words by elements. Also, Σ _k,k denotes a truncated factor-by-factor matrix solution, and V ^T _k,d denotes a truncated matrix of elements by documents. k means ranks of singular values.

5 is a second reference diagram for explaining the function of a topic extraction unit provided in the target analysis device. 5 illustrates the operation of TSVD in the Latent Semantic Analysis (LSA) algorithm. In FIG. 5, reference numerals 210 to 230 denote portions cut at U _t,k , Σ _k,k and V ^T _k,d , respectively.

Orthonormality U ^T U = I and V ^T V = I (where I is a k * k identity matrix) is defined as term loadings (L _T ) and document loadings; It is used to obtain L _D ). This can be expressed as an equation as follows.

In the above, L _T denotes a term-by-factor matrix representing the association between words and latent topics, and L _D denotes the relationship between documents with potential topics. It refers to a document-by-term matrix that is represented.

When the optimal number of dimensions is detected and text data is condensed through TSVD, the topic extraction unit 120 improves the interpretability of initial results. To do this, a post-LSA quantitative analysis is performed using varimax rotations. 6 shows word loadings and document loadings in an articulated factor space. 6 shows term-rotated loading operations and document-rotated loading operations and their outputs.

In the present invention, serialized topic models exist. Each topic consists of a set of distributions of words (keywords). All words in the document are ranked according to the amount of distribution. The term topic weight is assigned a weight corresponding to each topic. When 25 topics are extracted, 25 word topic weights are calculated for each word. Since each topic is of SVD dimension, word topic weights represent the coordinate value of the word in SVD space. 7 is an exemplary diagram showing topics extracted in relation to a stress factor of a woman in her 30s and weights of the topics.

In FIG. 7, a term cutoff refers to a threshold score that determines whether a word belongs to a topic, and a document cutoff refers to a threshold score that determines whether a document belongs to a topic. it means. Words and documents are assigned as topics (topic labeling) when they have a word topic weight and a document topic weight greater than these cutoffs.

The topic-based network analysis unit 130 performs a topic-based social network analysis (topic-based SNA) to visualize targets by age. The topic-based network analysis unit 130 also performs a function of identifying the relevance of topics based on similarity related to relational attributes between topics.

Social Network Analysis (SNA) refers to the process of investigating social structures based on networks and graph theory. Social network analysis includes nodes (individual actors, people, things, etc.), ties, edges, or links connecting them (relationships). , Interactions, etc.). Social network analysis focuses on the structure of ties within a set of social actors, such as people, groups, organizations, countries or persons performing human activity. It has characteristics.

The topic-based network analysis unit 130 performs a topic-based social network analysis (topic-based SNA) that focuses on their relationship structure within a set of stress factors in order to explain a target (ex. stress factors of women). . This topic-based network analysis unit 130 is to enable discovery of relationships between targets and overall patterns of targets (ex. stress factors). Visualizing a topic-based network may be performed.

The topic-based network analysis unit 130 is a useful tool for refining and arranging topics and sequentially performs three analytic produres.

As a first analysis procedure, the topic-based network analysis unit 130 performs a function of discovering and visualizing comprehensive patterns of the green network (corresponding to the third procedure) using the cosine similarity coefficient.

To this end, the topic-based network analysis unit 130 calculates cosine similarities to determine similarities between topics T _x and T _y . This can be expressed as an equation as follows.

In the above, tft means a term (keyword) frequency of a word (keyword) in a topic. For each of the topics, the degree of similarity is calculated between a specific topic included in a specific sample and other topics included in other samples.

As a second analysis procedure, the topic-based network analysis unit 130 is a topic-based, expressed by nodes (a set of serialized topics) and links (a set of degrees between nodes). Perform social network analysis. The topic-based network analysis unit 130 may generate a sociomatrix X(N, N) through this. Sociomatrix X (N, N) is a basic matrix type of social network data, and can be generated in the form of N * N using daily topics and cosine similarities. have. 8 is an exemplary diagram of a sociomatrix for topic-based social network analysis.

As a third analysis procedure, the topic-based network analysis unit 130 performs a function of configuring a weighted network. The weighted network is composed of weighted degree centrality of nodes to which weight is assigned. This value is defined as the sum of the weights of the connected links, and is used as the node size in the present invention.

The topic-based network analysis unit 130 may calculate the connection degree centrality of the nodes to which the weight is assigned using the following equation.

In the above, w denotes a weighted adjacency matrix, and w _ij has a value greater than 0 when node i is connected to node j.

When performing a topic-based social network analysis, the topic-based network analysis unit 130 may visualize topics by age and perform the above function. This will be described below.

9 is a table diagram of stress factors according to ages labeled as topics. The topic-based network analysis unit 130 may visualize by arranging nodes (topics) according to a predetermined criterion as shown in FIG. 10 based on the topics (stress factors) arranged in FIG. 9. . The topic-based network-based analysis unit 130 calculates cosine similarities based on Equation 5 in order to analyze the structure of topics related to the stress factor, and then generates a sociomatrix to be used as input data for topic-based social network analysis. I can.

When visualizing nodes as shown in FIG. 10, the topic-based network analysis unit 130 may visualize nodes by arranging them on different axes for each age. For example, the topic-based network analysis unit 130 includes topics related to 20s, topics related to 30s, topics related to 40s, and 50s on the Y+ axis, X+ axis, Y-axis and X-axis. You can arrange topics.

The topic-based network analysis unit 130 applies the same color to the same topic, adjusts the size of the node according to the weight (ex. stress level) of each topic, and visually displays targets for each age (ex. stress factor) on the graph. Can be displayed. In addition, the topic-based network analysis unit 130 may visualize by arranging nodes according to year order with 0 as a starting point. In addition, the topic-based network analysis unit 130 may connect nodes having the same color to each other to visualize the relationship between different ages and when certain targets appear frequently by age.

As shown in FIGS. 11 and 12, the topic-based network analysis unit 130 may visualize targets (ex. stress factors) by arranging nodes related to a specific age in a circular shape according to the year order. 11 and 12 are visualizations of stress factors related to their 30s and 40s, respectively, arranged in a circular shape according to the year order.

FIG. 11(a) shows all the stress factors related to the thirties, and FIG. 11(b) shows the stress factors having a weight greater than or equal to the threshold value among all the stress factors related to the thirties. Similarly, (a) of FIG. 12 shows all the stress factors related to the 40s, and (b) of FIG. 12 shows the stress factors having a weight greater than or equal to the threshold value among all the stress factors related to the 40s.

In FIGS. 11 and 12, the node may be represented in the form of age_year_detailed topic. For example, when there is a work life balance as a topic of the age of 35 in 2018, this node may be represented in the form of 35_2018_work life balance in (a) and (b) of FIG. 11.

According to (a) of FIG. 11, as a stress factor in their 30s, family (27.27%) was the highest, followed by work (24.24%) and marriage (Marriage, 15.15%). From this, it can be seen that the stressors in their 30s are related to forming a new family through marriage, raising children, and working to support parents. In the case of love, as a stress factor, it was 40.91% in their 20s, but significantly decreased to 3.03% in their 30s, indicating that in their 30s, the marriage-related stress turned into a major stress rather than dating as it became a marriageable age. have.

Referring to FIG. 11B, node 1 (14_4_mother in law), node 2 (16_5_parenting) and node 3 (09_3_work life balance) are interconnected. This is because it is difficult for a woman in her 30s to maintain a work life balance while working and life (child rearing) after marriage, and conflicts that arise when family members such as mother-in-law raise children instead are a major stress factor in their 30s. It indicates that you are holding it.

One embodiment of the present invention has been described above with reference to FIGS. 1 to 12. Hereinafter, a preferred embodiment of the present invention that can be inferred from such an embodiment will be described.

13 is a conceptual diagram schematically showing the internal configuration of a target analysis device according to a preferred embodiment of the present invention.

According to FIG. 13, the target analysis device 300 includes a message processing unit 310, a keyword detection unit 320, a target derivation unit 330, a node display control unit 340, a power supply unit 350, and a main control unit 360. do.

The power supply unit 350 performs a function of supplying power to each component constituting the target analysis device 300.

The main control unit 360 performs a function of controlling the entire operation of each component constituting the target analysis device 300.

The message processing unit 310 performs a function of collecting and pre-processing text messages related to an event. The message processing unit 310 is a concept corresponding to the data collection and preprocessing unit 110 of FIG. 1.

The message processing unit 310 extracts words containing meaning from sentences included in text messages, assigns weights to these words, and normalizes words to which weights are assigned to preprocess text messages. have.

The message processing unit 310 may extract words from sentences using tokenization in which tags are attached by identifying parts of speech of words.

The message processing unit 310 may assign weights to words based on a first frequency of each word appearing in one text message and a second frequency of each word appearing in all text messages.

The message processing unit 310 may normalize words based on a weight assigned to each word and the number of text messages.

The keyword detection unit 320 detects topics as keywords from documents related to text messages based on weights assigned to terms included in text messages as having meaning. Performs the function of The keyword detection unit 320 is a concept corresponding to the topic extraction unit 120 of FIG. 1.

The keyword detection unit 320 uses SVD (Singular Value Decomposition) using a first matrix related to the relationship between words, a second matrix related to the relationship between documents, and a third matrix related to the relationship between words and documents, and Topics may be detected as keywords based on TSVD (Truncated SVD) using the first frequency of each word.

The keyword detector 320 may detect topics among words based on cutoffs related to weights.

The keyword detector 320 may sequentially apply SVD, TSVD, and varimax rotation to documents, and then detect by labeling predetermined information on topics based on cutoffs.

The target derivation unit 330 performs a function of deriving targets for each age by executing a social network analysis on topics. The target derivation unit 330 is a concept corresponding to the topic-based network analysis unit 130 of FIG. 1.

The target derivation unit 330 may calculate a cosine similarity related to a similarity between topics based on the first frequency of the topics, and may derive targets by performing a social network analysis based on the cosine similarity. .

The target derivation unit 330 generates a sociomatrix based on the cosine similarity related to the similarity between the topics and the relationship information between the topics, and performs a social network analysis based on the sociomatrix. Can be derived.

The target analysis device 300 may further include a node display control unit 340.

The node display control unit 340 performs a function of visually displaying nodes by arranging nodes related to targets for each age on each axis of the graph according to the year order. The node display control unit 340 is a concept corresponding to the topic-based network analysis unit 130 of FIG. 1.

The node display controller 340 may visually display the nodes on a graph by applying nodes of the same color to the same target and adjusting the size of the nodes based on weights related to each topic.

Next, a method of operating the stress factor analysis device 300 will be described.

First, the message processing unit 310 collects and preprocesses text messages related to the event (STEP A).

Thereafter, the keyword detection unit 320 uses the weights given to terms included in the text messages as keywords that contain meanings from documents related to text messages. Detect (STEP B).

Thereafter, the target derivation unit 330 derives targets for each age by executing a social network analysis on the topics (STEP C).

Meanwhile, after STEP C, the node display control unit 340 may visually display nodes by arranging nodes related to targets for each age on each axis on the graph according to the year order (STEP D).

Even if all the components constituting the embodiments of the present invention described above are described as being combined into one or operating in combination, the present invention is not necessarily limited to these embodiments. That is, within the scope of the object of the present invention, all the constituent elements may be selectively combined and operated in one or more. In addition, although all of the components may be implemented as one independent hardware, a program module that performs some or all functions combined in one or more hardware by selectively combining some or all of the components. It may be implemented as a computer program having In addition, such a computer program is stored in a computer readable media such as a USB memory, a CD disk, a flash memory, etc., and is read and executed by a computer, thereby implementing an embodiment of the present invention. The recording medium of the computer program may include a magnetic recording medium or an optical recording medium.

In addition, all terms, including technical or scientific terms, have the same meaning as commonly understood by those of ordinary skill in the art, unless otherwise defined in the detailed description. Terms generally used, such as terms defined in the dictionary, should be interpreted as being consistent with the meaning of the context of the related technology, and are not interpreted as ideal or excessively formal meanings unless explicitly defined in the present invention.

The above description is merely illustrative of the technical idea of the present invention, and those of ordinary skill in the technical field to which the present invention belongs can make various modifications, changes, and substitutions within the scope not departing from the essential characteristics of the present invention. will be. Accordingly, the embodiments disclosed in the present invention and the accompanying drawings are not intended to limit the technical idea of the present invention, but are for illustration, and the scope of the technical idea of the present invention is not limited by these embodiments and the accompanying drawings. . The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of the present invention.

Claims (17)

A message processing unit that collects and preprocesses text messages related to an event;
A keyword detection unit that has meaning and detects topics as keywords from documents related to the text messages based on weights assigned to terms included in the text messages; And
And a target derivation unit for deriving targets by age by executing a social network analysis on the topics,
The message processing unit extracts the words that have the meaning from sentences included in the text messages, assigns weights to the words, and normalizes the words to which weights are assigned to generate the text messages. Target analysis device, characterized in that pre-processing. delete The method of claim 1,
And the message processing unit extracts the words from the sentences using tokenization in which a part of speech of the words is identified and tagged. The method of claim 1,
The message processing unit assigns weights to the words based on a first frequency of each word appearing in one text message and a second frequency of each word appearing in all of the text messages. . The method of claim 1,
The message processing unit normalizes the words based on a weight assigned to each word and the number of text messages. The method of claim 1,
The keyword detection unit SVD (Singular Value Decomposition) using a first matrix related to the relationship between the words, a second matrix related to the relationship between the documents, and a third matrix related to the relationship between the words and the documents. ) And a TSVD (Truncated SVD) using the first frequency of each word, and detecting the topics as keywords. The method of claim 1,
And the keyword detection unit detects the topics among the words based on a cutoff related to the weights. The method of claim 7,
The keyword detection unit sequentially applies SVD, TSVD and varimax rotation to the documents, and then detects by labeling predetermined information on the topics based on the cutoff. Device. The method of claim 1,
The target derivation unit calculates a cosine similarity related to the similarity between the topics based on a first frequency of the topics, and performs the social network analysis based on the cosine similarity to derive the targets. Target analysis device characterized in that. The method of claim 1,
The target derivation unit generates a sociomatrix based on the cosine similarity related to the similarity between the topics and the relationship information between the topics, and executes the social network analysis based on the sociomatrix. Target analysis device, characterized in that deriving targets. The method of claim 1,
A node display control unit that visually displays the nodes by arranging nodes related to targets for each age on each axis on the graph in order of year
Target analysis device, characterized in that it further comprises. The method of claim 11,
The node display control unit applies nodes of the same color to the same target, adjusts a size of a node based on a weight associated with each topic, and displays the nodes visually on a graph. Pre-processing by a message processing unit collecting text messages related to an event;
The keyword detection unit detects topics as keywords from documents related to the text messages based on weights given to terms included in the text messages as having meaning. step; And
A target derivation unit deriving targets for each age by performing a social network analysis on the topics; includes,
In the pre-processing, the words containing the meaning are extracted from sentences included in the text messages, weights are assigned to the words, and the words to which weights are assigned are normalized to the text message. Target analysis method, characterized in that pre-processing. delete The method of claim 13,
In the deriving step, a cosine similarity related to a similarity between the topics is calculated based on a first frequency of the topics, and the targets are derived by executing the social network analysis based on the cosine similarity. Target analysis method, characterized in that. The method of claim 13,
The node display control unit visually displays the nodes by arranging the nodes related to targets for each age on each axis on the graph according to the year order
Target analysis method further comprising a. A computer program stored in a computer-readable medium for executing the target analysis method according to any one of claims 13, 15 and 16 on a computer.

Images