KR100880001B1 - Mobile device for managing personal life and method for searching information using the mobile device - Google Patents

Abstract

A mobile device capable of managing the every day of the individual and an information searching method in the mobile device are provided to search information related to a keyword inputted by a user. A mobile device(100) comprises the followings. An information collecting part(110) collects the log information showing the life pattern of a user. A semantic information extracting part(120) analyzes the collected log information, and then extracts the semantic information. An input unit(130) receives a keyword from the user. A storage(140) stores a probability model including the semantic information. A probability model updating unit(150) renews the probability model based on the collected log information and the extracted semantic information.

Description

Mobile device for managing personal life and a method for searching information on the mobile device {mobile device for managing personal life and method for searching information using the mobile device}

The present invention relates to a mobile device that can manage an individual's daily life, and more particularly, a mobile device capable of receiving a keyword from a user and searching information related to the keyword based on information collected through the mobile device. And a method for retrieving information in the mobile device.

Mobile devices such as digital cameras and mobile phones can collect various information such as call logs, photo shoots, music file playback, location information, and the like, and the user can always carry the user's daily information effectively. In addition, since the mobile device is a strong personality equipment, it can be adapted and specialized according to the preferences or preferences of the individual. As such, if the information collected through the mobile device is efficiently used, more various services may be provided to users who use the mobile devices as life recorders.

As part of such technology, techniques for collecting log information through a mobile device, modeling cognitive behavior of a user based on the collected log information, and detecting and predicting Landmark have been studied.

However, the technology of searching for personalized information to the user according to the user's query based on the information collected through the mobile device is not dealt with in detail.

The technical problem to be achieved by the present invention is a mobile device that can manage the daily life of the individual, which can retrieve information personalized to the user based on the information collected through the mobile device, and in the device An information retrieval method and a computer readable recording medium having recorded thereon a program for executing the information retrieval method are provided.

Furthermore, the technical problems to be achieved by the present invention are not limited to those mentioned above, and other technical problems that are not mentioned will be clearly understood by those skilled in the art from the following description.

In order to solve the above technical problem, a mobile device capable of managing a daily life of an individual according to the present invention comprises: an information collecting unit collecting log information representing a life pattern of a user; A storage unit for storing a probabilistic model including semantic information which is meaningful form of information related to at least some of the collected log information; An input unit to receive a keyword from a user; A searcher for searching for semantic information based on the probability model according to the received keyword; And a display unit for displaying the searched result.

The mobile device may further include a semantic information extractor configured to analyze the collected log information to extract semantic information that is meaningful information.

The mobile device may further include a probability model updating unit for updating the probability model according to the collected log information and the extracted semantic information.

The semantic information extracting unit may include: a rule analyzing unit configured to extract semantic information satisfying a specific rule by analyzing the collected log information; A statistical analysis unit for extracting semantic information representing a statistical meaning by analyzing the collected log information; And a probability analyzer extracting semantic information through the probability model based on the collected log information or the semantic information extracted from the rule analyzer or the statistical analyzer.

In addition, the probabilistic model is in the form of a modular Bayesian network, each module constituting the probabilistic model has a virtual node for reflecting probabilistic evidence, the modular Bayesian network has a hierarchical structure having level information for each node Can be.

In addition, the probability model updating unit receives the collected log information and the extracted semantic information, defines a set of nodes to be learned and levels of each node, calculates correlations between nodes, and ranks nodes according to the results. It can be characterized in that, and learning the Bayesian network structure. Here, the learning of the Bayesian network structure may be performed using a K2 learning algorithm.

The search unit may further include a keyword expansion unit generating a keyword that is the basis of the search by using the probability model and a predetermined rule according to the input keyword; And a semantic information search unit for searching semantic information according to the received keyword and the generated keyword. Here, the rule already defined may follow an ontology model indicating an association between semantic information.

The display unit may display location information corresponding to the searched semantic information on the map image when the searched semantic information is related to the location information.

In order to solve the other technical problem, an information retrieval method in a mobile device that can manage the individual's daily life according to the present invention, (a) collecting log information indicating the life pattern of the user; (b) receiving a keyword from a user; searching for semantic information based on a probabilistic model including semantic information that is meaningful information related to at least some of the collected log information according to the received keyword; And (d) displaying the searched results.

The information retrieval method may further include (a2) extracting semantic information that is meaningful information by analyzing the collected log information.

The information retrieval method may further include (a3) updating the probability model according to the collected log information and the extracted semantic information.

The step (a2) may include: (a21) extracting semantic information satisfying a specific rule by analyzing the collected log information; (a22) extracting semantic information representing a statistical meaning by analyzing the collected log information; And (a23) extracting semantic information through the probability model based on the collected log information or the semantic information extracted in the step (a21) or the step (a22).

In addition, the probabilistic model is in the form of a modular Bayesian network, each module constituting the probabilistic model has a virtual node for reflecting probabilistic evidence, the modular Bayesian network has a hierarchical structure having level information for each node Can be.

In addition, the step (a3) receives the collected log information and the extracted semantic information, defines the set of nodes to be learned and the level of each node, calculates the association between the nodes, and accordingly the phase of the nodes. Determining the rank and learning the Bayesian network structure. Here, the learning of the Bayesian network structure may be performed using a K2 learning algorithm.

In addition, the step (c) may further include generating a keyword that is the basis of the search by using the probability model and a predetermined rule according to the received keyword; And searching for semantic information according to the received keyword and the generated keyword. Here, the rule already defined may follow an ontology model indicating an association between semantic information.

In addition, in the step (d), when the searched semantic information is related to the location information, location information corresponding to the searched semantic information may be displayed on a map image.

In order to solve the another technical problem, a computer-readable recording medium having a program for executing the above-described information retrieval method is provided.

According to the present invention, information personalized to a user according to a user's query can be effectively searched in consideration of a user's real life pattern.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description and the accompanying drawings, the substantially identical components are represented by the same reference numerals, and thus redundant description will be omitted. In addition, in the following description of the present invention, if it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted.

1 is a block diagram of a mobile device capable of managing a daily life of an individual according to an embodiment of the present invention. The mobile device 100 according to the present embodiment is a portable or mobile digital device, and may include, for example, a laptop, a digital camera, a mobile phone, a smartphone, a portable multimedia player, a digital broadcast receiver, a vehicle navigation device, and the like.

The mobile device 100 according to the present embodiment collects data indicative of a user's life pattern, and organizes and provides the user's life pattern to the user. When a user inputs a predetermined keyword and commands a search through the mobile device 100, the user provides semantic information by using the collected data. In the present embodiment, the semantic information means not only collected data but also meaningful types of information related thereto, for example, relaxation, sleep, meal, study, exercise, school, school, school, class, nightlife, dinner, travel, mountain climbing, walking, A user's behavioral state such as shopping or eating out, or an emotional state such as joy, anger or irritation, or temporal or spatial information or related person information.

Referring to FIG. 1, the mobile device 100 according to the present exemplary embodiment may include an information collecting unit 110 collecting log information representing a user's life pattern, and analyzing collected log information to mean meaningful information. A semantic information extractor 120 for extracting the information, an input unit 130 for receiving a keyword from a user, a storage unit 140 for storing a probability model including semantic information, collected log information, and extracted semantic information. A probabilistic model updating unit 150 for updating the probabilistic model, a search unit 160 for searching for semantic information based on at least the probabilistic model according to the input keyword, and a display unit 170 for displaying the searched result. .

The information collecting unit 110 collects log information representing a life pattern of the user. For example, the information collecting unit 110 collects user location information. To this end, the information collecting unit 110 may include a global positioning system (GPS). The satellite navigation apparatus receives a coordinate value indicating the position of the user. In addition, the information collecting unit 110 may collect information such as weather, temperature, wind speed, and news through the web. In addition, the information collecting unit 110 may collect usage information of the mobile device 100, that is, log information related to call history, short message service (SMS) transmission, image capturing, and multimedia content reproduction. Specifically, when the user transmits a short message, the information collection unit 110 collects data such as the content of the short message, the short message receiver, and the time when the short message is transmitted. In the case of a call history, the information collecting unit 110 collects data on, for example, a call partner, a talk time, a call volume, and the like. In addition, when a music file (DMB, video file, etc.) of the multimedia content is reproduced, the information collecting unit 110, for example, the genre, title, artist name (actor) of the reproduced music file (DMB, video file, etc.) Information), the number of times of play, the time of play, etc. are collected. In addition, the information collecting unit 110 may collect charging state information of the mobile device 100. The type of information collected by the information collecting unit 110 may vary depending on a function or a sensor included in the mobile device 100.

The storage unit 140 converts the log information collected by the information collecting unit 110 into an appropriate form and stores the converted log information. In this case, a short term storage method may be used for simply collecting information, and a server having a large capacity connected to the mobile device 100 through a wireless or wired network for preservation of information already stored in the storage 140. After being moved to a non-shown, it can be stored using a long term storage method. As will be described later, the storage 140 stores semantic information extracted by the semantic information extractor 120 in addition to the log information, and also includes semantic information related to at least some of the extracted semantic information and the log information. Stochastic models that summarize the user's life patterns are stored.

The semantic information extractor 120 analyzes log information collected by the information collector 110 to extract semantic information. The extracted semantic information is stored in the storage 140. 2 is a block diagram illustrating a specific embodiment of the semantic information extractor 120. Referring to FIG. 2, the semantic information extractor 120 includes a rule analyzer 121, a statistical analyzer 122, and a probability analyzer 123.

The rule analyzer 121 analyzes the collected log information and extracts semantic information that satisfies a specific rule. For example, the rule analyzer 121 extracts semantic information about a place corresponding to a coordinate value represented by latitude and longitude when the collected log information is a coordinate value indicating a user's location, that is, GPS location information. Alternatively, when the collected log information is a phone number, for example, a call log from 010-XXX-XXXX, the rule analysis unit 121 may call 'call from a friend' according to a predefined rule for the call list. Extract semantic information such as 'call from stranger'.

The statistical analyzer 122 analyzes the collected log information and extracts semantic information representing a statistical meaning. For example, if the collected log information is a call volume, the statistical analysis unit 122 analyzes this, and when a call volume is higher than a normal value in a specific time period, the call frequency is high or the call frequency is very high. Generate semantic information.

The probability analysis unit 123 extracts semantic information through the probability model stored in the storage unit 140 based on the collected log information or the semantic information extracted from the rule analysis unit 121 or the statistical analysis unit 122. do. In this case, as the probabilistic model used, a probabilistic model previously constructed as an initial model is used, or a probabilistic model updated by the probabilistic model updater 150 described later is used.

Here, the probabilistic model refers to a model that can calculate how much probability the given information exists through the conditional probability association between the information. In this embodiment, a probability model of a Bayesian network type may be used as the probability model. Bayesian networks are in the form of directed acyclic graphs (DAGs) that represent the connectivity of nodes, and efficiently represent and represent many probability relationships by a conditional probability table (CPT) defined according to this structure. The model can be calculated. In a Bayesian network, each node represents a random variable, and an arc represents an association between each node. A detailed description of the Bayesian network probabilistic model is given in [K. B. Korb, and A. E. Nicholson, Bayesian Artificial Intelligence, Chapman & Hall / CRC, 2003]. In addition, in the present embodiment, the Bayesian network is preferably a hierarchical structure having level information for each node.

When the probability of the semantic information obtained through the probabilistic model exceeds or exceeds a predetermined reference value, the probability analysis unit 123 may regard the semantic information as semantic information representing a user's situation or behavior. 3 shows an example of a meal related probability model as an example of the Bayesian network probability model according to the present embodiment. Referring to FIG. 3, nodes related to a user's previous behavior, nodes related to time, nodes related to a place, and nodes related to a user's behavior are hierarchically structured. As shown in the drawing, nodes having a hierarchical structure can be largely divided into input nodes and output nodes. Here, the input node means a node that affects a specific output node, and the output node means a node that is affected by at least one input node. In FIG. 3, the concentrations of the input node and the output node are different, that is, the output node is displayed in a dark concentration. When the probability model is a meal related probability model, the probability analysis unit 123 extracts semantic information such as 'during meal' or 'during meal' through the probability model based on the collected log information.

In addition, in this embodiment, the Bayesian network probability model is preferably a modular structure. 4 is a conceptual diagram of a modular Bayesian network according to the present embodiment. As shown in FIG. 4, the Bayesian network is modularized according to the divided domains PN ₁ , PN ₂ ,..., PN _n . Here, the Bayesian network may be modularized according to a predetermined criterion, for example, a meal related probability module and a traffic situation probability module. In this case, the probability analysis unit 123 extracts semantic information such as 'during meal' or 'during out' through a meal related probability module (eg, PN ₁ ), and a traffic state probability module (eg, PN). ₂ ) extracts semantic information related to the current means of transportation or traffic jams. In this process, the probability calculation method used in the representative Bayesian network probability model can be used.

Furthermore, when the probability model is divided into several probability modules, it is efficient to perform probability calculation in consideration of the association between the modules. To this end, as shown in FIG. 4, a virtual concatenation technique may be used to store semantic information extracted from each probability module and its probability value and input and reflect it as probability evidence. In the virtual concatenation technique, a virtual node is added to reflect probabilistic evidence, and a probability of the evidence is applied through a conditional probability value (CPV) of the added node. 5A and 5B are conceptual views illustrating a method of using probability evidence in a virtual connection technique. 5A illustrates a method of directly modifying a CPT value without a virtual node when a node into which probability evidence is input is a root node. FIG. 5B illustrates a case in which a node containing probability evidence is not a root node, and a method of converting a CPT value of the node by adding a virtual node as a child node.

This virtual connection scheme is described in [K.-S. Hwang, S.-B. Cho and Jong-Ho Lea, "A Bayesian inference model for landmarks detection on mobile devices," Journal of Korea Information Science Society: Computing Practices, vol. 13, no. 1, pp. 35-45, 2007. 02.] detailed description thereof will be omitted.

Referring back to FIG. 1, the probability model updater 150 stores the probability model stored in the storage 140 according to the log information collected by the information collector 110 and the semantic information extracted by the semantic information extractor 120. Update the. Here, the update of the probabilistic model may be performed periodically, or may be performed when data that is the basis of the update is collected at a predetermined level or more, or may be performed when the user designates it.

6 is a flowchart illustrating a process of updating the probability model by the probability model update unit 150 according to an embodiment of the present invention.

In operation 610, the probability model update unit 150 receives a set of input semantic information and target semantic information and defines a set of nodes to be learned. Here, the input semantic information is semantic information corresponding to the input node of the Bayesian network probabilistic model, and the log information collected by the information collecting unit 110 or the meaning extracted by the rule analyzing unit 121 or the statistical analyzing unit 122. Information is available. The target semantic information is semantic information corresponding to an output node and is semantic information to be analyzed based on the input semantic information. For example, the current user is 'eating out' or 'meaning school' meaning information. The target semantic information may include semantic information extracted by the probability analysis unit 123, an annotation input by a predefined rule or user, a query response result or feedback value obtained from a user, or a log that has already been collected or extracted. Semantic information generated by analyzing the information or semantic information may be used. Step 610 is a step of defining a set of nodes that will appear in the updated probability model, and includes both nodes corresponding to input semantic information and target semantic information. That is, the input semantic information is L = {l ₁ , l ₂ ,... }, The target semantic information is LM = {lm ₁ , lm ₂ ,... }, The set X of the entire nodes to be learned in step 610 is X = L∪LM = {X1, X2,... } Is defined.

The probabilistic model updater 150 defines the level of each node of the Bayesian network in step 620. In constructing a hierarchical Bayesian network probabilistic model, it is a necessary process to naturally determine the inference direction when generating the probabilistic model. For example, the level of the node corresponding to the target semantic information may be defined as 0, and the level of the node corresponding to the input semantic information may be defined as 1. In this case, the level of the node can be expressed as follows.

V = {v ₁ , v ₂ ,... ∀v _i = 1, if X _i ∈L, ∀v _i = 0, if X _i LM, i = 1,. }

In operation 630, the probability model updater 150 calculates an association between nodes. In operation 640, the probability model updater 150 determines the phase ranking of the nodes. Step 630 is a process required to determine the phase ranking of the node in step 640, step 640 is a step performed because the K2 algorithm used in step 660 described later uses the node's phase ranking as a parameter, The topological rank of Hs has an important effect on the performance of the probabilistic model.

In operation 640, the probability model updater 150 determines that nodes having a higher level and higher correlations have a higher priority based on the correlation between nodes calculated in operation 630. In other words, the higher the level, the higher the priority of the node, and the higher the correlation, the higher the level of the node. Mutual information may be used as a method for calculating the association between nodes. In this case, when X and Y are nodes, the association M (X; Y) between node X and node Y may be calculated by the following Equation 1. . Here, x and y denote values of nodes X and Y, respectively.

And, the correlation of node Xi with other nodes Xj can be calculated by the following equation (2).

In operation 640, the nodes may be sorted according to the phase ranking based on the level of each node and the correlation of each node calculated according to Equation 2.

Mutual Information is described in [J. Su, H. Zhang, "Full Bayesian network classifiers," Proc. of international conference on Machine learning, pp. 897-904, 2006. The detailed description thereof will be omitted.

In step 650, the probability model update unit 150 divides the set of all nodes to be learned to define each module. As already explained, according to a certain criterion, it divides into a meal related probability module, a traffic situation probability module, etc., for example. Each module defined is for example M ₁ = {X ₂ , X ₅ ,... },… , M _d = {X ₃₄ , X ₄₉ ,.. } In the form of.

In operation 660, the probability model updater 150 learns a modular probability model structure. 7A to 7C are conceptual views illustrating the learning of a modular probability model structure, in which 'E' represents an evidence node (input node), 'L' represents a specificity node (node representing target semantic information), and ' V 'represents a virtual evidence node.

First, a Bayesian network structure is learned using data given as a result of performing the above-described steps 610 to 650. Here, the K2 algorithm may be used as the learning algorithm. In addition to the K2 algorithm, other structural learning algorithms may be used as the learning algorithm. For the K2 algorithm, see [G. F. Cooper and E. Herskovits, "A Bayesian method for the induction of probabilisitic networks from data," Machine Learning, vol. 9, pp. 309-347, 1992, detailed description thereof will be omitted. 7A shows the results of training a modular Bayesian network.

Next, for each module, {(a) specificity node to divide, (b) specificity node parented to (a), (c) evidence node directly connected to (a), (d) said (c ), Leaving only the evidence nodes directly connected to the This leaves only the desired specificity and evidence node for each module, other specificity nodes that directly affect it, and an arc between the remaining nodes. This result is shown in Figure 7b.

The node corresponding to the above (b) is defined as a virtual evidence node. This makes it possible to accept evidence for specificity inference in other modules. This result is shown in FIG. 7C.

Referring back to FIG. 6, in step 670, the probabilistic model updater 150 learns probability parameters of the modular Bayesian network. At this time, the parameter can be learned through histogram analysis on the probability distribution, which is a commonly used method.

By performing the above-described steps 610 to 670 in the probability model updating unit 150, the probability model updated according to the log information collected by the information collecting unit 110 and the semantic information extracted by the semantic information extractor 120 is updated. Is generated.

Referring back to FIG. 1, the input unit 130 is a portion for receiving a user's command and may include a plurality of keys. In particular, the input unit 130 receives a keyword, which is a search base from the user.

The searcher 160 searches for semantic information related to the keyword based on the probability model stored in the storage 140 according to the keyword received from the inputter 130. Meanwhile, the storage 140 stores a rule already defined by an administrator or a user, for example, an ontology model indicating an association between semantic information, and the retrieval unit 160 includes a rule already defined as well as a probability model. The semantic information related to the input keyword can be searched based on.

8 is a block diagram illustrating a specific embodiment of the search unit 160. Referring to FIG. 8, the search unit 160 according to the present embodiment includes a keyword expansion unit 161 and a semantic information search unit 162.

The keyword expansion unit 161 further generates keywords based on the search using the probability model and the ontology model stored in the storage 140 according to the keyword input from the input unit 130. The keyword expansion unit 161 includes a probability model based keyword expansion unit 163 and an ontology-based keyword expansion unit 165.

The probabilistic model based keyword expansion unit 163 finds a node corresponding to the input keyword and the cause or effect relationship in the probabilistic model and calculates the degree of robustness of the connection relationship. Specifically, the node directly connected to the node including the input keyword is found and the degree of robustness of the connection relationship is calculated using the conditional probability parameter given in the probability model.

The connection strength S _i representing the degree of robustness of the connection relationship with the i th node among the nodes connected to the current node may be calculated by the following equation.

Here, p _i is a conditional probability when only cause x _i of several causes of the current node is activated and has a value according to the following equation.

Here, y means the expressed value of the current node.

The probability model based keyword expansion unit 163 generates semantic information of the node corresponding to the case where the connection strength _Si is a predetermined threshold value, for example, 0.5 or more, as an extended keyword.

The ontology-based keyword expansion unit 165 further generates a keyword that is the basis of the search by using the ontology according to the input keyword. An ontology is a structure that can express a lot of information and rules based on facts, and basically consists of words and relationships. As an ontology model, see, for example, ConceptNet-A Practical Commonsense Reasoning Tool-Kit, by H. Liu and P. Singh (BT technology journal, v.22 no.4, 2004, pp. 211-226). The ConceptNet is mentioned. ConceptNet represents numerically related keywords and their degree of association with a given keyword. 9 shows an example of a semantic network designed for the operation of ConcepNet. Of course, the semantic network for the ontology model can be designed in several forms besides those presented in ConceptNet. 10 is a result of the numerical value representing the degree of association with the highly relevant keywords when the keyword 'living room' is entered in ConceptNet.

The ontology-based keyword expansion unit 165 may use the following functions in expanding keywords by the ontology model.

1) Affective: A set of concepts with high emotional similarity to the input concept A = {a ₁ , a ₂ ,. } Is obtained.

2) Consequences: A set of concepts that occur concurrently or sequentially with input concepts. C = {c ₁ , c ₂ ,... } Is obtained.

3) Details: A set of concepts that have characteristics similar to those of an input concept, or a set relationship, sub-event, etc. D = {d ₁ , d ₂ ,. } Is obtained.

4) Spatial: A set of place concepts related to the input concept S = {s ₁ , s ₂ ,... } Is obtained.

The ontology-based keyword expansion unit 165 uses these functions to set a set of keywords Q = {q ₁ , q ₂ ,... } To obtain a set of extended keywords CQ = {QA, QC, QD, QS} for each of the semantic information (for example, the semantic information included in the above-described input semantic information set L and the target semantic information set LM). The correlation is calculated and used as a criterion for keyword selection.

The following algorithm shows an example of an algorithm for calculating the correlation R.

for (int a = 0; a <m a ++) {

Ra = 0;

Ra + = w _a ㅧ Calculate_relationship (QA, L, LM);

Ra + = w _c ㅧ Calculate_relationship (QC, L, LM);

Ra + = w _d Calc Calculate_relationship (QD, L, LM);

Ra + = w _s ㅧ Calculate_relationship (QS, L, LM);

}

float Calculate_relationship (CQ, L, LM) {

U = L∪ LM

float Value = 0.0f;

for each (cq _i ∈CQ) {

for each (u _j ∈U) {

if (cq _i == u _j ) Value + = cq _i

}

}

retun Value;

}

In the above algorithm, w _a , w _c , w _d , and w _s mean the weights for each of the above four functions, and can be defined as 0.1, 0.1, 0.1, or 1.0, respectively. In this way, if the weight w _s is defined as a relatively large value, the keyword is expanded based on the concept of place. Further, the ontology-based keyword expansion unit 165 generates and outputs semantic information obtained as a correlation keyword having a degree of correlation of a predetermined threshold or more as an extended keyword.

Referring back to FIG. 8, the semantic information search unit 167 searches for semantic information according to a keyword input through the input unit 130 and a keyword generated by the keyword expansion unit 161, and returns the result to the display unit 180. Output In this case, the semantic information search unit 167 may not only search the semantic information that is simply searched, but also the connection strength indicating the degree of robustness of the connection relationship for the extended search results based on the probability model, and the extended search results based on the ontology model. Correlation values can also be output. In addition, the connection strength or correlation may be output by normalizing to a value between 0 and 1. FIG.

Referring back to FIG. 1, the display unit 170 serves to rationally output a user's command processing result. The display unit 170 may be implemented as, for example, a flat panel display device such as a liquid crystal display (LCD). In particular, the display unit 170 effectively displays the search results of the search unit 160, that is, the searched semantic information on the screen. In this case, if the search result is a search result through an extended keyword, the value or degree of connection strength or correlation may be displayed on the screen. Preferably, the display unit 170 provides a map image. If the searched semantic information is related to the location information, the display unit 170 may display location information corresponding to the searched semantic information on the map image. The connection strength or correlation of the searched result may be expressed by, for example, the size, density, color, etc. of a dot, a figure, or a picture displayed on the screen.

11 and 12 illustrate examples of screens on which a search result is displayed by the display unit 170 according to an exemplary embodiment. FIG. 11 is a screen displayed using a general map, and FIG. 12 is a screen displayed using a satellite map. 11 and 12, search results according to keywords input from a user are displayed as movement trajectories and semantic information on a map image, and the collected multimedia information, for example, pictures and videos, is displayed in a separate list. Can be displayed. When the multimedia displayed through the separate list is selected by an action such as a click, the collection position may be displayed on the map image. In addition, the semantic information is simply displayed as text on the map image, and more detailed information may be displayed when the mouse is moved or clicked on the corresponding position. In addition, the display unit 170 may have a function of filtering and displaying semantic information according to an option such as a date desired by a user.

Referring back to FIG. 1, the controller 180 connects and controls the elements in the mobile device 100. For example, when a keyword and a search command are input from the user through the input unit 130, the search unit 170 and the display unit 180 may perform a search by referring to the data stored in the storage 140 and display the searched result. And so on.

13 is a flowchart of an information retrieval method in a mobile device capable of managing an individual's daily routine according to an embodiment of the present invention. The information retrieval method according to this embodiment consists of the steps processed in the mobile device described with respect to FIG. Therefore, although omitted below, the contents described with reference to FIG. 1 are also applied to the information retrieval method according to the present embodiment.

First, in operation 1310, the mobile device 100 collects log information indicating a user's life pattern.

In operation 1320, the mobile device 100 analyzes the collected log information to extract semantic information that is meaningful information.

In operation 1330, the mobile device 100 updates the probability model stored in the storage 140 according to the collected log information and the extracted semantic information. Prior to the update of the probabilistic model in this step, the storage unit 140 stores a probabilistic model including semantic information related to at least some of the collected log information. This step may be performed periodically, or when log information is collected at a predetermined level or more, or may be performed when a user command is issued.

In operation 1340, the mobile device 100 receives a keyword and an information search command from the user.

In operation 1350, the mobile device 100 searches for semantic information based on the ontology model indicating an association between a probability model and a predetermined rule, for example, semantic information, according to the input keyword.

In operation 1360, the mobile device 100 displays the searched result in operation 1350 so that the user can recognize it.

FIG. 14 is a flowchart illustrating a more specific embodiment of step 1320 shown in FIG. 13. In operation 1320, the semantic information extractor 120 described with reference to FIG. 2 is configured. Therefore, even if omitted below, the content described with reference to FIG. 2 is also applied to step 1320 according to the present embodiment.

In operation 1410, the mobile device 100 analyzes the collected log information to extract semantic information that satisfies a specific rule.

In operation 1420, the mobile device 100 analyzes the collected log information and extracts semantic information representing a statistical meaning.

In operation 1430, the mobile device 100 extracts semantic information through a probabilistic model based on the collected log information or the semantic information extracted in operation 1410 or 1420.

FIG. 15 is a flowchart illustrating a more specific embodiment of step 1350 illustrated in FIG. 13. Operation 1350 according to the present exemplary embodiment includes steps processed by the searcher 160 described with reference to FIG. 8. Therefore, even if omitted below, the contents described with reference to FIG. 8 are applied to step 1350 according to the present embodiment.

In operation 1510, the mobile device 100 may further generate a keyword that is the basis of the search by using a probabilistic model and an ontology model indicating an association between a predetermined rule, for example, semantic information, according to a keyword input from the user.

In operation 1520, the mobile device 100 searches for semantic information according to the received keyword and the keyword further generated in operation 1510.

In the above-described embodiment of the present invention, the probabilistic model including semantic information related to the log information collected through the mobile device 100 may probabilistically include the semantic correlation of various information in the daily life of the user. do. Therefore, according to the embodiment described above, since the semantic information related to the keyword input from the user is searched based on the probability model including semantic information related to at least some of the collected log information, the search considering the user's actual life pattern Can provide results. Furthermore, since the semantic information related to the input keyword is searched in consideration of not only such a probabilistic model but also a pre-defined rule such as an ontology model, that is, a generally accepted rule, a more substantial search result can be provided. In addition, since the probability model is updated according to the collected log information and the semantic information extracted by analyzing the information, the more the mobile device 100 is used by the user, the more faithfully the user's actual life pattern can be considered. The quality is improved.

Meanwhile, the above-described embodiments of the present invention can be written as a program that can be executed in a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium. The computer-readable recording medium may be a magnetic storage medium (for example, a ROM, a floppy disk, a hard disk, etc.), an optical reading medium (for example, a CD-ROM, DVD, etc.) and a carrier wave (for example, the Internet). Storage medium).

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

1 is a block diagram of a mobile device capable of managing a daily life of an individual according to an embodiment of the present invention.

2 is a block diagram illustrating a specific embodiment of the semantic information extractor 120.

3 shows an example of a meal related probability model as an example of the Bayesian network probability model according to the present embodiment.

4 is a conceptual diagram of a modular Bayesian network according to the present embodiment.

5A and 5B are conceptual views illustrating a method of using probability evidence in a virtual connection technique.

6 is a flowchart illustrating a process of updating the probability model by the probability model update unit 150 according to an embodiment of the present invention.

7A to 7C are conceptual diagrams for explaining learning of a modular probability model structure.

8 is a block diagram illustrating a specific embodiment of the search unit 160.

9 shows an example of a semantic network designed for the operation of ConcepNet.

FIG. 10 shows the result of numerical values indicating the degree of association with highly related keywords when keywords are entered into ConceptNet.

11 and 12 illustrate examples of screens on which a search result is displayed by the display unit 170 according to an exemplary embodiment.

13 is a flowchart of an information retrieval method in a mobile device capable of managing an individual's daily routine according to an embodiment of the present invention.

FIG. 14 is a flowchart illustrating a more specific embodiment of step 1320 shown in FIG. 13.

FIG. 15 is a flowchart illustrating a more specific embodiment of step 1350 illustrated in FIG. 13.

Claims (25)

In the mobile device which can manage an individual's daily life, An information collector configured to collect log information representing a user's life pattern; A storage unit for storing a probabilistic model including semantic information which is meaningful form of information related to at least some of the collected log information; An input unit for receiving a keyword from a user; A searcher for searching for semantic information based on the probability model according to the received keyword; And And a display unit for displaying the searched result. The method of claim 1, And a semantic information extractor configured to analyze the collected log information to extract semantic information which is meaningful information. The method of claim 2, And a probabilistic model updater for updating the probabilistic model according to the collected log information and the extracted semantic information. The method of claim 2, wherein the semantic information extraction unit, A rule analyzing unit for extracting semantic information satisfying a specific rule by analyzing the collected log information; A statistical analysis unit for extracting semantic information representing a statistical meaning by analyzing the collected log information; And And a probability analyzer extracting semantic information through the probability model based on the collected log information or the semantic information extracted from the rule analyzer or the statistical analyzer. The method of claim 3, And the probability model is in the form of a modular Bayesian network. The method of claim 5, Each module constituting the probabilistic model has a virtual node for reflecting probabilistic evidence. The method of claim 5, The modular Bayesian network is a hierarchical structure having level information for each node. The method of claim 7, wherein The probability model updating unit receives the collected log information and the extracted semantic information, defines a set of nodes to be learned and levels of each node, calculates correlations between nodes, and determines a node's phase ranking according to the result. And learning the Bayesian network structure. The method of claim 8, The learning of the Bayesian network structure uses a K2 learning algorithm. The method of claim 1, The search unit may further include a keyword expansion unit generating a keyword that is the basis of the search by using the probability model and a predetermined rule according to the received keyword; And And a semantic information search unit for searching semantic information according to the received keyword and the further generated keyword. The method of claim 10, And the rule already defined follows an ontology model indicative of the association between semantic information. The method of claim 1, And the display unit displays location information corresponding to the retrieved semantic information in a map image when the retrieved semantic information is related to the location information. In the information retrieval method in a mobile device that can manage the individual's daily life, (a) collecting log information representing a life pattern of the user; (b) receiving a keyword from a user; searching for semantic information based on a probabilistic model including semantic information that is meaningful information related to at least some of the collected log information according to the received keyword; And (d) displaying the searched result. The method of claim 13, (a2) analyzing the collected log information and extracting semantic information which is meaningful information. The method of claim 14, (a3) updating the probability model according to the collected log information and the extracted semantic information. The method of claim 14, wherein step (a2), (a21) extracting semantic information satisfying a specific rule by analyzing the collected log information; (a22) extracting semantic information representing a statistical meaning by analyzing the collected log information; And (a23) extracting semantic information through the probability model based on the collected log information or the semantic information extracted in the step (a21) or the step (a22). The method of claim 15, The probability model is an information retrieval method characterized in that the modular Bayesian network form. The method of claim 17, Wherein each module constituting the probability model has a virtual node for reflecting probabilistic evidence. The method of claim 17, And wherein said modular Bayesian network is a hierarchical structure having level information for each node. The method of claim 19, In step (a3), the collected log information and the extracted semantic information are input to define a set of nodes to be learned and the level of each node, calculate an association between the nodes, and calculate the phase ranking of the nodes according to the result. Determine, and learn the Bayesian network structure. The method of claim 20, The learning of the Bayesian network structure uses a K2 learning algorithm. The method of claim 13, wherein step (c) comprises: Generating a keyword that is the basis of the search by using the probability model and a predetermined rule according to the received keyword; And And searching for semantic information according to the received keyword and the generated keyword. The method of claim 22, And the rule already defined follows an ontology model indicating an association between semantic information. The method of claim 13, In the step (d), when the searched semantic information is related to the location information, location information corresponding to the searched semantic information is displayed on a map image. A computer-readable recording medium having recorded thereon a program for executing the information retrieval method according to any one of claims 13 to 24.

Images