DESC:RELATED APPLICATION
Benefit is claimed to India Provisional Application No. 3397/CHE/2013 titled "PATIENT KNOWLEDGE GRAPH" by BHAUMIK, Sandip et Al., filed on 30th July 2013, which is herein incorporated in its entirety by reference for all purposes.

FIELD OF INVENTION
The embodiments herein relate to pervasive medical computing and more particularly relates to a method and system for generating medical recommendations based on a knowledge graph.
BACKGROUND
During recent years, the field of context and intelligence is considered as a topic of great momentum. Typically, Context-Aware Intelligence is applied to understand a situation of a user and the user’s behavior for providing adaptive services related to their situation. Particularly, in medical domain, Context-Aware health architectures are being designed to provide improved healthcare solutions through intelligent use of patient health data, and continuous monitoring of patient vital signs.
In existing Context-Aware systems, behavior modeling of patients is performed using computing and communication systems that are worn on the patient’s body. Another existing system builds knowledge profiles of the user by measuring the user’s expertise based on answers received in response to different questions. In some existing systems, medical contents are recommended to a patient relevant to medical profiles of the patient.
Another existing system recommends medical information to the user based on intent provided by the user. For example, when the user provides the intent ‘diabetes in pregnancy’, the information regarding the diabetes is recommended to the user. The information recommended by these systems lack trust, as the data or information is provided to the user without considering source details. Further, these systems do not consider the user’s knowledge or expertise on those particular topics while recommending the information.
In another case, the intent provided by the user may not be in standard form. Hence, the recommended information or the data may not provide exact or required information that is desired by the user. For example, some of latent topics have high correspondence with medical topics. The user has to manually investigate all such possible correspondence or associations which require large amount of time and effort. Thus, there remains a need of robust and simple system and method for recommending trusted medical information or results to a user considering user’s expertise level, knowledge, and so on.

OBJECT OF INVENTION
The principal object of the embodiments herein is to provide a method and system for recommending trusted medical information to a user using a knowledge base search across a knowledge graph, considering the user’s propositional knowledge.
Another object of the embodiments herein is to assign weight to a response received for intent, where the weight is assigned based on a semantic similarity between the response and each knowledge object of a knowledge graph.
Another object of the embodiments herein is to construct a knowledge graph by extracting label data associated with one or more knowledge objects.
Another object of the embodiments herein is to construct a knowledge graph by determining latent data in a knowledge object and identifying the frequency of the latent data associated with a plurality knowledge objects.

SUMMARY
Accordingly the invention provides a method for generating at least one recommendation for a user. The method comprising constructing a knowledge graph representing at least one knowledge object associated with a node among a plurality of nodes, wherein each node represents a duster of information. The method further includes receiving at least one intent associated with the user. Furthermore, the method includes assigning at least one weight to at least one knowledge object of the knowledge graph, wherein the weight is assigned based on at least one semantic similarity between the intent and each knowledge object of the knowledge graph. The method further includes generating at least one recommendation for the user based on the weightage, wherein at least one recommendation comprises at least one knowledge object.
Accordingly the invention provides a system for generating at least one recommendation for a user. The system comprising a server, and an electronic device. The server is configured to construct a knowledge graph representing at least one knowledge object associated with a node among a plurality of nodes, wherein each node represents a duster of information. The system is configured to receive at least one intent associated with the user and assign at least one weight to at least one knowledge object of the knowledge graph, wherein the weight is assigned based on at least one semantic similarity between the intent and each knowledge object of the knowledge graph. Further, the system is configured to generate at least one recommendation for the user based on the weightage, wherein the at least one recommendation comprises the at least one knowledge object.
Accordingly the invention provides a computer program product comprising computer executable program code recorded on a computer readable non-transitory storage medium, the computer executable program code when executed causing the actions including constructing a knowledge graph representing at least one knowledge object associated with a node among a plurality of nodes, wherein each node represents a duster of information. The computer executable program code when executed causing further actions including receiving at least one intent associated with the user and assign at least one weight to said at least one knowledge object of the knowledge graph, wherein the weight is assigned based on at least one semantic similarity between the intent and each knowledge object of knowledge graph. The computer executable program code when executed causing further actions including generating at least one recommendation for the user based on the weightage, wherein the at least one recommendation comprises the at least one knowledge object.
These and other aspects of the embodiments herein will be better appreciated and understood when considered in conjunction with the following description and the accompanying drawings. It should be understood, however, that the following descriptions, while indicating preferred embodiments and numerous specific details thereof, are given by way of illustration and not of limitation. Many changes and modifications may be made within the scope of the embodiments herein without departing from the spirit thereof, and the embodiments herein include all such modifications.

BRIEF DESCRIPTION OF FIGURES
This invention is illustrated in the accompanying drawings, throughout which like reference letters indicate corresponding parts in the various figures. The embodiments herein will be better understood from the following description with reference to the drawings, in which:
FIG. 1 illustrates a high level overview of a system for generating medical recommendations for a user using a knowledge base search across a knowledge graph, according to embodiments disclosed herein;
FIG. 2 is block diagram illustrating various modules of a server, according to embodiments disclosed herein;
FIG. 3 is block diagram illustrating various modules of an electronic device, according to embodiments disclosed herein;
FIG. 4 shows an illustration of constructing knowledge graph representing one or more knowledge objects, according to embodiments disclosed herein;
FIG. 5 is a schematic representation illustrating a data model of a knowledge graph, according to embodiments disclosed herein;
FIG. 6 is a diagrammatic representation illustrating construction of a knowledge graph, according to embodiments as disclosed herein;
FIG. 7 is a schematic representation illustrating relations between a user (u), a document (d), a medical topic (m) and an LDA topic (l) in an exemplary patient knowledge graph, according to embodiments as disclosed herein;
FIG. 8 shows a schematic representation illustrating a knowledge graph structure involving users and documents, according to embodiments as disclosed herein;
FIG. 9 illustrates relationship between a keyword w and an LDA topic l, according to embodiments as disclosed herein.
FIG. 10 is a schematic representation of association between a medical category (m) and an LDA topic (l), according to embodiments as disclosed herein;
FIG. 11 is a flow diagram illustrating a method for generating and displaying a recommendation for a user by considering user’s propositional knowledge and a knowledge graph, according to embodiments as disclosed herein;
FIG. 12 is a flow diagram illustrating a method for assigning and displaying a weight to a response associated with a node, according to embodiments as disclosed herein; and
FIG. 13 illustrates a computing environment implementing the method and system for recommending trusted medical information to a user using a knowledge base search across a knowledge graph, according to embodiments as disclosed herein.

DETAILED DESCRIPTION OF INVENTION
The embodiments herein and the various features and advantageous details thereof are explained more fully with reference to the non-limiting embodiments that are illustrated in the accompanying drawings and detailed in the following description. Descriptions of well-known components and processing techniques are omitted so as to not unnecessarily obscure the embodiments herein. Also, the various embodiments described herein are not necessarily mutually exclusive, as some embodiments can be combined with one or more other embodiments to form new embodiments. The term “or” as used herein, refers to a non-exclusive or, unless otherwise indicated. The examples used herein are intended merely to facilitate an understanding of ways in which the embodiments herein can be practiced and to further enable those of skill in the art to practice the embodiments herein. Accordingly, the examples should not be construed as limiting the scope of the embodiments herein.
Prior to describing the embodiments in detail, it is useful to provide definitions for key terms and concepts used herein. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by a personal having ordinary skill in the art to which this invention belongs.
Knowledge graph or Med-Tree: Refers to a knowledge base that may be represented by using a visually appealing graphical presentation. Knowledge graph organizes information in the form of nodes, topics, sub-topics, keywords in the knowledge objects associated with a plurality of users. The users, the documents, medical concepts and extracted topics are stored in the Med-Tree through tree nodes and their relations.
Knowledge object: Refers to information related to a topic of interest or a domain knowledge comprising a document set, a medical topic, keyword, Latent Dirichlet Allocation (LDA) topic, concept text, and a concept type and the like.
Duster of information: A duster of information is a set of related knowledge objects in which each knowledge object has high information gain with respect to at least one other knowledge object. Each node in the knowledge graph represents a duster of information.
Propositional knowledge: The knowledge gained by user by reading articles, documents and the like.
Medical-based assistive network: Refers to a network that assists a user in retrieving medical information quickly and easily and enables the user to take decision effectively. The assistive network enables the user to provide intent and allows a knowledge base search across the knowledge graph based on the user’s propositional knowledge and the intent.
User: Refers to a person who provides intent by performing an activity on an electronic device for retrieving information from one or more knowledge objects in the medical based assistive network.
Intent: Refers to a topic of interest that a user is looking for by performing an activity on the electronic device. The intent can be specified either implicitly or explicitly by the user in the electronic device by performing one or more activities on an application.
Word vector: Refers to the magnitude and direction for determining the context of current topic based on keywords or label data identified in the knowledge graph.
Token: Refers to a unique identifier that identifies the keyword in the information source.
Semantic similarity: Refers to analyzing the keywords, topics in the knowledge object for determining semantically meaningful terminology associated with the extracted items in the knowledge object.
The embodiments herein achieve a method and system for recommending trusted medical information to a user using a knowledge base search across a knowledge graph. In an embodiment, the recommendations can be generated considering parameters such as, but not limited to, the user’s propositional knowledge (node or user’s knowledge graph), topic, subject, expertise level, and interest of the user, and the like. The method includes constructing a knowledge graph which represents plurality of knowledge objects associated with the users. Further, the method includes receiving intent associated with the user. Furthermore, the method includes assigning a weight to the knowledge objects of the knowledge graph based on a semantic similarity between the intent and each knowledge object. Furthermore, the method includes generating one or more recommendations for the user based on the weight assigned to the knowledge objects.
Unlike the conventional systems, the proposed system and method includes providing the recommendations (medical information) to the user considering user’s intent along with user’s propositional knowledge, and user’s expertise level in the given topic and so on. Hence, the recommended results are more focused which further avoids unnecessary information regarding the topic of interest.
In an embodiment, the method includes determining the intent provided by the user. The method further includes receiving a response for the intent from one or more users associated in the communication network. Furthermore, the method includes assigning a weight to each of the response received, where the weight is assigned based on semantic similarity between the response and each knowledge object of the knowledge graph. Furthermore, the method includes displaying the assigned weights to the user in user’s electronic device. Unlike the conventional systems, the method includes assigning weights to the responses that are received for the intent. Hence, the user can easily identify the trusted medical information among the plurality of received responses.
Referring now to the drawings, and more particularly to FIGS. 1 through 13, where similar reference characters denote corresponding features consistently throughout the figures, there are shown preferred embodiments.
FIG. 1 illustrates a high level overview of a system 100 for generating medical recommendations for a user using a knowledge base search across a knowledge graph, according to embodiments disclosed herein. As depicted in the FIG. 1, the system 100 includes a server 101, a plurality of electronic devices 1021-N (herein after referred as electronic device 102), and a medical based assistive network 103.
In an embodiment, the server 101 can be configured to construct a knowledge graph representing a plurality of knowledge objects associated with different nodes in the medical based assistive network 103. Further, the server 101 can be configured to assign weights to the knowledge objects in the knowledge graph, in which weights can be assigned to generate recommendations to the user. In an embodiment, the server 101 can be configured to assign weights to one or more responses received for the user’s intent. Furthermore, the server 101 can be configured to store the standard medical ontology for knowledge extraction and identification of medical information while generating the recommendations using the knowledge graph.
The electronic device 102 can be configured to display the recommendations generated by the system 100 to the user. Further, the electronic device 102 can be configured to provide the knowledge object associated with the user to the server 101. Furthermore, the electronic device 102 can be configured to store the user’s information in form of a knowledge graph (herein after referred as a node knowledge graph) and allows the user to perform the user activity in order to capture the intent of the user.
In an embodiment, the electronic device 102 can be any kind of computing device, such as, but not limited to, a laptop computer, Personal Digital Assistant (PDA), mobile phone, smart phone, or any electronic computing device which has been configured to perform the functions disclosed herein.
In an embodiment, the medical based assistive network 103 can be accessed by using any suitable network, such as, but not limited to, wireless network, wire line network, public network such as the Internet, private network, general packet radio network (GPRS), local area network (LAN), wide area network (WAN), metropolitan area network (MAN), cellular network, public switched telephone network (PSTN), personal area network, and the like.
FIG. 2 is block diagram illustrating various modules of a server 101, according to embodiments disclosed herein. The server 101 includes a data analyzer module 201, a semantic analyzer module 202, a query interpreter/builder module 203, a weight assigning module 204, a knowledge graph module 205, a recommendation module 206, a communication module 207, and a storage module 208.
The data analyzer module 201 can be configured to extract label data, and keywords from the knowledge objects associated with each user and analyze the data present in each knowledge object. The semantic analyzer module 202 can be configured to analyze the label data, and keywords for semantic correctness and create word vectors and tokens from the extracted keywords. In an embodiment, the semantic analyzer module 202 uses Latent Dirichlet Allocation (LDA) algorithm to identify latent data associated with each of the knowledge objects. Further, a list of the word vectors depicting each topic present within the knowledge object is identified. Furthermore, an indexing module which uses keywords (sets of keywords) present within each word vector is used to determine frequency of the latent data appearance in a plurality of knowledge objects of the knowledge graph. This would form an index denoting a set of word vectors with corresponding location identifiers within each knowledge object.
The Query interpreter/builder module 203 can be configured to interpret extracted latent data and further build a localized query based on the extracted latent data. Some of the latent topics may have high correspondence with some medical topics. The query interpreter/builder module 203 automatically finds the possible correspondence and associations of the latent topics by using the stored medical ontology.
The weight assigning module 204 can be configured to assign weights to the knowledge objects of the knowledge graph on receiving intent from the user. In an embodiment, the weight assigning module 204 can be configured to assign weights to one or more responses received for the user’s intent.
Based on the extracted data and the intent of the user, the knowledge graph module 205 can be configured to depict information in the form of a knowledge graph representing one or more knowledge objects associated with the users. The recommendation module 206 can be configured to generate recommendations to the user based on weights assigned to each of the knowledge objects of the knowledge graph.
The communication module 207 can be configured to provide the generated recommendations to the user by using suitable communication channel with the user’s electronic device 102. The storage module 208 stores the constructed knowledge graph along with the standard medical ontology.
FIG. 3 is block diagram illustrating various modules of an electronic device 102, according to embodiments disclosed herein. The electronic device 102 includes an input/output module 301, a knowledge graph module 302, a communication module 303, and a storage module 304.
In an embodiment, the input/output module 301 can be configured to receive intent from the user. In an embodiment, the user can provide the intent either implicitly or explicitly. In an embodiment, an implicit intent can be provided by the user by performing an activity on an application running on an electronic device 102. In an embodiment, an explicit intent can be provided by the user by specifying a localized query on an application running on the electronic device 102. The input/output module 301 can be configured to display the recommendations provided by the server 101.
In an embodiment, the knowledge graph module 302 can be configured to construct user’s knowledge graph (node knowledge graph) based on user’s browsing history, online access history, online subscriptions, online blogging or micro blogging, social feeds and so on related to medical topics. The knowledge graph module 302 can also be configured to frequently monitor the activities associated with the user and update the activities to the user’s personalized knowledge graph. The storage module 304 stores the user’s knowledge graph in the electronic device 102. The communication module 303 can be configured to provide communication sessions with the server 101 by using the medical based assistive network 103.
FIG. 4 shows an illustration of constructing knowledge graph 400 representing one or more knowledge objects, according to embodiments disclosed herein. In an embodiment, a medical/healthcare oriented knowledge graph, (also known as Med-Tree) is generated leveraging the user’s knowledge object. In an embodiment, the user’s knowledge object can be a document set (such as health records), a medical topic, keyword, LDA topic, concept text, and concept type, and so on. In an embodiment, the knowledge graph can be constructed by extracting the label data associated with each knowledge object (supervised extraction). The label data can be extracted based on the medical ontology stored in the server 101. Further, the extracted label date is stored in the knowledge graph through nodes and their relations. For example, the nodes in the knowledge can be given as
Documents (such as, asthma-living-managing.txt and so on)
Medical Topics (such as Stroke and so on)
Keywords (such as care, symptom, liver and so on)
LDA Topics (such as Topic 0)
Concept Texts (such as wheezing, coughing and so on)
The nodes in the knowledge graph can be connected through different relation types. For example, the relation types can be as follows:
(Document) CONTAINS (Keyword)
(Document) IN_MED_CATEGORY (Medical Topic)
(Document) IS_IN (LDA Topic)
(Keyword) MAY_INDICATE (LDA Topic)
(Document) CONTAINS_C (Concept Text)
(Concept Text) IS_OF_TYPE (Concept Type)
(LDA Topic) LDA_TO_MED (Medical Topic)
(Concept Text) CT_TO_MED (Medical Topic)
The information is analyzed to gather insights into user’s belief or knowledge quotient on different medical subjects.
In an embodiment, the knowledge graph can be constructed by identifying the latent data associated with the knowledge objects (unsupervised extraction). Further, the frequency of the latent data appearance in plurality of knowledge objects is determined. Furthermore, a semantic similarity among the plurality of knowledge objects is determined based on the frequency of the latent data.
In an embodiment, the semantic similarity between a medical topic (m) and a latent topic (l) can be determined using the following equation:
a(l,m)=?_(d? µ(m))¦? ?_(w? W_F (l))¦?tf_l (w,l) log??N/(df_l (w,l) ) ?(d,l) (tf(w,d))/(tf_c (w))? ?? ----(1)
where,
W_(F(l) )denotes the frequent keywords in LDA topic l
µ(m)denotes the documents belonging to the medical topic m
?(l) denotes the documents belonging to the LDA topic l
tf(w,d) denotes the number of times w occurs in document d
tf_l (w,l) denotes the number of times w occurs in the LDA topic l. i.e. tf_l (w,l)=?_(d??(l))¦?tf(w,d)?
df_l (w) is the number of documents in ?(l) that contain the word w.
?(d,l) denotes the proportion of belongingness of the document d in l.
tf_c (w) denotes the number of times w occurs in the entire collection.
In an embodiment, correspondence between the concept texts and the medical topics also can be defined. Concept texts are the problems/symptoms, medical tests, or medications for different diseases. If the system has access to the patient’s health records containing the symptoms, the medical tests and so on then the concept texts present in the records can be extracted. The strength of association between the concept texts and the medical topics is computed using the following equation:
ß(c,m)=(|D_m(c) |)/(|D(c)|)-------------------(2)
Here,
D(c) is the set of documents that contain the concept text c. |D(c)| denotes the number of documents in this set.
D_m (c) denotes the documents that belong to the medical topic m and contain the concept text c. |D_m (c)| denotes the number of documents in this set.
FIG. 5 is a schematic representation illustrating a data model of a knowledge graph, according to embodiments disclosed herein. As depicted in the FIG. 5, a combination of supervised and unsupervised extractions are used for identifying different nodes and creating relationships between nodes to form the knowledge graph. Further, the FIG. 5 depicts the document level view of the knowledge graph showing different knowledge objects and the relationships between them. The FIG. 5 also indicates the phases in which the objects/relations are identified.
FIG. 6 is a diagrammatic representation illustrating a knowledge graph, according to embodiments as disclosed herein. The FIG. 6 depicts the details and documents regarding two persons (Person A and Person B). The knowledge of the persons is extracted by many sources including, but not limited to, social networks and personal data sources. The extracted details and the documents are queried and retrieved by a query interface. The information is then stored in a form a graph comprising different types of nodes. For example, at the document level, the knowledge graph contains following types of nodes, viz. Documents (for example, files such as asthma-living-managing.txt and so on), Medical Topics (for example, Stroke), Keywords (for example, care, symptom, liver, and so on), Latent Dirichlet Allocation (LDA) Topics, Concept Texts (for example, wheezing, coughing and so on) and Concept Types (for example, treatment, test, problem).
FIG. 7 is a schematic representation illustrating relations between a user (u), a document (d), a medical topic (m), and an LDA topic (l) in an exemplary patient knowledge graph, according to embodiments as disclosed herein. The FIG. 7 depicts a graph structure. For each user, interest or knowledge quotient can be defined for each medical and LDA topic. For example, if a user has read many documents on lung cancer, then the user knows about lung cancer. In another example, if the user is authoring many documents on diabetes, it indicates that the user is knowledgeable about diabetes. As depicted in the FIG. 7, if the user ‘u’ has read the document ‘d’ and ‘d’ is in medical category ‘m’ or LDA category l, then the user ‘u’ has gathered some knowledge about the category m (or l). This information can be captured by the Relation type ‘KNOWS_ABOUT’ captures. The knowledge quotient (e) of the user u is stored as a property of the KNOWS_ABOUT relation.
In an embodiment, the knowledge quotient of a user on a particular topic is modeled using the following equation:
e(u,c)= ?_(u accesses d )¦belongingness(d,c) ---------(3)
where, c denotes a topic (medical topic or LDA topic).
FIG. 8 shows a schematic representation illustrating a knowledge graph structure involving users and documents, according to embodiments as disclosed herein. The FIG. 8 depicts a user has accessed the document. The document may belong to a medical topic or a LDA topic. Further, the user knows about the information or the data. Hence, the relation is shown as KNOWS_ABOUT in the knowledge graph. The FIG. 8 also illustrates the weigh factors associated with the document (knowledge object).
FIG. 9 illustrates relationship between a keyword w and an LDA topic l, according to embodiments as disclosed herein. For an instance, importance of a keyword (w) for an LDA topic (l) is high if:
w occurs frequently in the documents that are in l.
w does not occur frequently in other LDA topics.
In an embodiment, importance of w for l can be defined as: imp(w,l) = tfl(w,l) * log10 (N/dfl(w,l)), where tfl(w,l) is number of times w occurs in l, and dfl(w,l) is number of topics l where w is marked as frequent (during topic modeling). It can be noted that, importance is high when tfl(w,l) is high but dfl(w,l) is low.
FIG. 10 is a schematic representation of association between a medical category (m) and an LDA topic (l), according to embodiments as disclosed herein. In an embodiment, a LDA topic (l) has high correspondence with a medical category (m) if: a) Many documents (d) that are in m also belong to l; and b) ‘d’ contains important topical keywords w from l. Further, the keywords w are also important in d. In an embodiment, the correspondence index is given as
Correspondence index a (l, m):

where, tf (w, collection) is a number of times w occurs in an entire collection.
FIG. 11 is a flow diagram illustrating a method 1100 for generating and displaying a recommendation for a user by considering user’s propositional knowledge and a knowledge graph, according to embodiments as disclosed herein. At step 1101, the method 1100 includes constructing a knowledge graph representing the knowledge objects by measuring user’s knowledge quotient or expertise in different medical domains. In an embodiment, the knowledge objects can be, but not limited to, a document set (health records), a medical topic, keyword, Latent Dirichlet allocation (LDA) topic, concept text, and concept type, and the like. Further, the knowledge objects are associated with the nodes (users) in the medical based assistive network 103. In an embodiment, the term node represents a ‘duster of information’. Further, each node may have a person with an electronic device 102 which includes a friend electronic device, a friends-of-friend electronic device, a group electronic device, a department electronic device, a community electronic device, a company electronic device, an organization electronic device, a customer care electronic device, and an expert electronic device. The method 1100 allows the knowledge graph module 205 in the server 101 to construct the knowledge graph representing the knowledge objects of the users. In an embodiment, each user in the medical based assistive network 103 have own knowledge graph constructed and stored in their electronic devices. In an embodiment, the constructed knowledge graph comprises a semantic relationship among the plurality of nodes. Further, the semantic relationship is computed, for each node pair, by comparing a type of relation associated with a first node in the node pair and a type of relation associated with a second node in the node pair.
At step 1102, the method 1100 includes receiving intent associated with the user. The user can provide the intent through the electronic device 102 connected in the medical based assistive network 103. At step 1103, the method 1100 includes extracting an item from the intent received. In order to determine the medical topics of the intent, the query interpreter/builder module 203 determines whether any medical name is present in the received intent. For example, the user provides direct medical topic details like osteoarthritis. In some cases, the user may specify a generic name (for example, joint pain) or symptoms associated with the query (for example, pain instead of arthritis). For example, the user specifies the intent as ‘joint pain therapy’. Further, the query interpreter/builder module 203 searches the standard medical ontology stored in the storage module 208 for extracting the items in the intent received. As in the medical standards, the term ‘joint pain’ is not defined, the query interpreter/builder module 203 resolves the intent (query) associated with the medical topics as ‘osteoarthritis’ and ‘rheumatoidarthritis’. In an embodiment, the intent is mapped to the LDA topics that define about treatment, cure or medication.
At step 1104, the method 1100 includes computing a semantic similarity between the item and the each knowledge object of the knowledge graph. The method 1100 allows the semantic analyzer module 202 to compute the semantic similarity by analyzing the extracted item and the data associated with each of the knowledge object of the knowledge graph.
At step 1105, the method includes assigning weights to the knowledge object of the knowledge graph based on parameters associated with the user. In an embodiment, the parameters can be, but not limited to, a knowledge graph, topic, subject, knowledge, expertise level, and interest of the user. In an embodiment, the parameter can be stored locally in the user’s electronic device 102. The weight assigning module 204 assigns the weights based on the computed semantic similarity between the intent and each knowledge object of the knowledge graph.
In an embodiment, the weight can be assigned to each knowledge object as follows: For each document d in the candidate set S of a knowledge object, the document’s proportion of belongingness (?) to the LDA topics L(q) is considered. Then the weight of d is computed by using the following equation:
weight(d)= ?_(i=1)^k¦?(?_(w?q)¦tf(w,d)idf(w) ) ?(d,L_i ) ? ---(4)
idf(w) is the inverse document frequency of w. Documents with highest scores are returned are results. For example, the user provides the intent ‘avoid constipation diet’. After computing the semantic similarity between the item ‘constipation’ and the knowledge objects in the knowledge graph, the method 1100 infer that the intent provided by the user relates to the category of ‘ibs’ (irritable bowel syndrome). Also, the user is seeking information on suggested food habits to avoid or prevent the ‘ibs’. Hence more weightage is assigned to the category ‘ibs’.
At step 1106, the method 1100 includes generating a recommendation for the user based on the weight assigned to each of the knowledge object. The method 1100 allows the recommendation module 206 to generate the recommendation for the user containing the knowledge object that assigned the highest weight. At step 1107, the method 1100 includes displaying the recommendation to the user on user’s electronic device 102. The method allows the input/output module 301 in the electronic device 102 to display the recommendation generated by the server 101. At step 1108, the method 1100 includes updating the parameters of the user in the electronic device 102 of the user. For example, the user has read many documents on lung cancer, the user’s expertise level about the lung cancer is improved. Hence, the node knowledge graph of the user is updated accordingly in the electronic device 102. At step 1109, the method 1100 includes frequently monitoring activities associated with the user. The method 1100 allows the knowledge graph module 302 to frequently monitor the activities associated with the user such as, but not limited to, intent, browsing history, online access history, online subscriptions, online blogging or micro blogging, social feeds and so on related to medical topics. At step 1110, the method 1100 goes to step 1103 in response to determining that modifications identified in the activities of the user. At step 1110, the method 1100 goes to step 1109 in response to determining that no modifications are identified in user activities.
The various actions, acts, blocks, steps, and the like in method 1100 may be performed in the order presented, in a different order or simultaneously. Further, in some embodiments, some actions, acts, blocks, steps, and the like may be omitted, added, modified, skipped, and the like without departing from the scope of the invention.
FIG. 12 is a flow diagram illustrating a method 1200 for assigning and displaying a weight to a response associated with a node, according to embodiments as disclosed herein. At step 1201, the method 1200 includes receiving a request from the user. In an embodiment, the request is received from an application (for example, a social networking site) associated with the user electronic device. At step 1202, the method 1200 includes extracting an item from the request received. In an embodiment, the item can be a word vector or a token. In order to determine the medical topics, the query interpreter/builder module 203 determines whether any medical name is present in the received request. Further, the query interpreter/builder module 203 searches the standard medical ontology for extracting the items. At step 1203, the method 1200 includes determining intent based on the extracted items. For example, a request ‘How to treat diabetes in pregnancy’ is received from a user. The name of the disease ‘diabetes’ is extracted as the term is included in the request. However, the request contains a subset of diabetes document that discuss about pregnancy. Hence, the intent is taken as ‘diabetes in pregnancy’. In an embodiment, the intent can be a topic, subject, knowledge, or the interest of the user.
At step 1204, the method 1200 includes receiving a response from a node in response to the intent associated with the user. For example, the users connected with the medical based assertive network 103 can provide recommendations or opinions or suggestions to the intent posted, through the electronic devices 102 associated with each of the user. At step 1205, the method 1200 includes computing a semantic similarity between the response and each knowledge object of a knowledge graph. In an embodiment, a medical/healthcare oriented knowledge graph, (also known as Med-Tree) is generated leveraging the user’s knowledge object. In an embodiment, the user’s knowledge object can be a document set (such as health records), a medical topic, keyword, LDA topic, concept text, and concept type, and so on. In an embodiment, the knowledge graph can be constructed by extracting the label data associated with each knowledge object (supervised extraction). In an embodiment, the knowledge graph can be constructed by identifying the latent data associated with the knowledge objects (unsupervised extraction). The method 1200 allows the semantic analyzer module 202 to compute the semantic similarity between the response and the knowledge objects by analyzing the data or information associated with the response.
At step 1206, the method 1200 includes assigning weight to the node associated with the response. In an embodiment, the weight is assigned to the node based on the semantic similarity computed between the response of the node and the knowledge objects of the knowledge graph. For example, a node associating with the highest similarity response is given the highest weightage. At step 1207, the method 1200 includes displaying the assigned weights to the user. The user can identify the potential or more related responses by viewing the weights assigned to each response.
At step 1208, the method 1200 includes frequently monitoring the responses received to the intent associated with the user. The method 1200 allows the knowledge graph module 205 to frequently monitor whether any new responses are received for the intent associated with the user. At step 1209, the method 1200 goes to the step 1205 in response to determining that new responses are received for the intent. At step 1209, the method 1200 goes to the step 1208 in response to determining that there are no responses received for the intent.
The various actions, acts, blocks, steps, and the like in method 1200 may be performed in the order presented, in a different order or simultaneously. Further, in some embodiments, some actions, acts, blocks, steps, and the like may be omitted, added, modified, skipped, and the like without departing from the scope of the invention.
FIG. 13 illustrates a computing environment implementing the system and methods described herein, according to embodiments as disclosed herein. As depicted in the figure, the computing environment 1300 includes at least one processing unit 1501 that is equipped with a control unit 1302 and an Arithmetic Logic Unit (ALU) 1303, a memory 1304, a storage unit 1305, plurality of networking devices 1306 and a plurality Input output (I/O) devices 1307. The processing unit 1301 is responsible for processing the instructions of the algorithm. The processing unit 1301 receives commands from the control unit 1302 in order to perform its processing. Further, any logical and arithmetic operations involved in the execution of the instructions are computed with the help of the ALU 1303.
The algorithm comprising of instructions and codes required for the implementation are stored in either the memory unit 1304 or the storage 1305 or both. At the time of execution, the instructions may be fetched from the corresponding memory 1304 and/or storage 1305, and executed by the processing unit 1301.
In case of any hardware implementations various networking devices 1307 or external I/O devices 1306 may be connected to the computing environment to support the implementation through the networking unit and the I/O device unit. The embodiments disclosed herein can be implemented through at least one software program running on at least one hardware device and performing network management functions to control the network elements.
The embodiments disclosed herein can be implemented through at least one software program running on at least one hardware device and performing network management functions to control the elements. The elements shown in Figs. 1 and 13 include blocks which can be at least one of a hardware device, or a combination of hardware device and software module.
The embodiment disclosed herein specifies a method and system for recommending trusted medical information to a user using a knowledge base search across a knowledge graph, considering the user’s propositional knowledge. The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and/or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of preferred embodiments, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope of the embodiments as described herein.
,CLAIMS:
1. A method for generating at least one recommendation for a user, the method comprising:
constructing a knowledge graph representing at least one knowledge object associated with a node among a plurality of nodes, wherein each said node represents a duster of information;
receiving at least one intent associated with said user;
assigning at least one weight to said at least one knowledge object of said knowledge graph, wherein said weight is assigned based on at least one semantic similarity between said intent and each said knowledge object of said knowledge graph; and
generating at least one recommendation for said user based on said weightage, wherein said at least one recommendation comprises said at least one knowledge object.

2. The method of claim 1, wherein constructing a knowledge graph comprises:
extracting at least one label data associated with said at least one knowledge object, wherein said at least one label data is extracted based on an ontology; and
constructing said knowledge graph by storing said extracted label data.

3. The method of claim 1, wherein constructing a knowledge graph comprises:
identifying at least one latent data associated with said at least one knowledge object;
determining frequency of said at least one latent data appearance in a plurality of knowledge objects;
determining a semantic similarity among said plurality of knowledge objects based on said frequency of said at least one latent data; and
constructing said knowledge graph by storing said latent data based on said semantic similarity.

4. The method of claim 1, wherein said knowledge graph comprises semantic relationship among said plurality of nodes, wherein said semantic relationship is computed, for each node pair, by comparing a type of relation associated with a first node in said node pair and a type of relation associated with a second node in said node pair.
5. The method of claim 1, wherein assigning at least one weight to said at least one object of said knowledge graph comprises:
extracting at least one item from said at least one intent;
computing said at least one semantic similarity between said at least one item and each said knowledge object of said knowledge graph based on at least one parameter associated with said user; and
assigning said at least one weight to each said knowledge object of said knowledge graph based on said semantic similarity.

6. The method of claim 5, wherein said parameter comprises at least one of: said user knowledge graph, topic, subject, knowledge, expertise level, and interest of said user, wherein said parameter associated with said user is stored locally on an electronic device of said user.
7. The method of claim 1, wherein said method further comprises displaying said at least one recommendation on said electronic device of said user, wherein said at least one recommendation is displayed along with said at least one assigned weight to said at least one knowledge object of said knowledge graph.
8. The method of claim 1, wherein said method further comprises:
frequently monitoring at least one activity associated with said user; and
generating recommendations of said at least one knowledge object based on an output of said monitoring.

9. The method of claim 1, wherein said method further comprises dynamically updating said at least one parameter associated with said user based on said output of said monitoring.
10. A method for assigning a weight to at least one response associated with at least one node in a communication network including a plurality of nodes, the method comprises:
receiving at least one response from said at least one node, wherein said at least one response is received in response to determining an intent associated with a user;
computing at least one semantic similarity between said at least one response and each said knowledge object of at least one knowledge graph; and
assigning a weight to said at least one node associated with said at least one response based on said at least one semantic similarity.

11. The method of claim 10, wherein said at least one knowledge graph represents said at least one knowledge object associated with said node among a plurality of nodes, wherein each said node represents a duster of information.
12. The method of claim 10, wherein said method further comprises:
extracting at least one label data associated with said at least one knowledge object, wherein said at least one label data is extracted based on an ontology; and
constructing said knowledge graph by storing said extracted label data.

13. The method of claim 10, wherein said method further comprises:
identifying at least one latent data associated with said at least one knowledge object;
determining frequency of said at least one latent data appearance in a plurality of knowledge objects in a corpus;
determining a semantic similarity among said plurality of knowledge objects based on said frequency of said at least one latent data; and
constructing said knowledge graph by storing said latent data based on said semantic similarity.

14. The method of claim 10, wherein said knowledge graph comprises semantic relationship among said plurality of nodes, wherein said semantic relationship is computed, for each node pair, by comparing a type of relation associated with a first node in said node pair and a type of relation associated with a second node in said node pair.

15. The method of claim 10, wherein determining said intent comprises:
receiving at least one request from said user, wherein said request is received from an application;
extracting at least one item from said at least one request; and
determining said intent of said user based on said at least one extracted item, wherein said intent indicates at least one of: topic, subject, knowledge, and interest of said user.

16. The method of claim 10, wherein computing said at least one semantic similarity between said at least one response and each said knowledge object of at least one knowledge graph comprises:
extracting at said least one item from said at least one response; and
computing said at least one semantic similarity between said at least one item and each said knowledge object of at least one knowledge graph based on said at least one parameter associated with said at least one node.

17. The method of claim 16, wherein said parameter comprises at least one of: said node knowledge graph, topic, subject, knowledge, expertise level, and interest of said node, wherein said parameter associated with said user is stored locally on an electronic device of said user.
18. The method of claim 10, wherein said method further comprises displaying said at least one assigned weight to said at least one node associated with said at least one response based on said at least one semantic similarity.
19. The method of claim 10, wherein said method further comprises:
frequently monitoring at least one response associated with said node; and
dynamically assigning said at least one weight to said at least one node t based on an output of said monitoring.

20. A system for generating at least one recommendation for a user, the system comprising a server, an electronic device, wherein said server is configured to:
construct a knowledge graph representing at least one knowledge object associated with a node among a plurality of nodes, wherein each said node represents a duster of information;
receive at least one intent associated with said user;
assign at least one weight to said at least one knowledge object of said knowledge graph, wherein said weight is assigned based on at least one semantic similarity between said intent and each said knowledge object of said knowledge graph; and
generate at least one recommendation for said user based on said weightage, wherein said at least one recommendation comprises said at least one knowledge object.

21. The system of claim 20, wherein said server is configured to construct a knowledge graph by:
extracting at least one label data associated with said at least one knowledge object, wherein said at least one label data is extracted based on an ontology; and
constructing said knowledge graph by storing said extracted label data.
22. The system of claim 20, wherein said server is configured to construct a knowledge graph by:
identifying at least one latent data associated with said at least one knowledge object;
determining frequency of said at least one latent data appearance in a plurality of knowledge objects;
determining a semantic similarity among said plurality of knowledge objects based on said frequency of said at least one latent data; and
constructing said knowledge graph by storing said latent data based on said semantic similarity.

23. The system of claim 20, wherein said knowledge graph comprises semantic relationship among said plurality of nodes, wherein said semantic relationship is computed, for each node pair, by comparing a type of relation associated with a first node in said node pair and a type of relation associated with a second node in said node pair.
24. The system of claim 20, wherein said server is configured to assign at least one weight to said at least one object of said knowledge graph by:
extracting at least one item from said at least one intent;
computing said at least one semantic similarity between said at least one item and each said knowledge object of said knowledge graph based on at least one parameter associated with said user; and
assigning said at least one weight to each said knowledge object of said knowledge graph based on said semantic similarity.

25. The system of claim 24, wherein said parameter comprises at least one of: said user knowledge graph, topic, subject, knowledge, expertise level, and interest of said user, wherein said parameter associated with said user is stored locally on an electronic device of said user.

26. The system of claim 20, wherein said electronic device is further configured to display said at least one recommendation on said electronic device of said user, wherein said at least one recommendation is displayed along with said at least one assigned weight to said at least one knowledge object of said knowledge graph.

27. The system of claim 20, wherein said system is further configured to:
frequently monitor at least one activity associated with said user; and
generate recommendations of said at least one knowledge object based on an output of said monitoring.

28. The system of claim 20, wherein said electronic device is further configured to dynamically update said at least one parameter associated with said user based on said output of said monitoring.

29. A system for assigning a weight to at least one response associated with at least one node in a communication network including a plurality of nodes, the system comprises a server, an electronic device, wherein said server is configured to:
receive at least one response from said at least one node, wherein said at least one response is received in response to determining an intent associated with a user;
compute at least one semantic similarity between said at least one response and each said knowledge object of at least one knowledge graph; and
assign a weight to said at least one node associated with said at least one response based on said at least one semantic similarity.

30. The system of claim 29, wherein said at least one knowledge graph represents said at least one knowledge object associated with said node among a plurality of nodes, wherein each said node represents a duster of information.

31. The system of claim 29, wherein said server is further configured to:
extract at least one label data associated with said at least one knowledge object, wherein said at least one label data is extracted based on an ontology; and
construct said knowledge graph by storing said extracted label data.

32. The system of claim 29, wherein said server is further configured to:
identify at least one latent data associated with said at least one knowledge object;
determine frequency of said at least one latent data appearance in a plurality of knowledge objects in a corpus;
determine a semantic similarity among said plurality of knowledge objects based on said frequency of said at least one latent data; and
construct said knowledge graph by storing said latent data based on said semantic similarity.

33. The system of claim 29, wherein said knowledge graph comprises semantic relationship among said plurality of nodes, wherein said semantic relationship is computed, for each node pair, by comparing a type of relation associated with a first node in said node pair and a type of relation associated with a second node in said node pair.

34. The system of claim 29, wherein said server is further configured to determine said intent by:
receiving at least one request from said user, wherein said request is received from an application;
extracting at least one item from said at least one request; and
determining said intent of said user based on said at least one extracted item, wherein said intent indicates at least one of: topic, subject, knowledge, and interest of said user.

35. The system of claim 29, wherein said server is further configured to compute said at least one semantic similarity between said at least one response and each said knowledge object of at least one knowledge graph by:
extracting at said least one item from said at least one response; and
computing said at least one semantic similarity between said at least one item and each said knowledge object of at least one knowledge graph based on said at least one parameter associated with said at least one node.

36. The system of claim 35, wherein said parameter comprises at least one of: said node knowledge graph, topic, subject, knowledge, expertise level, and interest of said node, wherein said parameter associated with said user is stored locally on an electronic device of said user.

37. The system of claim 29, wherein said electronic device is configured to display said at least one assigned weight to said at least one node associated with said at least one response based on said at least one semantic similarity.

38. The system of claim 29, wherein said server is further configured to:
frequently monitor at least one response associated with said node; and
dynamically assign said at least one weight to said at least one node t based on an output of said monitoring.