Abstract: A method and a system of selective encryption of a test dataset is disclosed. In an embodiment, the method may include determining a relevancy grade associated with each of a plurality of datapoints within a test dataset by comparing the test dataset with a common heat map, wherein the common heat map is generated using a plurality of training datasets. The method may further include calculating, based on the relevancy grade, an encryption level associated with each of the plurality of datapoints. The method may further include selectively encrypting at least one datapoint from the plurality of datapoints based on the encryption level associated with each of the plurality of datapoints. The at least one data point is rendered to a user after being decrypted.

Abstract: This disclosure relates to method and system for providing explanation for output generated by artificial intelligence (AI) model. The method may include receiving encrypted input data and a public encryption key from a client device, wherein the encrypted input data is encrypted using the public encryption key. The method may further include generating an encrypted AI model by encrypting an AI model using the public encryption key. The method may further include generating an encrypted output and an encrypted feature data based on the encrypted input data using the encrypted AI model, and generating an encrypted explanation for the encrypted output based on the encrypted feature data. The method may further include providing the encrypted output and the encrypted explanation to the client device for rendering, wherein the encrypted output and the encrypted explanation are decrypted by the client device using a private encryption key.

Abstract: The present disclosure relate to effectively determining spatial and temporal features for extracting low-level and high-level features from image frames of a multimedia content. A plurality of image frames are received from an imaging unit. Spatial filters are applied on each image frame to generate a first set of activation maps which provide spatial features in the image frames. Further, a temporal filter is applied on the plurality of image frames at a plurality of levels to generate one or more second set of activation maps corresponding to each level for determining temporal features in the image frames. Thereafter, the spatial feature from each image and temporal feature of the plurality of image frames from each level is extracted, which represent low-level and high-level spatial and temporal features.

Abstract: A method and a system of determining a relation between two or more entities in a text document is disclosed. In an embodiment, the method may include receiving training text data annotated with two or more entities, and creating one or more n-grams based on the training text data. The method may further include generating a Convolutional Neural Network (CNN) model using the one or more n-grams, and creating an entity vector using at least one of a word embedding and a numeric embedding based on the training text data. The method may further include generating a relation-entity model using the CNN model and the entity vector.

Abstract: A method and a system for controlling devices through voice interaction are disclosed. In an embodiment, the method may include identifying at least one feature of a target device and an action to be performed on the at least one feature, based on an intent and an object determined from a voice input received from a user. The method may further include determining a correspondence between the at least one feature and the action to be performed using a trained neural network. The method may further include comparing a current operational state of the at least one feature with an operational threshold of the at least one feature. The method may further include performing the action on the at least one feature based on the determined correspondence, when the current operational state is within limits of the operational threshold.

Abstract: The present invention relates to the field of dynamic content creation and management and more particularly to dynamically create and insert immersive promotional content in a multimedia. The multimedia requested by a user is procured from a media server (105) and the one or more objects present in the multimedia is determined. The content of the multimedia is analyzed, and a decision is taken to immerse the promotional content in the multimedia based on the analysis of the multimedia. Further, a promotional content is created in real time using the components present in the asset database (106) based on the personalized user profile. The created promotional content is overlaid and blended with a at least one object from the one or more objects in the multimedia. The created promotional content and the multimedia is joined to form a composite media and is streamed to the user device.

Abstract: A method and a system of selective encryption of a test dataset is disclosed. In an embodiment, the method may include determining a relevancy grade associated with each of a plurality of datapoints within a test dataset by comparing the test dataset with a common heat map, wherein the common heat map is generated using a plurality of training datasets. The method may further include calculating, based on the relevancy grade, an encryption level associated with each of the plurality of datapoints. The method may further include selectively encrypting at least one datapoint from the plurality of datapoints based on the encryption level associated with each of the plurality of datapoints. The at least one data point is rendered to a user after being decrypted.

Abstract: The present disclosure relate to effectively determining spatial and temporal features for extracting low-level and high-level features from image frames of a multimedia content. A plurality of image frames are received from an imaging unit. Spatial filters are applied on each image frame to generate a first set of activation maps which provide spatial features in the image frames. Further, a temporal filter is applied on the plurality of image frames at a plurality of levels to generate one or more second set of activation maps corresponding to each level for determining temporal features in the image frames. Thereafter, the spatial feature from each image and temporal feature of the plurality of image frames from each level is extracted, which represent low-level and high-level spatial and temporal features.

Abstract: A method and a system of determining a relation between two or more entities in a text document is disclosed. In an embodiment, the method may include receiving training text data annotated with two or more entities, and creating one or more n-grams based on the training text data. The method may further include generating a Convolutional Neural Network (CNN) model using the one or more n-grams, and creating an entity vector using at least one of a word embedding and a numeric embedding based on the training text data. The method may further include generating a relation-entity model using the CNN model and the entity vector.

Abstract: The present invention relates to the field of dynamic content creation and management and more particularly to dynamically create and insert immersive promotional content in a multimedia. The multimedia requested by a user is procured from a media server (105) and the one or more objects present in the multimedia is determined. The content of the multimedia is analyzed, and a decision is taken to immerse the promotional content in the multimedia based on the analysis of the multimedia. Further, a promotional content is created in real time using the components present in the asset database (106) based on the personalized user profile. The created promotional content is overlaid and blended with a at least one object from the one or more objects in the multimedia. The created promotional content and the multimedia is joined to form a composite media and is streamed to the user device.

Abstract: This disclosure relates to method and system for providing explanation for output generated by artificial intelligence (AI) model. The method may include receiving encrypted input data and a public encryption key from a client device, wherein the encrypted input data is encrypted using the public encryption key. The method may further include generating an encrypted AI model by encrypting an AI model using the public encryption key. The method may further include generating an encrypted output and an encrypted feature data based on the encrypted input data using the encrypted AI model, and generating an encrypted explanation for the encrypted output based on the encrypted feature data. The method may further include providing the encrypted output and the encrypted explanation to the client device for rendering, wherein the encrypted output and the encrypted explanation are decrypted by the client device using a private encryption key.

Abstract: A method and system are described for generating structured relations between words in sentences. The method includes generating encoded hidden state vectors using a single layer bi-directional Long Short Term Memory (LSTM) neural network. The method includes generating current hidden state vectors based on word embedding associated with each word at a time stamp ‘t’. The method includes computing attention distribution of each word based on encoded hidden state vectors and current hidden state vectors. The method includes computing context vector of sentences based on attention distribution of each word and the encoded hidden state vectors. The method includes computing vocabulary distribution at time stamp “t” based on context vector and current hidden state vectors. The method includes computing probability distribution of words based on encoded hidden state vectors, current hidden state vectors, and vocabulary distribution.

Abstract: A method and a system for controlling devices through voice interaction are disclosed. In an embodiment, the method may include identifying at least one feature of a target device and an action to be performed on the at least one feature, based on an intent and an object determined from a voice input received from a user. The method may further include determining a correspondence between the at least one feature and the action to be performed using a trained neural network. The method may further include comparing a current operational state of the at least one feature with an operational threshold of the at least one feature. The method may further include performing the action on the at least one feature based on the determined correspondence, when the current operational state is within limits of the operational threshold.

Abstract: An electronic apparatus for providing an on-screen keyboard including a plurality of keys are provided. The electronic apparatus comprises a touch interface configured to receive a touch input of a user and a processor configured to, in response to the touch input being received though the touch interface, determine a touch area where the touch input is received, in response to the plurality of keys are included in the touch area, identify a key corresponding to a touch pattern of the user among the plurality of keys, and display the identified key on a display of the electronic apparatus.