Tata Consultancy Services Limited

Manasi Samarth Patwardhan

Shirish Subhash Karande

Beena Rai

Parijat Dilip Deshpande

Jayita Dutta

Shankar Balajirao Kausley

Shashank Madhukar Deshmukh

Smita Subhash Ghaisas

Kanika Kalra

Abhishek Sainani

Richa Sharma

Silpa Vadakkeeveetil Sreelatha

Sai Sree Bhargav Kurma

Shrikant Arjunrao Kapse

Priya Kedia

Shabbirhussain Hamid Bhaisaheb

Lovkesh Vig

Rajaswa Ravindra Patil

Shashank Deshmuk

This disclosure relates generally to program synthesis for weakly-supervised multimodal question answering using filtered iterative back-translation (FIBT). Existing approaches for chart question answering mainly address structural, visual, relational, or simple data retrieval queries with fixed-vocabulary answers. The present disclosure implements a two-stage approach where, in first stage, a computer vision pipeline is employed to extract data from chart images and store in a generic schema. In second stage, SQL programs for Natural Language (NL) queries are generated in dataset by using FIBT. To adapt forward and backward models to required NL queries, a Probabilistic Context-Free Grammar is defined, whose probabilities are set to be inversely proportional to SQL programs in training data and sample programs from it. Compositional similarity-based filtration strategy employed on the NL queries generated for these SQL programs enables synthesizing, filtering, and appending NL query-SQL program pairs to training data, iteratively moving towards required NL query distribution.

System and method for program synthesis for weakly-supervised multimodal question answering using filtered iterative back-translation

Traditional food quality monitoring systems fail to monitor the variation of food quality in real-time scenarios. Existing machine learning approaches require dedicated data models for different classes of food items due to differences in characteristics of different food items. Also, to generate such data models, a lot of annotated data is required per food item, which are expensive. The disclosure herein generally relates to monitoring and shelf-life prediction of food items, and, more particularly, to system and method for real-time monitoring and shelf-life prediction of food items. The system generates a data model using a knowledge graph indicative of a hierarchical taxonomy for a plurality of categories of the plurality of food items, which in turn contains metadata representing similarities in physio-chemical degradation pattern of different classes of the food items. This data model serves as a generic data model for real-time shelf-life prediction of different food items.

Method and system for real-time shelf-life prediction of food items

This disclosure relates generally to visio-linguistic understanding. Conventional methods use contextual visio-linguistic reasoner for visio-linguistic understanding which requires more compute power and large amount of pre-training data. Embodiments of the present disclosure provide a method for visio-linguistic understanding using contextual language model reasoner. The method converts the visual information of an input image into a format that the contextual language model reasoner understands and accepts for a downstream task. The method utilizes the image captions and confidence score associated with the image captions along with a knowledge graph to obtain a combined input in a format compatible with the contextual language model reasoner. Contextual embeddings corresponding to the downstream task is obtained using the combined input. The disclosed method is used to solve several downstream tasks such as scene understanding, visual question answering, visual common-sense reasoning and so on.

Method and system for visio-linguistic understanding using contextual language model reasoners

This disclosure relates generally to a system and method for monitoring and quality evaluation of perishable food items in quantitative terms. Current technology provides limited capability for controlling environmental conditions surrounding the food items in real-time or any quantitative measurement for the degree of freshness of the perishable food items. The disclosed systems and methods facilitate in quantitative determination of freshness of food items by utilizing sensor data and visual data obtained by monitoring the food item. In an embodiment, the system utilizes a pre-trained CNN model and a RNN model, where the pertained CNN model is further fine-tined while training the RNN model to provide robust quality monitoring of the food items. In another embodiment, a rate kinetic based model is utilized for determining reaction rate order of the food item at a particular post-harvest stage of the food item so as to determine the remaining shelf life thereof.

System and method for monitoring and quality evaluation of perishable food items

This disclosure relates generally to data processing, and more particularly to a method and system for generating ground truth labels for ambiguous domain specific tasks. The system generates reference data corresponding to a regulation statement being processed, using a crowd sourcing mechanism and then processes the reference data using an Expectation Maximization (EM) model. The EM model determines consensus with respect to ambiguity of terms/phrases, validity of questions, and validity of answers, and then based on the determined consensus, provides questions and answers to disambiguate the regulation statement.

Method and system for generating ground truth labels for ambiguous domain specific tasks

Growing community of inventors

Pune, India

Manasi Samarth Patwardhan