Sonia Khetarpaul

Bio: Sonia Khetarpaul is an academic researcher from Shiv Nadar University. The author has contributed to research in topics: Computer science & Recommender system. The author has an hindex of 4, co-authored 15 publications receiving 72 citations. Previous affiliations of Sonia Khetarpaul include Indian Institute of Technology Delhi & Indian Institutes of Technology.

Mining GPS data to determine interesting locations

Abstract: It is possible to obtain fine grained location information fairly easily using Global Positioning System (GPS) enabled devices. It becomes easy to track an individual's location and trace her trajectory using such devices. By aggregating this data and analyzing multiple users' trajectory a lot of useful information can be extracted. In this paper, we aim to analyze aggregate GPS information of multiple users to mine a list of interesting locations and rank them. By interesting locations we mean the geographical locations visited by several users. It can be an office, university, historical place, a good restaurant, a shopping complex, a stadium, etc. To achieve this various relational algebra operations and statistical operations are applied on the GPS trajectory data of multiple users. The end result is a ranked list of interesting locations. We show the results of applying our methods on a large real life GPS dataset of sixty two users collected over a period of two years.

Bus Arrival Time Prediction Using a Modified Amalgamation of Fuzzy Clustering and Neural Network on Spatio-Temporal Data

Abstract: This paper presents a dynamic model that can provide prediction for the estimated arrival time of a bus at a given bus stop using Global Positioning System (GPS) data. The proposed model is a hybrid intelligent system combining Fuzzy Logic and Neural Networks. While Neural Networks are good at recognizing patterns and predicting, they are not good at explaining how they decide their input parameters. Fuzzy Logic systems, on the other hand, can reason with imprecise information, but require linguistic rules to explain their fuzzy outputs. Thus combining both helps in countering each other’s limitations and a reliable and effective prediction system can be developed. Experiments are performed on a real-world dataset and show that our method is effective in stated conditions. The accuracy of result is 86.293% obtained

SHEG: summarization and headline generation of news articles using deep learning

Abstract: The human attention span is continuously decreasing, and the amount of time a person wants to spend on reading is declining at an alarming rate. Therefore, it is imperative to provide a quick glance of important news by generating a concise summary of the prominent news article, along with the most intuitive headline in line with the summary. When humans produce summaries of documents, they not only extract phrases and concatenate them but also produce new grammatical phrases or sentences that coincide with each other and capture the most significant information of the original article. Humans have an incredible ability to create abstractions; however, automatic summarization is a challenging problem. This paper aims to develop an end-to-end methodology that can generate brief summaries and crisp headlines that can capture the attention of readers and convey a significant amount of relevant information. In this paper, we propose a novel methodology known as SHEG, which is designed as a hybrid model. It works by integrating both extractive and abstractive mechanisms using a pipelined approach to produce a concise summary, which is then used for headline generation. Experiments were performed on publicly available datasets, viz. CNN/Daily Mail, Gigaword, and NEWSROOM. The results obtained validate our approach and demonstrate that the proposed SHEG model is effectively producing a concise summary as well as a captivating and fitting headline.

Mining GPS traces to recommend common meeting points

Abstract: Scheduling a meeting is a difficult task for people who have overbooked calendars and many constraints. The complexity increases when the meeting is to be scheduled between parties who are situated in geographically distant locations of a city and have varying travel patterns. In this paper, we present a solution that identifies a common meeting point for a group of users who have temporal and spatial locality constraints that vary over time. The problem entails answering an Optimal Meeting Point (OMP) query in spatial databases. Under Euclidean space OMP query solution identification gets reduced to the problem of determining the geometric median of a set of points, a problem for which no exact solution exists. The OMP problem does not consider any constraints as far as availability of users is concerned whereas that is a key constraint in our setting. We therefore focus on finding a solution that uses daily movements information obtained from GPS traces for each user to compute stay points during various times of the day. We then determine interesting locations by analyzing the stay points across multiple users. The novelty of our solution is that the computations are done within the database by using various relational algebra operations in combination with statistical operations on the GPS trajectory data. This makes our solution scalable to larger groups of users and for multiple such requests. Once this list of stay points and interesting locations are obtained, we show that this data can be utilized to construct spatio-temporal graphs for the users that allow us efficiently decide a meeting place. We perform experiments on a real-world dataset and show that our method is effective in finding an optimal meeting point between two users.

Object and Currency Detection with Audio Feedback for Visually Impaired

Abstract: In recent years, orientation and navigation technologies and support for people with visually impaired disabilities have increased. This paper proposes a model for the detection of objects for visually impaired individuals. We have developed a module for blind people to identify and match items scanned to a list of objects, such as currency notes, person and bottle, and the angle and distance of the object being detected from the camera. The model is based on image recognition from a database of objects. The results of the image recognition shall be communicated to the user by system audio feedback.

Availability, Reliability, and Security in Information Systems and HCI

Abstract: In a hospital, information exchange is essential to save lives and to prevent life-endangering mistakes. Information exchange is supported by a hospital information system (HIS). From a theoretical perspective, the deployment of an HIS is promising because it reduces errors and duplication of information. In practice, however, there are some major problems concerning the usage of such a system. One way to deal with these problems is introduced in this paper: the integration of sensor data into social media. The paper concentrates on the conceptual benefits and risks such an integration may generate. It focuses on the case of a patient room.

Inferring high-level behavior from low-level sensors

Abstract: We present a method of learning a Bayesian model of a traveler moving through an urban environment. This technique is novel in that it simultaneously learns a unified model of the traveler’s current mode of transportation as well as his most likely route, in an unsupervised manner. The model is implemented using particle filters and learned using Expectation-Maximization. The training data is drawn from a GPS sensor stream that was collected by the authors over a period of three months. We demonstrate that by adding more external knowledge about bus routes and bus stops, accuracy is improved.

Bus arrival time prediction using mixed multi-route arrival time data at previous stop

Abstract: The primary objective of this paper is to develop models to predict bus arrival time at a target stop using actual multi-route bus arrival time data from previous stop as inputs. In order to mix and fully utilize the multiple routes bus arrival time data, the weighted average travel time and three Forgetting Factor Functions (FFFs) – F1, F2 and F3 – are introduced. Based on different combinations of input variables, five prediction models are proposed. Three widely used algorithms, i.e. Support Vector Machine (SVM), Artificial Neutral Network (ANN) and Linear Regression (LR), are tested to find the best for arrival time prediction. Bus location data of 11 road segments from Yichun (China), covering 12 bus stops and 16 routes, are collected to evaluate the performance of the proposed approaches. The results show that the newly introduced parameters, the weighted average travel time, can significantly improve the prediction accuracy: the prediction errors reduce by around 20%. The algorithm comparison demonstrates that the SVM and ANN outperform the LR. The FFFs can also affect the performance errors: F1 is more suitable for ANN algorithm, while F3 is better for SVM and LR algorithms. Besides, the virtual road concept in this paper can slightly improve the prediction accuracy and halve the time cost of predicted arrival time calculation. First published online 02 May 2017

Unsupervised Interesting Places Discovery in Location-Based Social Sensing

Abstract: This paper presents an unsupervised approach toaccurately discover interesting places in a city from location-basedsocial sensing applications, a new sensing applicationparadigm that collects observations of physical world fromLocation-based Social Networks (LBSN). While there are alarge amount of prior works on personalized Point of Interests(POI) recommendation systems, they used supervised learningapproaches that did not work for users who have little orno historic (training) data. In this paper, we focused onan interesting place discovery problem where the goal isto accurately discover the interesting places in a city thataverage people may have strong interests to visit (e.g., parks, museums, historic sites, etc.) using unsupervised approaches. Inparticular, we develop a new Physical-Social-aware InterestingPlace Discovery (PSIPD) scheme which jointly exploits thelocation's physical dependency and the visitor's social dependencyto solve the interesting place discovery problem using anunsupervised approach. We compare our solution with state-of-the-art baselines using two real world data traces from LBSN. The results showed that our approach achieved significantperformance improvements compared to all baselines in termsof both estimation accuracy and ranking performance.