Durga Toshniwal

Feature based Summarization of Customers’ Reviews of Online Products☆

Abstract With the growing availability and popularity of opinion-rich resources such as review forums for the product sold online, choosing the right product from a large number of products have become difficult for the user. For trendy product, the number of customers’ opinions available can be in the thousands. It becomes hard for the customers to read all the reviews and if he reads only a few of those reviews, then he may get a biased view about the product. Makers of the products may also feel difficult to maintain, keep track and understand the customers’ views for the products. Several research works have been proposed in the past to address these issues, but they have certain limitations: The systems implemented are completely opaque, the reviews are not easier to perceive and are time consuming to analyze because of large irrelevant information apart from actual opinions about the features, the feature based summarization system that are implemented are more generic ones and static in nature. In this research, we proposed a dynamic system for feature based summarization of customers’ opinions for online products, which works according to the domain of the product. We are extracting online reviews for a product on periodic bases, each time after extraction, we carry out the following work: Firstly, identification of features of a product from customers’ opinions is done. Next, for each feature, its corresponding opinions’ are extracted and their orientation or polarity (positive/negative) is detected. The final polarity of feature-opinions pairs is calculated. At last, feature based summarizations of the reviews are generated, by extracting the relevant excerpts with respect to each feature-opinions pair and placing it into their respective feature based cluster. These feature based excerpts can easily be digested by the user.

Aspect based Summarization of Context Dependent Opinion Words

Abstract Popularity and availability of opinion-rich resources in e-commerce platform is growing rapidly. Before buying any product, one is interested to know the opinion of other people about that product. For any product, there are hundreds of reviews available online so it becomes very difficult for the customers to read all the reviews. Also, one cannot set his mind based on reading some of the review since it gives him a biased view about that product. So we need to automate this process. As we know, there are lots of opinion words present in the sentences of a review which will tell about the polarity of that product. Out of all the opinion words, some words behave in the same manner means they have the same polarity in all contexts, but some words are context dependent means they have different polarity in different context. In this paper, we proposed an Aspect Based Sentiment Analysis and Summarization (ASAS) System, which handles the context dependent opinion words that has been the cause of major difficulties. For finding the opinion polarity, first, we used an online dictionary for classifying the context independent opinion word. Second, we used natural linguistic rules for assigning the polarity to maximum possible context dependent words. These steps create the training data set. Third, for classification of the remaining opinion words, we used opinion words and feature together rather than opinion words alone, because the same opinion word can have different polarity in the same domain. Then we used our Interaction Information method to classify the feature-opinion pairs. Fourth, as negation plays a very crucial role, we found negation words and flipped the polarity of the corresponding opinion word. Finally, after classifying each opinion word, the system generated a short summary for that particular product based on each feature

Geospatial sentiment analysis using twitter data for UK-EU referendum

Abstract Brexit i.e. “British Exit” is one of the major events in the history of economics and British politics. When EU referendum took place, 52% of votes were in favor of United Kingdom leaving the European Union. This was a major event affecting the overall economy of Britain and was also one of the most talked about events. Twitter is a social network platform where users from all over the world discussed a lot about Brexit and British politics. Analysis of such user generated text can reveal a lot about the event and what people think of it. Sentiment analysis is a trending technique to get insights from any written text. Our aim in this study is to analyze the tweets geospatially and then perform Geo-spatial Sentiment analysis based on the location of the events verses the geospatial tweets distribution for that particular event on global level. Also shows the keywords and hashtags were mostly used by people during that time and how they were being used i.e. which hashtag was used positively, which one was used negatively and which one carried neutral sentiment with itself. This study also tries to find out the British politicians who were being talked about the most and what people think of them sentimentally. Also, the tweets intended for these famous British politicians are analyzed geospatially and their sentiment distribution is visualized. The geospatial sentiment analysis of the whole dataset for “leave” and “remain” tags is also plotted on the atlas map.

Parallel Framework for Unsupervised Classification of Seismic Facies

Seismic facies classification plays a critical role in characterizing & delineating the various features present in the reservoirs. It aims at determining the number of facies & their description for the available seismic data. During the past few decades, seismic attributes have been widely used for the task of seismic facies identification. It helps geologists to determine different lithological and stratigraphical changes in the reservoir. With the increase in the seismic data volume & attributes, it becomes difficult for the interpreters to examine each seismic line. One of the solutions given to this problem was to use some computer-assisted methods such as k-means, self-organizing map, generative topographic map and artificial neural network for analyzing the seismic data. Even though these computer-assisted methods performed well but due to the size of the 3-D seismic data the overall classification process becomes very protracted. In this paper, we introduce a parallel framework for unsupervised classification of the seismic facies. The method begins by calculating four different seismic attributes. Spark & Tensorflow based implementation of unsupervised facies classification algorithms are then used to identify the seismic facies based on the 4-D input attributes data. Further, the comparison of results (in terms of execution time & error) of Spark & Tensorflow based algorithms with already existing approach show that the proposed approach provides results much faster than previously existing MPI based approach.

Predicting Particulate Matter for Assessing Air Quality in Delhi Using Meteorological Features

Air pollution is one of the biggest threats to the environment. According to statistics of World Health Organization, more than 80% of people living in urban areas inhale poor air quality levels. Hence assessing air quality is important especially in urban areas where people suffer more health problems due to poor air quality. Data mining techniques can serve to be very useful for analyzing the air quality data. In the past, several research works were done for various developing countries of the world, except a few for developing countries, like India. Specifically for Delhi, where high concentrations of Oxides of Nitrogen, Oxides of Sulphur, Benzene, Toluene, Particulate Matter etc. are reported in its atmosphere. The presence of certain meteorological conditions in the atmosphere can be very helpful to identify the presence of such pollutants. Particulate matter with a diameter of 2.5 \(\upmu \)m or less (\(PM_{2.5}\)) is focused upon in this work. Data mining techniques like multivariate linear regression model and regression trees etc. to identify the relationship between meteorological features and air quality are deployed. Further, the use of ensemble techniques such as random forests are also given in the present research work. Evaluation is done over root mean square error metrics and results are found to be promising.

Location Prediction Using Sentiments of Twitter Users

This study aims to predict the next location of a twitter user only by using his past tweets. Twitter is a very popular micro-blogging platform and a lot of people tweet about different topics varying from personal day-to-day activities to some global event. This provides us with the opportunity to perform sentiment analysis on their past tweets for prediction of their next visit. Sentiment analysis helps in revealing the opinion, desire or intentions of a person looking at the text that they write. In this paper, a new model called Sentiments based Labeled LDA model (SLLDA) is proposed to predict users’ next location within a given geo-spatial range. This kind of prediction can be used by various establishment owners for the targeted promotions of their products. This can also be helpful for personalized recommendation. Various experiments have been performed to evaluate the performance of the proposed model. The proposed model outperforms in every set of experiments and is better than each baseline model considered in the study. The accuracy comparison has also been done for different window lengths of past tweets and different radii of query. The performance of the proposed model turned out to be better for each set of experiments.

Identifying the Local Business Trends in Cities Using Data Mining Techniques

With the growing popularity of social media, and several users using them, humongous amount of information is being generated. This information may be 140 words like tweets, posts, and images shared on Facebook or reviews written on Yelp. It would be valuable to both consumers and businesses if the current local business trends followed by the people of different cities can be identified. Local business reviews are available, which when mined can be used to find out the local business trends across cities. With this information, the users can watch out for the prevalent local businesses (yoga, beauty and spas, ballet, restaurant, etc.) in each city, and the businesses can chart their b-plans, accordingly. To accomplish this, the present work uses data mining technique of clustering on benchmark dataset.

Visibility Prediction in Urban Localities Using Clustering

Various research works in computer science utilize data generated from cities, to make it smart and intelligent. Some of these works make use of data mining, sensor networks, internet of things, web of things, cloud computing techniques and machine learning techniques. In this work, smart mobility using data mining is mainly focused upon. Smart mobility is one of the crucial aspects of smart city addressing efficient movement of people and goods from one place to another. In the present work, several literature works on smart mobility have been discussed along with suggested improvement in mobility for several locations in India. This work focuses upon the issue of reduction in visibility in environment, which is caused by presence of certain atmospheric conditions. Reduction in visibility hinders traffic and causes accidents, thus affecting smooth movement of people and goods. The present work examines the humidity content of several locations in India, specifically metropolitan city of Bangalore have been considered in this research. Furthermore, clustering have been performed to investigate the humidity trends at these locations and results are found to be promising.

Privacy Preserving Data Mining Techniques for Hiding Sensitive Data: A Step Towards Open Data

Privacy Preserving Data Mining (PPDM) is an area that deals with data mining techniques that allow the data to be mined while keeping its privacy intact. PPDM comes into play when different parties want to involve in collaborative data mining and wish to collectively mine their data to extract knowledge while keeping their data private. The preservation of privacy of data is also of utmost importance when we want to publish any data as open data. This study includes a detailed review of some of the recent techniques proposed in this area. The biggest challenge in devising any privacy preserving data mining technique is to achieve a good balance between privacy and utility of the privacy preserved data.

Large-Scale Distributed Sparse Class-Imbalance Learning

Abstract Class-imbalance learning is a classic problem in data mining and machine learning community. In class-imbalance learning, the idea is to learn the model so that it performs equally well on all the classes. Most of the work in literature so far have tackled this problem either in a centralized way or the work is limited to a particular domain such as intrusion detection. In the present paper, we propose to solve the class-imbalance learning problem on large-scale sparse data in a distributed setting. More specifically, we partition the data across examples and distribute each chunk of the data to different processing nodes. Each node runs a local copy of FISTA-like algorithm which is a distributed implementation of the prox-linear algorithm for cost-sensitive learning. We show the efficacy of the proposed approach on benchmark and real-world data sets and compare the performance with the state-of-the-art techniques in the literature.