Christian Rohrdantz

Visual opinion analysis of customer feedback data

Today, online stores collect a lot of customer feedback in the form of surveys, reviews, and comments. This feedback is categorized and in some cases responded to, but in general it is underutilized — even though customer satisfaction is essential to the success of their business. In this paper, we introduce several new techniques to interactively analyze customer comments and ratings to determine the positive and negative opinions expressed by the customers. First, we introduce a new discrimination-based technique to automatically extract the terms that are the subject of the positive or negative opinion (such as price or customer service) and that are frequently commented on. Second, we derive a Reverse-Distance-Weighting method to map the attributes to the related positive and negative opinions in the text. Third, the resulting high-dimensional feature vectors are visualized in a new summary representation that provides a quick overview. We also cluster the reviews according to the similarity of the comments. Special thumbnails are used to provide insight into the composition of the clusters and their relationship. In addition, an interactive circular correlation map is provided to allow analysts to detect the relationships of the comments to other important attributes and the scores. We have applied these techniques to customer comments from real-world online stores and product reviews from web sites to identify the strength and problems of different products and services, and show the potential of our technique.

Document Cards: A Top Trumps Visualization for Documents

Finding suitable, less space consuming views for a documents main content is crucial to provide convenient access to large document collections on display devices of different size. We present a novel compact visualization which represents the documents key semantic as a mixture of images and important key terms, similar to cards in a top trumps game. The key terms are extracted using an advanced text mining approach based on a fully automatic document structure extraction. The images and their captions are extracted using a graphical heuristic and the captions are used for a semi-semantic image weighting. Furthermore, we use the image color histogram for classification and show at least one representative from each non-empty image class. The approach is demonstrated for the IEEE InfoVis publications of a complete year. The method can easily be applied to other publication collections and sets of documents which contain images.

Visual sentiment analysis on twitter data streams

Twitter currently receives about 190 million tweets (small text-based Web posts) a day, in which people share their comments regarding a wide range of topics. A large number of tweets include opinions about products and services. However, with Twitter being a relatively new phenomenon, these tweets are underutilized as a source for evaluating customer sentiment. To explore high-volume twitter data, we introduce three novel time-based visual sentiment analysis techniques: (1) topic-based sentiment analysis that extracts, maps, and measures customer opinions; (2) stream analysis that identifies interesting tweets based on their density, negativity, and influence characteristics; and (3) pixel cell-based sentiment calendars and high density geo maps that visualize large volumes of data in a single view. We applied these techniques to a variety of twitter data, (e.g., movies, amusement parks, and hotels) to show their distribution and patterns, and to identify influential opinions.

Feature-Based Visual Sentiment Analysis of Text Document Streams

This article describes automatic methods and interactive visualizations that are tightly coupled with the goal to enable users to detect interesting portions of text document streams. In this scenario the interestingness is derived from the sentiment, temporal density, and context coherence that comments about features for different targets (e.g., persons, institutions, product attributes, topics, etc.) have. Contributions are made at different stages of the visual analytics pipeline, including novel ways to visualize salient temporal accumulations for further exploration. Moreover, based on the visualization, an automatic algorithm aims to detect and preselect interesting time interval patterns for different features in order to guide analysts. The main target group for the suggested methods are business analysts who want to explore time-stamped customer feedback to detect critical issues. Finally, application case studies on two different datasets and scenarios are conducted and an extensive evaluation is provided for the presented intelligent visual interface for feature-based sentiment exploration over time.

Event identification for local areas using social media streaming data

Unprecedented success and active usage of social media services result in massive amounts of user-generated data. An increasing interest in the contained information from social media data leads to more and more sophisticated analysis and visualization applications. Because of the fast pace and distribution of news in social media data it is an appropriate source to identify events in the data and directly display their occurrence to analysts or other users. This paper presents a method for event identification in local areas using the Twitter data stream. We implement and use a combined log-likelihood ratio approach for the geographic and time dimension of real-life Twitter data in predefined areas of the world to detect events occurring in the message contents. We present a case study with two interesting scenarios to show the usefulness of our approach.

Visual sentiment analysis of customer feedback streams using geo-temporal term associations:

Large manufacturing companies frequently receive thousands of web surveys every day. People share their thoughts regarding a wide range of products, their features, and the service they received. In addition, more than 190 million tweets (small text Web posts) are generated daily. Both survey feedback and tweets are underutilized as a source for understanding customer sentiments. To explore high-volume customer feedback streams, in this article, we introduce four time series visual analysis techniques: (1) feature-based sentiment analysis that extracts, measures, and maps customer feedback; (2) a novel way of determining term associations that identify attributes, verbs, and adjectives frequently occurring together; (3) a self-organizing term association map and a pixel cell–based sentiment calendar to identify co-occurring and influential opinion; and (4) a new geo-based term association technique providing a key term geo map to enable the user to inspect the statistical significance and the sentiment distribution of individual key terms. We have used and evaluated these techniques and combined them into a well-fitted solution for an effective analysis of large customer feedback streams such as web surveys (from product buyers) and Twitter (e.g. from Kung-Fu Panda movie reviewers).

Comparative Exploration of Document Collections: a Visual Analytics Approach

We present an analysis and visualization method for computing what distinguishes a given document collection from others. We determine topics that discriminate a subset of collections from the remaining ones by applying probabilistic topic modeling and subsequently approximating the two relevant criteria distinctiveness and characteristicness algorithmically through a set of heuristics. Furthermore, we suggest a novel visualization method called DiTop‐View, in which topics are represented by glyphs (topic coins) that are arranged on a 2D plane. Topic coins are designed to encode all information necessary for performing comparative analyses such as the class membership of a topic, its most probable terms and the discriminative relations. We evaluate our topic analysis using statistical measures and a small user experiment and present an expert case study with researchers from political sciences analyzing two real‐world datasets.

“Beautiful picture of an ugly place”. Exploring photo collections using opinion and sentiment analysis of user comments

User generated content in the form of customer reviews, feedbacks and comments plays an important role in all types of Internet services and activities like news, shopping, forums and blogs. Therefore, the analysis of user opinions is potentially beneficial for the understanding of user attitudes or the improvement of various Internet services. In this paper, we propose a practical unsupervised approach to improve user experience when exploring photo collections by using opinions and sentiments expressed in user comments on the uploaded photos. While most existing techniques concentrate on binary (negative or positive) opinion orientation, we use a real-valued scale for modeling opinion and sentiment strengths. We extract two types of sentiments: opinions that relate to the photo quality and general sentiments targeted towards objects depicted on the photo. Our approach combines linguistic features for part of speech tagging, traditional statistical methods for modeling word importance in the photo comment corpora (in a real-valued scale), and a predefined sentiment lexicon for detecting negative and positive opinion orientation. In addition, a semi-automatic photo feature detection method is applied and a set of syntactic patterns is introduced to resolve opinion references. We implemented a prototype system that incorporates the proposed approach and evaluates it on several regions in the World using real data extracted from Flickr.

Real-Time Visual Analytics for Text Streams

Combining automated analysis and visual-interactive displays helps analysts rapidly sort through volumes of raw text to detect critical events and identify surrounding issues.

Exploratory Text Analysis using Lexical Episode Plots

In this paper, we present Lexical Episode Plots, a novel automated text-mining and visual analytics approach for exploratory text analysis. In particular, we first describe an algorithm for automatically annotating text regions to examine prominent themes within natural language texts. The algorithm is based on lexical chaining to find spans of text in which the frequency of a term is significantly higher than its average in the document. In a second step we present an interactive visualization supporting the exploration and interpretation of Lexical Episodes. The visualization links higher-level thematic structures with content-level details. The methodological capabilities of our approach are illustrated by analyzing the televised US presidential election debates.