Web Based Interfaces

From CS2610 Fall 2014
Revision as of 17:44, 2 December 2014 by Jingtao Wang (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

slides1 slides2

Contents

Readings


Reading Critiques

Wenchen Wang 17:02:53 11/21/2014

<Opinion Space: A Scalable Tool for Browsing Online Comments > <Summary> This paper introduces a new visualization technique to represent people’s opinion called opinion space. <Paper Review> Every user’s opinion is represented by one dot. Similar opinion dot close to each other. And opposite opinion dots are separated apart. This kind of technique has been applied to US government political opinion web page. The key point of opinion space is to reduce a five-dimension space to two-dimension space, because each opinion is represented by x and y value in the final display. A technique called principal component analysis (PCA) is used in this paper to reduce dimensionality. In this way, people can focus their interesting opinions and ignore the unconcerned ones. I did a class project, which is similar to this idea in visualization class last semester. Similarly, we represent the questions on Stack Overflow to facilitate people for finding questions relevant to their own and exploring similar questions. The result is very promising and we got a good score in that project. <Zoetrope: Interacting with the Ephemeral Web> <Summary>This paper introduces a visualization technique Zoetrope to visualize the historical web pages. <Paper Review> I never saw this kind of idea before. I think it is very interesting. Zoetrope helps better multi-version web page management and utilization. Zoetrope allows users to review historical web page by move the slider of temporal lens. Lenses apply filters, such as filtering on time, keyword, amounts and their combination. In this way, users could find some relationship with the incidents of other web page. For example, by reviewing the traffic of a certain area on google map, people could find at which time, the traffic is heavy and find some regular pattern. And they will find the heaviness of the traffic is related to the baseball game. Another example, by reviewing the price of a product, user could review the past price of the same product to decide whether the buy it, comparing past prices.

yubo feng 1:16:53 11/24/2014

"Opinion Space: A Scalable Tool for Browsing Online Comments In this work, the authors present Opinion Space, an online tool that offers a new perspective for sharing opinions and comments. The authors also conduct a user study that evaluates the user engagement and usefulness of the system. The most attractive point of the system is it's neat user interface that uses a projected 3D space visualization. Based on the affordances from distance, size, and brightness, the system signifies the similarity/dissimilarity and insightfulness of the others' opinions. I believe the user can experience definitely different and, in some sense, better interaction with the others, through this ""exploring"" metaphor. The system requires apparently more actions from the user -- not only reading and writing, the user needs to assess the others' opinions. This would be another threshold to enter the system. However, this won't be a big matter if we consider the increasing participations and contributions of the Internet users. One obvious drawback of this technique is that, since every opinion is represented as a dot in the space, the user can't take advantages of the ""traditional"" online bulletin boards. For example, in the traditional board, one may see the titles of certain amount of postings at a glance. However, since all titles hide out, one can't take benefits from ""glancing"". Definitely, one may need more clicks or mouse movements to explore different postings. Also, I'd like to point out that the authors' argument that Opinion Space enables users to read a significantly more diverse set of comments seems contradict and meaningless. The visualization of the system is based on the distance between opinions. Thus, everything that Open Space highlights is the similar opinion and dissimilar opinions will be displayed as smaller and darker dots. In this case, it sounds more correct that clicks on the dissimilar dots happen rather random, not due to the design of Open Space. In addition, if it's random, there would be no differences between Open Space and the other perspectives (List or Grid). I believe this is what their statistical results signify (p-values around the hypothesis 3, on page 1181). However, the authors keep arguing that the user using the Space would read more diverse comments. I consider this is a groundless assertion and a wrong conclusion. In the article “Zoetrope: Interacting with the Ephemeral Web” by Adar, Dontcheva, Fogarty, and Weld a system is presented that enables interaction with historical web pages, links, and data. The system, Zoetrope, uses a set of operators to manipulate content streams allowing the user to interact with historical web data that would otherwise be lost due to how dynamic the web can be. Zoetrope gives users something that was not possible before, allowing them to access past versions of pages by using a lens that scans backwards through time. The way the authors structured the interface by using lenses and filters seem like a simple interaction technique that a user would be able to easily pick up on. The system implements three temporal lenses: visual, structural, and textual. The visual lenses allow the user to specify a region on the original page and by using a slider the user can view past visual data. The structural lenses allow the user to make a precise selection that track selected HTML content that is independent of the visual position. The textual lenses allow the user to track text regardless of where it appears on the page. The filters allow the user to focus on specific items of interest such as time, keywords, and amounts. Zoetrope collects data by crawling pages at semi-regular intervals. The authors state that the current interval that they crawl at is one hour but it could be determined by the amount of content change. It would be interesting for future work to develop an efficient algorithm that can determine when Zoetrope should go out and collect data on specific pages. I found the article to be lacking from an evaluation standpoint. The article goes into a lot of detail about implementation of the system and how the system functions but not much of a system evaluation. Part of the reason is that the concept of the system is new but it would have been interesting to see more of the reasoning behind some of the design decisions the authors made. I do however feel that Zoetrope is a good idea as it adds time to viewing content on the Web. "

Longhao Li 15:50:57 11/24/2014

Critique for Opinion Space: A Scalable Tool for Browsing Online Comments In general, this paper introduced a new system that makes the online comments visualized for users, which have great performance. The author also talked about the experiment that they conduct to evaluate if this system is good enough to beat the stage of the art. This new system is a very important achievement. Since nowadays there are more online comments generated than before, and this number is increasing. Viewing comments turn out to be a hard job for users. List view of the comments takes a lot of time for the users to conclude if the product is good or not. This new system, Opinion Space, make the viewing of online comments become easier. It highlights the comments found most insightful by users in a space that full of points of reviews. Users can easily browsing the reviews and focus on the highlight review to decide if they want to purchase the item. To evaluate their system, the author did some user study that comparing Opinion Space with list view and grid view. The result shows that Opinion Space turn out to be most satisfied by the users. It may change people’s understanding on viewing comments in the future. This new method to do visualization of contents may be able to expend to some new area other than online comments. For example, in education, after the grading of homework, if the common problems can be show in this way, instructor will be able to easily extract what should be explain again for the class to make most of the students understand what the materials said. Moreover, in medication, highlighting where new disease happened on map or highlighting the frequency of flu happen in worldwide can help doctors easily analysis the situation and make the right decision to treat patients and prevent disease expend. Critique for Zoetrope: interaction with the Ephemeral Web This paper introduced a new system that works to give user the chance to see the web content in the past. Since web content changing fast at this age, this system can help user to find information from Internet easier. Zoetrope is a system that helps users to see the web content not only the fresh one, but also the content in the past. This is a very useful tool for people to find content on the web. When people find a link in some published article that go public for several years, the link may not be able to pointing to the right content. At this time, Zoetrope will be lifesaver. Also some website update very frequently so that when people see some content that were interesting yesterday, they may not find it again in the future if they don’t using some tool like Zoetrope. The author not only introduced the idea, but also enhances it and made it be possible to use properly. The author added Lenses in the system, which is used to display the history of a certain area in the web content. It also embedded some filters to help user to find some certain content. It included visual lenses structural lenses and also textural lenses. These three lenses works together to help user to targeting what they want to see. There are a lot of filters in the system that based on time, keyword, amounts etc. Also users can bind different lenses and filters together to do accurate search. With these filters and lenses user can easily find what they want to see. The system also includes the feature that can help to see the timeline visualization of the change of content in some website. It enhanced the function of the system. To my understanding, some similar system have already in the market, they may not work on web content. They may works on history of change for local machine, like Apple’s Time machine. But I think this system works better since it gives the user more way to do the search. I think this function can be extended into other system.

phuongpham 20:03:26 11/24/2014

Opinion Space: A scalable tool for browsing online comments: the paper introduces opinion space, a GUI plotting comments on a 2D space with customized clusterring method. The authors have addressed some problems of current network communication and conducted user study to explore the advantages of Opinion Space. This is a standard research project where the authors conduct a user study, analyse the result based on their hypothesis. Further analysis was left to future work. It's interesting to know from the paper that there are several statistical test for hypothesis verification. However, I have some concerns about the experiment. First, it would be better if the starting question (Figure 2) focus on a topic rather than many unrelated topics. In other words, if we have experts identifying important aspects of a topic and collect users' opinion on those aspects, the plotting graph will be meaningful. We can also do it automatically but this is still an on going research in extracting important topics, aspects of a discussion. Second, I have a feeling that an interested readers would want to read most (or all) comments. And this time, it seems to be difficult to view every single topic in the opinion space compared to the traditional list view. We may address this problem by providing summary of a "closed" cluster or present a list of all comments in a cluster. Third, can we plot the opinion space in a 3d instead of 2d space? It may introduce other interesting questions to be answered. ***Zoetrope: Interacting with the ephemeral web: this paper presents zoetrope, a tool to visualize temporal web information. The authors have introduced lenses and filters as means to view temporal information on webpages. The technique is based on information retrieval over a set of temporal instances of the same webpage. The users can use lenses combine with filters to view interested information overtime. This is another application of history tool that we have seen in outpost and a tool for drawing. As standard historical tool, the authors use slider, image shots and support filters. The idea using filters and historical over webpage is interesting. However, I have some concerns about the approach. First, why do we use lense? We can view a total view of a webpage change overtime without the computational cost to extract a portion of a webpage. Using keyword search is not good because there are ambigious as the authors have pointed out, using HTML structure may depend on that HTML element. If the authors can give a good example why do we want to view only a smal portion of a webpage, I think the paper will become more convience. Second, I feel this approach will work well with a smaller application, e.g. tracing changes of documents inside an organization or personal. WWW is an overshot because we can't control the update time. While within a restricted environment, we can ask the source to push whenever changes occur. Third, the semantics mentioned in the paper depend much on DOM structure. However, with current webpage content, not all important information is tagged. Therefore, we can't use information as we hope with Zoetrope.

Yanbing Xue 21:59:55 11/24/2014

The first paper is mainly about a comments section in which users can leave notes for other uses about their thoughts on the subject being discussed. These comment sections are useful for gathering user opinions, especially as (nearly) the entire world has access to the internet, allowing many multiple viewpoints. Unfortunately, this is a double-edged sword. Because so many have access to comment, the comments section can quickly become extremely long; thousands of comments may be left on a single article within a matter of hours. These can be sorted in a variety of ways. Comments facilitate a basic human function in cyberspace: expressing opinion. The current “standard” listing system is inefficient. It favors the most recent posts and the most fringe opinions. Opinion space gets around this by presenting all comments at once in two-dimensional space with a third value designating their importance. However, such kind of interfaces are not able to scale very well and often ignore the in force of extreme opinions. opinion space is built to solve this problem. The design is delivered from previous work such as deliberative polling, dimension reduction and collaborative filtering. The aim of opinion space is to allow users to visualize and navigate through a diversity of comments. A good benefit from Opinion Space is that users are able to view comments from a range of perspectives. Users were significantly more engaged with the Space and Grid interfaces. Users would read significantly more diverse comments using the Space interface. It is inspiring to learn how the authors designed the system based on previous work the the experiments to evaluate their interface from many different aspects. ========== The second paper is mainly about a powerful tool to interact with webs. The idea behind Zoetrope is to provide a graphical tool that can track the changes of multiple websites and their elements through time. Traditionally, when we view the websites, we can only view the current screenshot of the page. However, it would be very helpful if we could also see the changes of this website in history. The biggest difficulty in implementing this technique is the crawling of the history pages. Crawling pages in real time would bring a very long latency, which would not be appreciated by the user. So the way the users do is that they take advantage of two plugins, screen grabbing, web page dumping. Zoetrope is based on a set of operators for manipulating content streams. In the paper, the author describes these primitives and the associated indexing strategies for handling temporal Web data. They form the basis of Zoetrope and enable our construction of new temporal interactions and visualizations. While the lenses define the region of content to be tracked, the secondary interaction of filtering and combining lenses define how the content is tracked. The filter allow people to track information based on time, keyword, amount, and also provide compound filters, trigger filters. Zoetrope also allows lenses binding and stacking to enrich the filtering condition.

Bhavin Modi 22:09:02 11/24/2014

Reading Critique On Opinion Space: A Scalable Tool for Browsing Online Comments A brief summary of the paper, creating a new platform for browsing online comments and comparing with existing implementation to see the effects of their innovative solution. Moving on to discussing exactly what the authors proposed, is called Opinion Space. The authors have highlighted the existing problems in online discussion and commenting forums, most of which use the list view to display the comments, which creates a problem trying to find the ones you may want to read, in simple words there is no flexibility. We have seen previously the paper on Stack Overflow, but that mainly included technical problems and people were not interested in diversity but specific problems solutions. Ratings help in this respect but not when on topics on politics where the views of people are more diverse and there is no one correct view. So now we have opinion space where they use PCA to map the diversity of opinions in relation to the user’s profile on a two dimensional scale. Distance indicates the how much the opinions differ and larger, brighter coloured spaces indicate opinions many people agree with and/or respect. This make the whole experience intuitive and gives the control to the user, I can say this by actually using it. Though from experiments conducted we see that there is no statistical significance when it comes to increasing the diversity of comments read by the user, but that has a subtle problem that even in list view comments are ordered randomly. The main idea lies in having control on the variation of diversity you want to view, from extremes to minor difference in opinions. Some problems that the authors have identified have proposed solutions in the future works of the paper, one of which creating a scale to view exactly where each opinion lies and add some background information, which could make using the application more informative by inserting landmarks. The second work proposed we have seen that idea does increase user interest which is including some competitive perks, like badges and ranks again like Stack Overflow. Finally to conclude the paper is well written in terms of clearly defined hypothesis and strong experimental data that is thoroughly researched and presented. A little more emphasis on the motivation and clearly defining existing solutions and their solution in respect to them would have helped clear exactly the contribution of the paper. -------------------------------------------------------------------------------------------------------- Reading Critique On Zoetrope: Interacting with the Ephemeral Web Zoetrope is a software created to interact with historical representations of web pages for purposes of collecting, viewing and representing relative data in an easily available and sophisticated manner. The paper literally encompasses all the details of implementation of the Zoetrope including the algorithms and libraries used, the concepts applied to solve problems and the even the languages. Though the application has been well constructed and is very innovative, there is no detailed study on the background of the research, again the existing solutions and the comparison that can be made between Zoetrope and them. The reason of creating such an application also remains unclear and there are existing methods to collect such data as analysis are generally maintained on sports sites, and new stories too are linked as a timeline. But one aspect of paper that is outstanding is that they have discussed every possible problem and how they have fixed it and tried to solve it. These problems are using filters and different types of lenses, Memory demands of such an implementation, the capturing mechanism modifications that are possible. The major issue that not every page remains fixed, modifications to the layout of webpages (Remains largely unaffected but the DOM almost vanishes after a year, but the location of elements have some consistency which helps) has been discussed extensively with data being provided. One other extension that we can use is maybe let the users view their own activity over time, and maybe other people activities too. This would corresponding to providing a mechanism that conforms to the previous papers on design space and history. No user study have been provided, maybe because this application is the first of its kind, but there is the mentioning of AT&T Difference Engine, so some baseline is there and they could compare the effectiveness of their application in a real life setting. Scenarios of usage provide a clear visualization of where and when the application can be used, so maybe more than saying that they created something that’s better, the focus is on new interaction techniques and innovations.

Eric Gratta 23:57:16 11/24/2014

Opinion Space: A Scalable Tool for Browsing Online Comments (2010) Siamak Faridani, Ephrat Bitton, Kimiko Ryokai, Ken Goldberg This paper offers a novel system for filtering and presenting comments to website users called Opinion Space. This topic is immediately interesting, as the issue of “cyberpolarization” mentioned in the paper is familiar to any frequent Internet user. What the authors seem most focused on is ensuring that thoughtful, moderate comments make their way on to the users’ screen in spite of a plethora of extreme and hateful comments. This work falls under the category of “discourse architecture” as an attempt to facilitate large-scale conversations. Interesting about this system is that, at least at first, it was released to the public for any users to come and contribute comment ratings. The system seems to explore different reward systems for encouraging good behavior in commenting and exposure to a diversity of comments. In the related work, authors draw on related work on social network analysis, sentiment analysis, polling techniques, social psychology, collaborative filtering, and other social computing concepts. To feasibly display a map of other users’ comments in Opinion Space, a mathematical technique called principal component analysis was used which was well-explained with Figure 4. The experimental setup was unique in that there were a total of five hypotheses considered and tested. This paper described its experimental setup well, and went into incredible detail about all of the statistical techniques used. Inferential statistics were calculated on all five of the hypotheses, and discussion on each was insightful for how users interact with comments. -------------------------------------------------------------------------- Zoetrope: Interacting with the Ephemeral Web Eytan Adar, Mira Dontcheva, James Fogarty, Daniel S. Weld This paper starts by identifying a unique feature of the Internet – its rapid change. Because websites are so frequently altered/updated, the authors believe that being able to navigate the “historical Web” would grant access to information that is currently lost in time. Providing the utmost clarity, this paper simply lists its contributions in a bulleted list. Features are well-explained with their figures. A figure should have been included for the structural lens feature because it was not as clear as others. The description of the Zoetrope system is grounded in the concept of content (data) streams, making it easier for other computer scientists to understand; all of the features are explained in terms of tuples and their transformations, concepts familiar from data stream management. The work in this paper seemed to be approaching the problem from the “wrong” direction. While the utility of many of the features of Zoetrope is clear, I think that adding historical features to websites would cover anything that Zoetrope is accomplishing. Site-specific historical features would also provide more consistent quality than Zoetrope because they would be developed with the specific topic domain of the website in mind. This paper sheds light on how historical access to the Web could be useful, but the history that the authors are interested in is largely content. These websites probably store historical content data anyway and could serve it on user request if historical access features and queries were added to their sites. The paper does at least contribute, then, methods for manipulating data streams.

Yingjie Tang 0:01:38 11/25/2014

“Opinion Space: A Scalable Tool for Browsing Online Comments” is a paper which proposed an approach to visualize the online comments. This paper give me a insightful idea in the comment analysis area. There are a lot of work like the Discourse Architecture and and Poligraph who have been invented by other researchers. The main technique in this area is not complicated, they use Dimensionality Reduction which is a mature method in machine learning area. When I read this paper, I feel that the fatal weakness of their work may lay in the evaluation part. Since the user study of this work is to compare the Score of ‘List’, ‘Grid’ and ‘Space’ Approach, how does the data mean by an enhancement of 30%. And there are so many related work in these area as the author mentioned, they can compare their method with some existing method and see what happens. In a word, the baseline is not strong enough to make a big contribution. Moreover, the evaluation part may be a tricky part. The data collected are called ‘Average Dwell Time’, ‘Average Agree with Rating’. I think what does the data mean in these part still remains un-unanswered. Does it mean the approach is useful when the users spend more dwelling time? Maybe the motivation is different for different users. Maybe some users wants to find some comments with high rating and this approach is the same as the ‘List’ approach. The paper also mentioned this point from the data collected from the questionnaire. People are accustomed to the ‘List’ way of the comments and the ‘List’ has a better affordance than the ‘Space’. There is one thing which really impressed me. The part of summary of results is novel to me because it adopt a hypothesis defend way. They proposed five hypothesis before the experiment and the defend their hypothesis with their data. I think this is a good way to analyze data because the user will be more clear about what the conclusions the author are going to draw from the data when they first see the hypothesis.—————————————————— ‘Zoetrope: Interacting with the Ephemeral Web’ is a paper which present an approach to enable users interact with historical Web. This is a paper of description and without any user study. I think the main reason for not adopting user study is that this approach is a new approach to address this problem. Users can choose the areas they want to explore and move the slider to see the historical data of the web pages. Although the author presented many ways to use the Zoetrope, the main goal he/she want to do is to verify that this approach is a valid solution to historical Web data. When I first read this paper and come to their idea to show the history data of the Web a concern appears in my mind: How could they dispose with the large amount of data for the Website’s history. However, at the implementation part, they address this problem. Since they only to store the whole page of the original site and don’t have to store the css style files for the descendent history. And a HTML will become around 30% size without the css style files. It seems that my concerned problem has been resolved. However, the changes of the websites are not just the data in the same CSS style. There are the cases that the web designers changed the whole style of the site. What if we get the data with ‘Zoetrope’ and with an un compatible CSS file? In this case, I think the regional query can still work while the whole history website will be invalid. We can project the single data like the data of stock changes and the data of oil price. These single data are tractable even thought the CSS style file changed. Although there exists some weakness, I like this idea not jus it proposed a valid solution to solve the ephemeral web but because it is one of the most profound manipulation of virtual reality that can really make the Web the most versatile source that could be used for any kind of research that one does.

zhong zhuang 0:16:21 11/25/2014

This article is about scalable online comments browsing system. In today’s Internet, users are allowed to leave comments for news, blogs, videos and virtually everything. When the number of comments hit certain level, the current comments listing will not scalable well. Just list comments in chorological order will leads to many problems. One of them is “cyberpolarization” which means comments will converge into two extreme views, and eventually become a “flame war”. The author proposed a new system: Opinion Space. This system is based on deliberative polling and opinion mining. In deliberative polling, users are polled on two sets of polls, after the first poll, the users are given a period of deliberative time, and then present the second poll. In this way, people will tend to avoid the extreme views. The system also apply visualization techniques to project a five dimensional profile to a two dimensional plane. The system display comments as points on a screen, if two points are far away from each other, that means the two user have different opinions. if a point has halo, that means the user is rated as active user. The bigger halo is, the more active the user is. Then the author conducted a user study, the study shows great improvement of the commenting system.

Mengsi Lou 1:29:46 11/25/2014

Opinion Space: A Scalable Tool for Browsing Online Comments Nowadays many website and bbs have the form of comments that the users can make comments of some article or topic, and also refer to other people’s comments. But the most common interface is a list sorted by time or by ratings. This pattern can not server well to users. There are problems with existing comments system. The first problem is that thoughtful moderates are often shouted down by extremists. Second, the amount of data can be overwhelming. Opinion Space is a new online tool designed to collect and visualize user opinions on topics ranging. This system aims to solve the problems by incorporating ideas from deliberative polling, dimensionality reduction, and collaborative filtering. For the design of the Opinion Space, there are several parts. When the users enters opinions they are s presented with five opinion profile propositions and asked to rate them on a continuous scale. And what’s interesting is that the map is a projection of the five- dimensional opinion profiles onto two dimensions using a technique known as principal component analysis. The user study worth study carefully. First they adopt three interface that are list interface, grid interface and space interface. Then they made five hypotheses. Here list only the first one that is Opinion Space will be significantly more engaging than List or Grid in terms of average dwell time and in terms of user ranking of overall preference. Then for the method, they designed a within-subject study using the Space interface as the experimental condition and the List and Grid interfaces as two control conditions. Evaluation of Hypotheses part, they used Analysis of Variance ANOVA for significance. Given n data sets, these tests produce a p-value that estimates the probability that the outcome is by chance that the sets were sampled from the same distribution; known as the null hypothesis. Lower p-values correspond to greater significance of the data. To assess hypothesis 1, they consider the data collected from the exit survey question that asked participants to rank the three interfaces by preference. The results of this analysis mildly support hypothesis 1 (for List-Space p- value= 0.01188, for Grid-Space p-value= 0.03884 and for Grid-List, p-value = 0.0209). /////////////////////////////////////////// Zoetrope: Interacting with the Ephemeral Web This paper discusses about the system Zoetrope that enables interaction with the historical Web (pages, links, and embedded data) that would otherwise be lost to time. The temporal Web can be manipulated, queried, and analyzed from the context of familiar pages. Zoetrope is based on a set of operators for manipulating content streams. The Zoetrope architecture includes a Web crawler that collects data, a set of databases to store collected content, and an interface that provides access to the stored data. Lenses enable viewing of Web content from different moments in time, but this exploration may be just the first part of satisfying an information need. Visualizations, like lenses, create a sequence of transforms, filters, and renderers to display results. Although visualizations can exist independently of a lens, a lens typically defines the data displayed in the visualization. Lenses can thus also be used as a prototyping tool for testing selections and aggregations.

Wei Guo 1:30:32 11/25/2014

Reading Critique for Zeotrope Zeotrope is a tool for interacting with the historical Web that would otherwise be lost to time. Zeptrope presents a new way to think about historical web data. It adds dimension of time to the familiar pages. Zeotrope also implements a number of interactions based in lenses and visualization that allow people to access and analyze historical information through contextualized selections and extractions. I watched the video from https://www.youtube.com/watch?v=7C-B7qdClak It presents the content in a webpage. User can define an area of the webpage, then it will allow user to drag a button to return to any specific history time of this area to see the content at that time. This is a 2D webpage, and adding a time axis in. Is that possible that we make it as a 3D visualization? Reading Critique for Opinion Space Opinion space is a new online interface incorporating ideas from deliberative polling, dimensional reduction, and collaborative filtering that allows participants to visualize and navigate through a diversity of comments. This Opinion space will automatically highlights the comments found most insightful by user rating from a range of perspectives. For the study design part, instead of having one hypotheses with several dimensions, the authors choose to have 5 hypothesizes. In this way, the analysis can be clear and concentrate on each point instead of messing everything up together. I think we can also try to do it this way in our final project. The Dimensional Reduction is another interesting point to me. I have never heard the low variance projection and maximal variance project before. The two pictures clearly explain the concepts. According to Wikipedia: “dimensional reduction or dimension reduction is the process of reducing the number of random variables under consideration, and can be divided into feature selection and feature extraction.” The calculation is hard. The understanding is simple. Projecting from different directions will make the different variances of some projects placed in 3D.

SenhuaChang 2:04:46 11/25/2014

The first paper "Opinion Space: A Scalable Tool for Browsing Online Comments " shows us a new online tool, which is incorporates ideas from deliberative polling, dimensionality reduction, and collaborative filering and also create a new interface for viewing online comments.It visualizes the comments into dots in a huge space and uses some similarity metrics to measure the relationship between different users. I think the most important part in this paper is the user study. They create three interfaces (includes this space interface) to make an evaluation. They use List and Grid interface to be the control groups. What’s more, the authors apply lots of statistical approaches to analyze the user study data, which provide a new direction when we want to evaluate our hypotheses. I think in this Opinion Space, organization may be a problem in it. When we have lots of opinions from different persons, there will be lots of dots in that space. So the traditional pointing problem will show up. < << >> > The second paper "Zeotrope: Interacting with the Ephemeral Web" shows a novel system called Zeotrope that enables interaction wich the historical web that would otherwise be lost to time. The idea behind Zoetrope is to provide a graphical tool that can track the changes of multiple websites and their elements through time. This allows for the quick cross-referencing of information across a large spectrum of data sources and quickly provides graphical analysis of the information. The user interface seems to be a patch programing UI associated with a 'sand box' type web viewer that allows multiple web pages to be rendered into a single working space. Zoetrope uses a number of novel interactions, the temporal Web can be manipulated, queried, and analyzed from the context of familar pages.

zhong zhuang 2:42:51 11/25/2014

This article is about a system for interacting with the ephemeral web. The word ephemeral is perfect for describing the World Wide Web today. Ten years from now, web is mostly static, information on websites will not change too much in a long time. But the situation changed dramatically in the past ten years. Now, most web site are dynamic, news website will change everyday, micro blogs such as tweeter will change every seconds. Most web sites only display the current state. Users are very difficult to view past versions. The authors presents a system to enable an easy way to view web site in a temporal space. The system is called Zoetrope. In order to view web sites in a multi-version fashion, Zoetrope implements a number of interactions based on “lens”. To view different type of web sites, users can use different lens. The system introduce a novel semantic for dealing with temporal web data and show a small number of operators that can be used for many different tasks. This language is tested and proven to be very easy to use.

Brandon Jennings 5:09:25 11/25/2014

Opinion Space With the popularity of the internet for expressing opinions, there becomes an increasing need for extracting useful data. The problem is that infrastructures like forums and blogs generally have a linearly organized layout for comments usually sorted by more recent post. On of the unique characteristics of the system proposed here is the combined features of questionnaire and textual comments. What this does is essentially give the comment some sort of weight. It gives the comment a rating based on the combination of different context questions associated with it. This is important for searching for relevant topics and comments. It also gives a basis for querying comments, using the questions and responses as filters. The scalability of this system is also bonus. Zoetrope Maybe as a developer one might find use for this system, but I do not see a whole lot of utility for this. The concept of keeping track of the past has been explored in some of our earlier papers. Those were about keeping track of stages in a development process. However any critical information that is needed could retrieved other than keeping track of websites. Interesting comparisons can be made using Zoetrope between the current state and past state of a website. I think what is more interesting in this paper is the visualization technique. The organization of windows and ability to manipulate visualizations are worth investigating.

Jose Michael Joseph 6:46:41 11/25/2014

Open Space: A scalable tool for browsing online comments This paper is primarily about the system Opinion Space that enables the users to see through the diversity of comments for a particular application without getting caught up viewing only those comments that are there in majority. It therefore saves users from being cyber polarized. There are many other research papers that were published before this that was focused on a similar topic. Deliberative Polling and Opinion Mining is one such task. Its drawback, according to the authors, is that it has an inherent bias. The authors have not explained why this bias exists and that would have been a good addition to this paper. One of the features of Opinion Space is that the users have to rate on various subjects and give comments regarding the same. But these comments can be changed at any time. Such a feature is volatile as people would change the comments frequently thereby decreasing the engagement of the page due to this increased confusion. The user study conducted by the authors involved 12 people out of which 10 were men and 2 women. This in itself poses a problem because if there is any gender specific interaction, the skewed ratios would ensure that one aspect of this experiment is under represented. Also all the participants were college students and were asked on issues such as marijuana legalization which is something that is particularly favorable with the younger population. For a more diverse user study for a system of such a large scale, more people should have been chosen from diverse backgrounds. The primary hypothesis of this paper, that users will be able to see more useful comments by using Open Space, was only marginally proven by their user study. Thus it shows that the system still has a lot of room for improvement as compared to its traditional counterpart.

Jose Michael Joseph 6:47:15 11/25/2014

Zoetrope: Interacting with the ephemeral web This paper is about using a system that stores historical content of websites so as to transcend the ephemeral nature of web page and enable the user to see content that was part of the website in the past but is not so anymore. The first feature the authors describe is a slider that controls how back the user can take the system and thus see the different changes in the website over time. This is done by the use of a web crawler that takes this information and stores it in a database. Lens are features which are used to restrict the content selection so as to see the modification of only that content over the long period of time. There are many forms of lenses but all have the same primary purpose. The use of lenses are that they help to restrict the selection and thus we can see the changes that take place in a particular segment of the page. The authors also introduce the concept of using visualization tools to further show the data the user needs in terms that the user can easily understand. The reasoning behind this is that the user will be more comfortable looking as graphs as opposed to simple changes in content over time. Unfortunately, although these ideas do sound cool and innovative there are already websites such as the WayBack machine that precisely do the basic functionality of viewing site history as shown in this paper.

Xiyao Yin 8:12:14 11/25/2014

‘Opinion Space: A Scalable Tool for Browsing Online Comments’ presents Opinion Space: a new online interface incorporating ideas from deliberative polling, dimensionality reduction, and collaborative filtering that allows participants to visualize and navigate through a diversity of comments. This is a scalable way to visualize the ‘opinion landscape’ and to operate as a self-organizing system that encourages participants to find and consider comments written by those who hold opinions different from their own. With Opinion Space, authors aim to address the above problems by incorporating ideas from deliberative polling, dimensionality reduction, and collaborative filtering. Authors created three interfaces: List, Grid, and Space and popluated each with a set of 200 randomly selected user comments from ‘in the wild’ experiment and recorded data from a random order of 12 study participants. They considered five hypothesis and the results used ANOVA on each hypothesis. This paper has a large part of future work, which provides many chances for future research. It shows that Opinion Space has advantages in reducing the barrier to entry by making the interface easy to use and more intuitive. The classification of three types of users: casual users, authors and gamers in this paper is also a good point. This can also be a good hint for our current project in HCI class. ‘Zoetrope: Interacting with the Ephemeral Web’ present Zoetrope, a system that enables interaction with the historical Web(pages, links, and embedded data) that would otherwise be lost to time. The World-Wide Web is highly dynamic, for a person who does not have access to historical data, or would find it difficult or impossible to recreate the Web is primarily one-dimensional, containing only one version of any page. With Zoetrope, one can explore past versions of entire pages or place lenses on any part of a webpage to scan backwards through time. Familiar pages, which have previously existed only in the present, now have the added dimension of time. Zoetrope architecture includes a Web crawler and an interface. Figures in this paper are concise and descriptive and we can know the Zoetrope easily. Futhermore, developers can use Zoetrope to rapidly prototype and test temporal queries through Zoetrope’s visual query interface. It seems that problems of big data needs to be concerned in Zoetrope.

yeq1 8:37:40 11/25/2014

Yechen Qiao Review for 11/25/2014 Zoetrope: Interacting with the Ephemeral Web In this paper, the authors introduced a new toolkit that allows the users to analyze features of websites across time. The system discretely samples through a continuous stream of websites across time. The analytics are done abstractly through conversion of time streams into operators and then displayed by renderers. Visual, structural and textual lens can be used to provide nested filters prior to visualization. Multiple visualization techniques are implemented to analyze different images. For example, timelines and movies are used to track changes of images across time, while clustering and time series allows visualization of grouped data list. While the main application doesn’t seem to be all that novel, I especially liked the introduction of this paper. This is one example where good anecdotes can create connections between the paper and the reader. In security publications, this technique is widely used and we even have agreed on different names for people in different roles (Alice, Bob, and Mallory, and Mallory is always the bad person). Another interesting point about this paper is the detail in describing the system architecture. This is somewhat typical for a system paper, but I haven’t seen many HCI papers like this. The architecture was first designed abstractly and mathematically before we get to see the details. This gives the implementation theoretical grounding on correctness. The implementation is further strengthened by consideration and trials of multiple alternatives. For example, the authors found out that using blank image to represent randomization may cause significant disruption to the user experience, and as such, the alternative of provide the immediate historical data item is more appealing. Yet another example is between the alternatives of displaying tracked changes on a page and displaying those changes on an item. The only problem I can find with this paper is that the motivation itself is not sufficiently argued: I know websites change constantly, but I don’t know why I would ever be bothered to track changes on a webpage. (I can think of multiple reasons, but none of them seems strong enough.) I actually participated in a similar argument before and the response I got was most research are useless anyway. I think this is an excuse and not a defense for lacking sufficient motivations. Opinion Space: A Scalable Tool for Browsing Online Comments In this paper, the authors attempted to resolve cyberpolarization in online comments to provide a friendlier discussion environment and more rationalized opinions. They proposed an alternative visualization technique that uses PCA to provide dimensionality reduction and provide two dimensional view of the opinion space across all users about the topic. They compared the approach with list and grid on multiple hypotheses ranging from usability and sociology of the community. In this paper, the discussion is modular, with each module being a different hypothesis. One of the common problem with research papers is that often they claim the approach is better is some aspects, but the difference is too small for it to gain traction as users have already grown accustomed to existing approach. Often, the new approach will require large amount of improvement to compensate for experience, potentially due to power law of practice? When this is not possible, the authors may circumvent this issue by listing multiple advantages instead of a single advantages, and argue instead that this approach is significantly better in multiple aspects. In this type of paper, the rigorousness of statistical analysis becomes an important part of the paper. However, the author chose to address “how well it works” more so than “why it works”, with little emphasis on discussions on pros and cons of each dimensionality reduction alternatives and simply chose a widely used technique PCA. It is unclear why the largest eigenvector works in this approach, especially how eigenvector on high variance relates to changes in user friendliness and moderation. I think it may be possible that this is because the question is too difficult to answer, and the authors simply used heavier evaluation to compensate for this…

Qiao Zhang 9:27:42 11/25/2014

Opinion space: a scalable tool for browsing online comments Their goal is to go beyond visualization of relationships in the graph, to find a more effective way to explore and analyze interactions in social networks. In other words, to learn about the relationships between each comment in a 2D space rather than in a linearly time fashion. The distance represents the similarity between two opinions, and all comments are displayed in a 2D space. Personally, I think this idea is better when applied to large displays where comments are no longer dots but actual texts. In their user study, the linear approach forces the users to browse comments one by one by using blurring. However, it is not true most of the time, because a user should be able to jump back and force as well as skim and skip between comments. Thus, I believe that representing comments in actual text, which requires more display space, is a better idea than representing in dots, which requires extra clicking/hovering efforts and prevents users from skimming. One technique used in this paper is that the author conducted two-way ANOVA to test if there exists carry-over effect in their user study, which is very persuasive. The data analysis part is a bit confusing to me because they conducted a handful of different tests for significance. I am wondering how to properly choose the type of significance test for different data, since different tests have different assumptions. The authors seem to be good in statistics, however I am wondering where should we start as a student in interdisciplinary area. =============================== Zoetrope: Interacting with the Ephemeral Web In this paper, the authors present a system called Zoetrope that enables interaction with the historical Web that would otherwise be lost to time. Using a number of novel interactions, the temporal Web can be manipulated, queried, and analyzed from the context of familar pages. Users can explore past versions of entire pages or place lenses on any part of a webpage to scan backwards through time by using a tooltip view called lenses. Furthermore, developers can use Zoetrope to rapidly prototype and test temporal queries through Zoetrope’s visual query interface. With the Zoetrope system, we introduce a novel semantics for dealing with temporal Web data and show a small number of operators that can be utilized for many different tasks. The idea of Zoetrope is tempting, however, as I observe, the changing of webpages sometimes is not relevant to its structure, i.e. the DOM tree. Most of the time, a newer version of webpage will result in complete deletion of nodes, new structure and so on. In this situation, the system might not be able to perform well.

Vivek Punjabi 9:54:13 11/25/2014

Opinion space: a Scalable tool for browsing online comments: The paper presents a new online interface for comments, Opinion Space, using deliberative polling, dimensionality reduction, and collaborative filtering. It allows participants to visualize and navigate through various comments. It is designed to collect and visualize user opinions on variety of topics. It provides an interactive map with many points each representing an user and comment, and with different color codes. The user activities included entering opinions using horizontal sliders and a textual comment, and browse and rate comments of other users. The final map projects this five dimensional opinion profiles into two dimensional map using principal component analysis. The user study was performed to compare list interface, grid interface and space (similar to Opinion Space) interface. The results well supported all the hypothesis except for the 5th which produced similar results as grid layout. Users found space interface more engaging than others, significantly. The idea for adding a scoring model to this interface is pretty amazing and motivational. This tool can also be used to understand user opinions and their behaviors using some learning algorithms. As its can collect lots of user data, it can be used further to enhance user experience. Zoetrope: interacting with the ephemeral web: The authors present a system, Zoetrope, that enables users to interact with the historical web, that is allow users to view old pages or links even after they are modified. It also allows users to place lenses on any part of a webpage to scan backwards through time. Its architecture includes a Web crawler that collects data, a set of databases to store collected content, and an interface that provides access to the stored data. It provides users with several temporal lenses such as visual, structural and textual lenses, which are sufficient for tracking and extracting a wide range of data types within a page. It also provides filters to focus on specific information of interest. Zoetrope has a web crawler that crawls the web at semi-regular intervals in order to track the frequent changes of any web pages. The visualizations are in the form of timelines, clustering and time series. Thus, Zoetrope adds a new dimension of time to the web and makes it more intuitive. The idea looks interesting and unique, however, I would like to know more applications of this technique in real time which can help motivate to move forward in this research direction. Also, it might lead to some storage problems as we will be storing all the changes at some servers and storing all the web pages and its metadata can be expensive.