Juan D. Velásquez

Opinion Mining and Information Fusion

A survey on Information Fusion applied to Opinion Mining.Information Fusion techniques in the Opinion Mining context.Text mining algorithms to extract relevant opinions from documental data bases. Interest in Opinion Mining has been growing steadily in the last years, mainly because of its great number of applications and the scientific challenge it poses. Accordingly, the resources and techniques to help tackle the problem are many, and most of the latest work fuses them at some stage of the process. However, this combination is usually executed without following any defined guidelines and overlooking the possibility of replicating and improving it, hence the need for a deeper understanding of the fusion process becomes apparent. Information Fusion is the field charged with researching efficient methods for transforming information from different sources into a single coherent representation, and therefore can be used to guide fusion processes in Opinion Mining. In this paper we present a survey on Information Fusion applied to Opinion Mining. We first define Opinion Mining and describe its most fundamental aspects, later explain Information Fusion and finally review several Opinion Mining studies that rely at some point on the fusion of information.

A novel deterministic approach for aspect-based opinion mining in tourism products reviews

This work proposes an extension of Bing Lius aspect-based opinion mining approach in order to apply it to the tourism domain. The extension concerns with the fact that users refer differently to different kinds of products when writing reviews on the Web. Since Lius approach is focused on physical product reviews, it could not be directly applied to the tourism domain, which presents features that are not considered by the model. Through a detailed study of on-line tourism product reviews, we found these features and then model them in our extension, proposing the use of new and more complex NLP-based rules for the tasks of subjective and sentiment classification at the aspect-level. We also entail the task of opinion visualization and summarization and propose new methods to help users digest the vast availability of opinions in an easy manner. Our work also included the development of a generic architecture for an aspect-based opinion mining tool, which we then used to create a prototype and analyze opinions from TripAdvisor in the context of the tourism industry in Los Lagos, a Chilean administrative region also known as the Lake District. Results prove that our extension is able to perform better than Lius model in the tourism domain, improving both Accuracy and Recall for the tasks of subjective and sentiment classification. Particularly, the approach is very effective in determining the sentiment orientation of opinions, achieving an F-measure of 92% for the task. However, on average, the algorithms were only capable of extracting 35% of the explicit aspect expressions, using a non-extended approach for this task. Finally, results also showed the effectiveness of our design when applied to solving the industrys specific issues in the Lake District, since almost 80% of the users that used our tool considered that our tool adds valuable information to their business.

FOCAS AUTOMATIC CATALOG MATCHING ALGORITHMS

This paper describes efficient algorithms that automatically take two or more catalogs of objects with instrumental coordinates and magnitudes and matches them. The challenges are that the instrumental coordinates may be only partially overlapping, at a different scale, rotated, or even inverted (flipped). The object magnitudes may be derived from different passbands so that the relative magnitudes of the objects differ. Also, the catalog may not contain all the same objects to to differences in separating close objects or to partial overlap between images. Finally, the catalog positions and magnitudes are subject to noise in the images from which they were derived. The algorithms are applicable to any automated cataloging system. However, the implementation described here is part of the Faint Object Classification and Analysis System (FOCAS). FOCAS automatically produces catalogs of objects from digital images. The algorithms described here first take a subsample of the brightest objects from the catalog, and other catalogs. Then all the objects in the catalogs are matched based on the transformed reference coordinates.

Text mining applied to plagiarism detection: The use of words for detecting deviations in the writing style

Plagiarism detection is of special interest to educational institutions, and with the proliferation of digital documents on the Web the use of computational systems for such a task has become important. While traditional methods for automatic detection of plagiarism compute the similarity measures on a document-to-document basis, this is not always possible since the potential source documents are not always available. We do text mining, exploring the use of words as a linguistic feature for analyzing a document by modeling the writing style present in it. The main goal is to discover deviations in the style, looking for segments of the document that could have been written by another person. This can be considered as a classification problem using self-based information where paragraphs with significant deviations in style are treated as outliers. This so-called intrinsic plagiarism detection approach does not need comparison against possible sources at all, and our model relies only on the use of words, so it is not language specific. We demonstrate that this feature shows promise in this area, achieving reasonable results compared to benchmark models.

Identifying Customer Preferences about Tourism Products Using an Aspect-based Opinion Mining Approach☆

Abstract In this study we extend Bing Lius aspect-based opinion mining technique to apply it to the tourism domain. Using this extension, we also offer an approach for considering a new alternative to discover consumer preferences about tourism products, particularly hotels and restaurants, using opinions available on the Web as reviews. An experiment is also conducted, using hotel and restaurant reviews obtained from TripAdvisor, to evaluate our proposals. Results showed that tourism product reviews available on web sites contain valuable information about customer preferences that can be extracted using an aspect-based opinion mining approach. The proposed approach proved to be very effective in determining the sentiment orientation of opinions, achieving a precision and recall of 90%. However, on average, the algorithms were only capable of extracting 35% of the explicit aspect expressions.

Combining eye-tracking technologies with web usage mining for identifying Website Keyobjects

This paper introduces a novel approach for collecting and processing data originated by web user ocular movements on a web page, which are captured by using an eye-tracking tool. These data allow knowing the exact web users eye position on a computer screen, and by combining them with the sequence of web page visits registered in the web log, significant insights about his/her behavior within a website can be extracted. With this approach, we can improve the effectiveness of the current methodology for identifying the most important web objects from the web users point of view, also called Website Keyobjects. It takes as input the websites logs, the pages that compose it and the interest of users in the web objects of each page, which is quantified by means of a survey. Subsequently, the data are transformed and preprocessed before finally applying web mining algorithms that allow the extraction of the Website Keyobjects. With the utilization of the eye-tracking technology, we can eliminate the survey by using a more precise and objective tool to achieve an improvement in the classification of the Website Keyobjects. It was concluded that eye-tracking technology is useful and accurate when it comes to knowing what a user looks at and therefore, what attracts their attention the most. Finally, it was established that there is an improvement between 15% and 20% when using the information generated by the eye tracker.

Web User Session Reconstruction Using Integer Programming

An important input for Web usage mining is Web user sessions that must be reconstructed from Web logs (sessionization) when such sessions are not otherwise identified. We present a novel approach for sessionization based on an integer program. We compare results of our approach with the timeout heuristic on Web logs from an academic Web site. We find our integer program provides sessions that better match an expected empirical distribution with about half of the standard error of the heuristic.

Towards the identification of keywords in the web site text content: A methodological approach

Since the creation of the web, the designers are looking for friendlier ways of make web page contents, which pictures, sounds, movies and free texts attract the users’ interest. Special attention receive the text content, because is the most frequently parameter used to retrieve information from the web. A simple way in order to understand the user’s text preferences, could be collect the words used in a searching. However, this information is only well‐know for the owner of the specific searching engine. In this paper we introduce a methodology in order to extract the most interest words for a user in a particular web site, based of the user browsing behavior and the web page text content. The methodology was tested using data originated in a bank web site showing the effectiveness of our approach.

Using Self Organizing Feature Maps to Acquire Knowledge about Visitor Behavior in a Web Site

When a user visits a web site, important information concerning his/her preferences and behavior is stored implicitly in the associated log files. This information can be revealed by using data mining techniques and can be used in order to improve both, content and structure of the respective web site.

A Knowledge Base for the maintenance of knowledge extracted from web data

By applying web mining tools, significant patterns about the visitor behavior can be extracted from data originated in web sites. Supported by a domain expert, the patterns are validated or rejected and rules about how to use the patterns are created. This results in discovering new knowledge about the visitor behavior to the web site. But, due to frequent changes in the visitors interests, as well as in the web site itself, the discovered knowledge may become obsolete in a short period of time. In this paper, we introduce a Knowledge Base (KB), which consists of a database-type repository for maintaining the patterns, and rules, as an independent program that consults the pattern repository. Using the proposed architecture, an artificial system or a human user can consult the KB in order to improve the relation between the web site and its visitors. The proposed structure was tested using data from a Chilean virtual bank, which proved the effectiveness of our approach.