Nonnegative matrix factorization (NMF) can be formulated as a minimization problem with bound constraints. Although bound-constrained optimization has been studied extensively in both theory and practice, so far no study has formally applied its techniques to NMF. In this letter, we propose two projected gradient methods for NMF, both of which exhibit strong optimization properties. We discuss efficient implementations and demonstrate that one of the proposed methods converges faster than the popular multiplicative update approach. A simple Matlab code is also provided.

/pdf/projected-gradient-methods-for-nonnegative-matrix-4jkbx8akf5.pdf

Projected Gradient Methods for Nonnegative Matrix Factorization

Cross-language information retrieval (CLIR) is an active sub-domain of information retrieval (IR). Like IR, CLIR is centered on the search for documents and for information contained within those documents. Unlike IR, CLIR must reconcile queries and documents that are written in different languages. The usual solution to this mismatch involves translating the query and/or the documents before performing the search. Translation is therefore a pivotal activity for CLIR engines. Over the last 15 years, the CLIR community has developed a wide range of techniques and models supporting free text translation. This article presents an overview of those techniques, with a special emphasis on recent developments.

Translation techniques in cross-language information retrieval

A document localization, management and delivery system in a computer environment. A preferred embodiment of the invention automatically determines the language and country of a Web site visitor and directs the Web server to deliver the appropriate localized content contained in a country/language database to the visitor's browser. The visitor's browser is notified of the proper font and content encoding needed to display the selected language and is allowed to download the font. A toolkit is provided which allows a master site to be built that is language and country-independent. The actual language and country content is placed in a language/country database where it is easily managed and maintained. When a visitor enters the site, the requested document is automatically served in the visitor's language and for the visitor's country by filling in a document template from the master site with the correct language content from the language/country database. A viewer allows the developer to view and debug the document template as it appears to a visitor in any of the available language content from the language/country database.

Multilingual electronic document translation, management, and delivery system

A translation management system in a computer environment. A preferred embodiment of the invention automatically detects when a document, data stream, or non-text file in the master language has been updated and notifies the user which corresponding documents, data streams, or non-text files in the other languages require translation which are then staged and dynamically routed and sequenced to individual translation resources where the actual translation is performed. Management status, reporting, scheduling, and accounting information is sent to the user as the translation process ensues. The user is notified of the completion of translation and the invention coordinates the delivery of the translated documents, data streams, or non-text files back to the user's site for installation and optional review. The invention makes a variety of translation resources instantly available to the user which include both automated translation tools as well as human translators. The translation resources are connected to the invention using a flexible architecture that can be deployed on intranets as well as the Internet.

Translation management system

Upon receipt of a document request, a client assistant examines its cache for the document. If not successful, a server searches for the requested document in its cache. If the server copy is still not fresh or not found, the server seeks the document from its host. If the host cannot provide the copy, the server seeks it from a document repository. Certain documents are identified from the document repository as being fresh or stable. Information about each these identified documents is transmitted to the server which inserts entries into an index if the index does not already contain an entry for the document. If and when this particular document is requested, the document will not be present in the server, however the server will contain an entry directing the server to obtain the document from the document repository rather than the document's web host.

System and method of accessing a document efficiently through multi-tier web caching

Fraud detection is of great importance to financial institutions This paper is concerned with the problem of finding outliers in time series financial data using Peer Group Analysis (PGA), which is an unsupervised technique for fraud detection The objective of PGA is to characterize the expected pattern of behavior around the target sequence in terms of the behavior of similar objects, and then to detect any difference in evolution between the expected pattern and the target The tool has been applied to the stock market data, which has been collected from Bangladesh Stock Exchange to assess its performance in stock fraud detection We observed PGA can detect those brokers who suddenly start selling the stock in a different way to other brokers to whom they were previously similar We also applied t-statistics to find the deviations effectively

Unsupervised Outlier Detection in Time Series Data

With the worldwide growth of the Internet, research on Cross-Language Information Retrieval (CLIR) is being paid much attention. Existing CLIR approaches based on query translation require parallel corpora or comparable corpora for the disambiguation of translated query terms. However, those natural language resources are not readily available. In this paper, we propose a disambiguation method for dictionary-based query translation that is independent of the availability of such scarce language resources, while achieving adequate retrieval effectiveness by utilizing Web documents as a corpus and using co-occurrence information between terms within that corpus. In the experiments, our method achieved 97% of manual translation case in terms of the average precision.

http://www.dl.is.ritsumei.ac.jp/~amaeda/pub/IRAL00-e.pdf

Query term disambiguation for Web cross-language information retrieval using a search engine

Since a library is inherently multi-lingual, a multi-linguid document environment is crucial for a digital library. In the near future, worldwide information sharing through digital libraries will be common. Currently, multi-lingual documents are poorly facilitated on computers and the Internet. It is impractical to consider installing fonts for all character sets in every user’s terminal. This paper presents a multilingual document browsing tool for a user with no multilingual fonts on his or her terminal. It discusses several methods for browsing multi-linguaJ documents and proposes a browser which sends a text string with the font glyphs required to display the text. It also gives the evaluation result of the browser.

A browsing tool of multi-lingual documents for users without multi-lingual fonts

national network, and a multilingual browser is an essential tool for international access to and sharing of global information. The Web has expanded very rapidly worldwide. We can easily access a document from a foreign site using an off-the-shelf browser. However, that browser is usually capable of showing only documents written in English and a local language, not those written in other languages. The principal problem preventing users from browsing documents written in a foreign A new technology allows users to browse multilingual documents on the Internet. language is the lack of both a font for the language and a display function to process multiple character codes. We developed a technology called MHTML to browse multilingual documents on an off-the-shelf Web browser, and applied the technology to a multilingual gateway service to browse foreign documents and to a multilingual electronic text collection of Japanese folk tales [2, 3, 4]. MHTML technology is composed of three elements: The MHTML document object is a package containing a source text string and the minimum set of glyphs required to display the text. An HTML text specified by a URL in an applet tag to invoke the MHTML viewer is converted into an MHTML document object on the fly and sent to the viewer by an MHTML server. (A glyph of a character is a graphical entity used to display or print the character. A minimum set of glyphs for a text is a set of glyphs for all of the distinct characters that appear in the text.) The MHTML viewer, implemented as a Java applet, displays an MHTML document object on a Web browser using only the glyphs enclosed in the object. The MHTML server, with the viewer applet and a font bank, converts an HTML document into MHTML using the glyphs defined in the font bank. The four major advantages of MHTML technology begin with its simple user environment where the user installs only a Java-enabled browser. Secondly, the number of distinct characters used in a source text is much smaller than the total number of characters of the text. We found the ratio between the length of an MTHML document object and its source text was approximately 2:1 to 3:1 in the case of scientific articles written in Japanese. (The source Japanese character is encoded in two bytes. The font glyph is the bitmap data of a character, and its size …

Viewing multilingual documents on your local Web browser

The anomaly detection problem has important applications in the field of fraud detection, network robustness analysis and intrusion detection. This paper is concerned with the problem of detecting anomalies in time series data using Peer Group Analysis (PGA), which is an unsupervised technique. The objective of PGA is to characterize the expected pattern of behavior around the target sequence in terms of the behavior of similar objects and then to detect any differences in evolution between the expected pattern and the target. The experimental results demonstrate that the method is able to flag anomalous records effectively.

/pdf/anomaly-detection-using-unsupervised-profiling-method-in-121gjesx7q.pdf

Akira Maeda

Papers

Unsupervised Outlier Detection in Time Series Data

Query term disambiguation for Web cross-language information retrieval using a search engine

A browsing tool of multi-lingual documents for users without multi-lingual fonts

Viewing multilingual documents on your local Web browser

Anomaly Detection Using Unsupervised Profiling Method in Time Series Data.