NAACL 2001

Exhibits Day:
June 6, 2001

The Industrial Exhibits Session will take place on June 6 in the Rangos Ballrooms, on the second floor.
In addition to the exhibit booths, which will be open all day, there will be a number of scheduled presentations.
We present here the list of booths, the schedule of presentations, and then the abstracts of the presentations.

Exhibit Booths

Scheduled Presentations

Abstracts of these presentations appear below this schedule.

Session 1: Multilingual and Multimedia Language Technologies
9:00 - 9:30: Chinese Text-Based Multimedia Content Search and Retrieval, Joe Zhou, Intel
9:30 - 10:00 Integrated Foreign Language Processing Tools for Monolingual Users, Mudar Yaghi, Apptek

Session 2: Knowledge Discovery, Information Search and Retrieval
10:45 - 11:15 Linguistic Approach to Information Management, Margalit Zabludowski, Manager, U.S. Linguistic Support and Development, LexiQuest
11:15 - 11:45 Uberdex, Tom Bougan, Applied Technical Systems
12:00 - 2:00 LUNCH
(Session 2 - continued)
2:00 - 2:30 Harnessing the Power of Ordinary Language in Search Technologies, Dr. Antonio Sanfilippo, Antonio Sanfilippo, Director of Application Development, David Ellworthy, Principal Search Architect, LingoMotors

Session 3: Machine Translation
2:30 - 3:00 The Coming Boom in Real-Time Translation Services, Robert Levin, Transclick
3:00 - 3:30 A New Generation CAT Tool, Dan Gervais, Executive Director, Gerry Gervais, Multicorpora R&D, Inc.
3:30 - 4:00 Translation Memory Technology: Application, Benefits and Limitations, Christina Spies, Director of Sales, U.S. East, Trados Corporation

Session 4: Question Answering Systems
4:30 - 5:00 Online Answering Engine with Sentience, Professor P.V.S. Rao, Adviser, Tata Infotech

Presentation Abstracts

The abstracts are listed alphabetically by company, with the time of the presentation included after the name.

Applied Technical Systems (11:15-11:45): ATS introduces Sinai, a knowledge discovery engine that offers easy access to large and complex datasets. Sinai is particularly effective with free text because it has the power to search for concepts rather than key words.

Sinai takes advantage of a unique data management model to analyze the free text and identify semantic similarities between words. A system that can understand the semantic similarity between "contract" and "agreement", "solar system" and "galaxy", is much more capable of helping people locate the right kind of information.

Sinai is designed to accommodate collections of documents relating to a specific field (law, horticulture, drug research, etc.). Each Sinai system is "trained" in the specific language of that field. In fact, Sinai works equally well with any language using an ASCII character set.

Sinai includes these concept search features:

 

Apptek (9:30 - 10:00): Foreign language information acquisition is vital as more is available on the Internet and in electronic form each year. English is no longer the prevalent language of the Internet. Yet Machine Translation alone is not enough to address this requirement. Considerable volume of translated text is difficult to exploit without linguistically intelligent tools. These tools augment MT. Translation memory, historically an independent track of software development, is converging with rule based MT systems to introduce a statistical and machine learning dimension to MT. Data extraction, name finding, name/event searches, thematic searches, morphologically enabled searches, and query translation are utilities in a suite of foreign language information acquisition environment. The introduction of a language suite for monolingual users would require an integration of distinct Natural Language Processing modules that share the same linguistic resources.
 

Intel (9:00 - 9:30): We have built a text-based video indexing and retrieval system. The system can search a video file for scenes that match the user's query. The video database consists of more than a hundred-hour general video news. Optimized for Intel architecture platform, the search response time is less than one second.

Through its easy-to-use interface, the user keys in a query in natural language to conduct a search. Though the current system supports Mandarin Chinese only, other natural languages are possible with certain modifications. Just like any other Internet search engine, the system goes through an offline indexing process, as well as an online search process. Its indexer and search engine are however NLP-driven.

In the offline mode, the indexer first identifies all the Chinese words and named entities in the video scripts, i.e., texts transcribed from the audio track. Then, it builds a comprehensive indexing table to represent the position and weight of each word or each entity name. The semantic relation between each pair of named entities is also identified.

In the online mode, the search engine analyzes the user's query expressed in natural language and performs a semantic match between the query and the videos scripts in the database, then outputs the most relevant video scenes.

Using our system, the user is liberated from the laborious manual rewinding and monitoring when trying to locate a video segment from a large raw video file. The text-based video search engine serves as a content auditor. It examines the video content quickly and retrieves the most relevant stories based on the user's request.
 

Lexiquest (10:45 - 11:15): This talk will relay how Lexiquest evolved from a linguistic R&D and project-oriented organization leading the way for NLP applications in France to become an international provider of linguistic solutions for the management and access of information. The presentation will chronicle the evolution of LexiQuest's product development and highlight customer applications to illustrate the role of linguistics in business. LexiQuest software allows users to communicate with technology on their own terms by enabling computers to understand human language. Over 100 customers in markets including finance, telecommunications and Information Technology, employ LexiQuest solutions. Customer applications include e-commerce; customer relationship management; search engine retrieval; corporate Intranets and human resources. Among LexiQuest's customers are France Telecom, Chase Manhattan Bank, HotBot, BNP Paribas, GE and PBS.

Founded in Paris under the name Erli SA, LexiQuest is the fruition of more than two decades of landmark research and development in computational linguistics and natural language processing. LexiQuest's groundbreaking core technology was developed in cooperation with the European Economic Commission (EEC). Steering a consortium of 11 companies from member countries, LexiQuest defined the "GENELEX Standard," which evolved into the company's linguistic software tools and applications of today.
 

LingoMotors (2:00 - 2:30): LingoMotors has created a natural language core technology that reads and understands text. This technology dramatically improves access to any data, providing relevant information and knowledge to the user.

LingoMotors' premier product, TurboSearch, revolutionizes the search experience. TurboSearch bolts on to search engines allowing questions to be asked in Ordinary Language(tm). This empowers the user to pose search questions in a "type the way you talk" format.

Search engines powered with TurboSearch process all forms of text, both structured and unstructured, to recognize the specific meaning of the search question. This allows the user to search for products, ideas, and concepts without the use or knowledge of specialized search algorithms. By understanding sentence concepts, this easy-to-use, next-generation technology provides superior Internet and Intranet search results with First-Page Relevance, eliminating pages of irrelevant data.
 

Multicorpora R&D, Inc (3:00 - 3:30): We will explain the differences between conventional Translation Memory Systems and a new generation of CAT tools, MultiTrans. This new generation is corpus-based, instead of Translation Memory based, which means that you don't have to align all sentences of legacy documents before using them. A quick demonstration of the software will be given.
 

Tata Infotech (4:30 - 5:00): OASIS is an interactive question - answering system, which answers questions in common English. Developed as a result of on-going research at the Tata Infotech's, Cognitive Systems Research Lab, OASIS is specifically oriented towards answering questions and providing information on a given topic, and is the first product of this type to be developed in India. The engine has been set-up to answer questions about Tata Infotech.

The system adopts a user-friendly conversational style. Each answer is preceded by a conversational 'prelude' which confirms to the user that the question has been properly understood. It also suggests other topics on which the user can ask questions. In its present form, OASIS will not deal with involved questions of certain types and extremely complicated sentences.

A variant of this has been implemented which answers questions relating to textual information as well as information from standard databases. The question in natural language is in this case converted into a query in SQL and the information obtained from the database is presented to the user, embedded in an English sentence.

OASIS is proposed to be made available both as a product and as a solution.
 

Trados Corporation (3:30 - 4:00): In today's global marketplace, successful companies communicate in the world's languages. TRADOS Solutions can assist you to breakdown the language barrier and manage your multilingual content.

TRADOS, founded in 1984, is the global leader in providing translation technology solutions that power multilingual business. With over 40,000 licenses representing the majority of the current translation tools market, TRADOS Solutions remain the leading choice among professional translators, enterprise solution providers, and global leaders such as Compaq, Oracle, Nortel Networks, Siemens, Volkswagen/Audi and Wal-Mart.

TRADOS Solutions stores, matches and recycles previously translated works in virtually any language or file format and thereby, avoids duplicative and costly translation processes.
 

Transclick (2:30 - 3:00): The leading real-time translation Application Service Provider, Transclick, will present both a stand-along Laptop version as well as a mobile Blackberry PDA version of its real-time machine translation system. The system allows for customizing dictionaries for specific domains to increase translation accuracy. Currently, 6 European languages paired with English translate text and HTML bi-directionally.
 


Exhibits Chair:
Lynn Carlson, lmcarls@super.org