Site: Matsushita Electric Industrial Co., Ltd.
Multimedia System Research Laboratory
Date Visited: 24 March 1998
WTEC Attendess: L. Goldberg (report author), T. Ager, B. Croft, M. Shamos,R.D. Shelton
Matsushita Electric is a conglomerate company in the consumer electric andelectronics field representing some 281 individual companies, including majornames, such as National, Panasonic, Technics, and Quasar. The company employs270,000 workers and had net sales of $62 billion in 1997. This visit wasconducted at the Multimedia Systems Research Laboratory, one of 13 specializedcompany laboratories that includes a Central Research Laboratory in Osaka. Thecompany has increased efforts in areas of telecommunications, home equipment,and industrial equipment, while seeing a decrease in the area of traditionalconsumer video equipment.
The Multimedia Systems Research Laboratory undertakes projects with a timehorizon of five years, in comparison to 10 years at the Central ResearchLaboratory. The mission of the laboratory is the development ofnetworking-based systems and the support of business creation. Three keyelements to achieving this are differentiating technology development,fundamental systems development, and creation of new business. The laboratorysits in an exceptionally modernistic building which houses business units,system integration units and an extensive display and demonstration area. Theconcept of the whole building is to become the systems business front end ofMatsushita, with a view to developing new applications and responding quicklyto needs of customers.
Mr. Oyama previewed a project to develop an English language search enginesoftware based on the company's prior work on Japanese language search engines.Researchers have developed a method of maximal word indexing that is consideredwell suited for indexing of Japanese text and affords improved accuracy inretrieval. The method utilized character-based indexing, and the size of theindex is about 2 times as big as the original text. The demonstration was thefull-text search of the U.S. patent literature. The search speed was said to bethe fastest in the industry, at an average of 700 million words/sec indexsearch for up to 1 million documents on the server. Compiling the index of thedocument set required approximately 1-1/2 hours for 1 GB of data. Similarefforts are under way employing their unique methods for the Chinese and Koreanlanguages (4 billion characters/sec in Chinese and 2.4 billion characters/secin Korean), as well as for other European languages.
Mr. Yasukawa then described another text retrieval project, involving100,000 documents, that is intended as an intelligent search finder on theWorld Wide Web. The functions include document ranking, by a similaritymeasure; automatic summary generation, with quoting of relevant parts of aretrieved document; and extracting related keywords, by selecting a set ofkeywords relevant for a query from the retrieved documents. The method usesspecial indexing techniques to compensate for finding words, as in Japanese,which have no spacing as word endings.
Mr. Kiyono described creation of a new business model for providinghigh-quality video digital information services over the Internet based uponthe use of Digital Video Disk (DVD) technology. The DVD ROM, residing with theend user, is encoded with disc identifiers, which when transmitted to the Webserver, enables automatic access to a specific Web server. The Web server inturn sends an HTML file to the user to generate up-to-date Web page informationand simultaneously provides MPEG-2 playout of the desired video data from theDVD. This service is in preparation, under the growing popularity of DVD with600 titles recorded. Panasonic has been playing an important role in DVD forum.Some 10-11 companies are involved in this area, and they do not see a problemin setting common formats for this technology. The range of applicationsenvisioned for these services is electronic publishing, electronic commerce,information service, and remote education.
Mr. Kato described development work on a visual system of retrieval forlarge video tape archives, such as those held by TV broadcast stations or evenin home video libraries. The video medium is registered and encoded by keywordsfor subsequent search. Currently, 63 (9 x 7) video clips within the medium aredigitized and displayed in a single composite screen in MPEG format.
Following the demonstrations, an extensive discussion period took placedealing with questions raised by the visiting WTEC group. The text followingeach question below is paraphrasing the hosts' replies.
[Mr. Yasukawa] The company is just starting such a project and has 6 or 7researchers in the natural language area. As to speech recognition, it isworking in Japanese, English and Chinese. It is planning cooperation with othercompanies and Japanese universities in this area.
[Mr. Nishikawa] Matsushita is interested, but since it is a businesscompany, it is hard for it to invest in such a large project. Matsushita isinterested in personal home video and document access.
Only for text search but none for visual data.
This lab doesn't deal with music, but it will be a very interestingarea.
The government supports promotion of multimedia in municipal systems, butcustomers aren't ready. This laboratory has little government support at thistime, but the government has asked for proposals to be submitted. There are alot of such opportunities.
We don't have any projects here, but other Matsushita labs are working withbroadcast studios.
This is very important for our consumer activities. Only commercialdivisions are pursuing it from the practical approach, e.g., designing controlbuttons on remote controls. Our laboratory focuses on the needs ofconsumers-hence it is more short term. For example, the National Museum ofEthnology's need for MPEG images drives that work.
[The following two questions elicited a thoughtful response by the seniorengineer, Mr. Tsumura, who had the longest experience in the company.]
The company is strong in the consumer electric field. It has been strong inthat field and we would like to remain a leader. The future is for a homeinformation infrastructure-with small personal computers in the home. We'd liketo build this up.
There will be a variety of types of information from the society to thehome. We need to have something for the homes; neither PCs nor TVs as currentlyused will be sufficient. We have been a company that provides hardware, but inthe future, we need content, technology to manage content, and hardware to playit. We are not so capable in the first two areas of content or management ofcontent.
[Mr. Nishikawa] In wireless phones, where the number of subscribers reached20 millions in only five years in Japan. Matsushita is looking at the wirelessinfrastructure and working on access systems in the home. The company isworking jointly with NTT.
The definition of basic research is becoming ambiguous. Multimedia is anapplied science that is important. Some work done at the Central ResearchLaboratory in speech recognition has been going on for the past 20 years.
[An issue of common interest was the level of cooperation betweenuniversities and industry in each country. This led to the followingquestion:]
[Mr. Nishikawa] Employees are educated at the Bachelor's and Master's level,and then trained by the company. Some companies said that they did not alwayshire professional people. They don't want workers trained in any way, becausethe companies want to go in a certain direction. They wanted collectivism, notindividualism. In this point, general education is more welcome. But things arechanging. To create new technology, we need specialists. So the role ofuniversities becomes critical. Generally speaking, their projects have beensmall, but some universities collaborate and form large projects. This trendhelps to accelerate new technology development.
[The young people sitting at the table were asked what their experiencelevel was before joining the company. Most were at the Master's level incomputer science related fields, and some had had subsequent experience inbasic research, such as at the Electrotechnical Laboratory in Tsukuba.]
DVD-Internet linkage system. Briefing material.
High-speed English language full text search/registrationprogram. Briefing material.
Intelligent WWW finder. Briefing material.
Matsushita Electric. (General brochure.)
Noguchi, N., Y. Kanno, M. Inaba, K. Kurachi. 1998. Newindices for Japanese text: A new work-based index of non-segmented text forfast full-text-search. Systems Trans. of Information Processing Soc.Japan, 39, 1.
Panasonic AV&CC Systems Square. (Multimediabrochure.)
Saato, Mitsuhiro. n.d. Text retrieval system based onMaximal-Word indexing method. Briefing material.
Video archive & browse system. Briefing material.