LM95 Workshop Information

From July 17 - August 25, 1995 the Center will be hosting an invited research workshop on Language Modeling for speech recognition. LM95 is a third workshop in a series sponsored by the Department of Defense, the previous two having taken place at Rutgers University in the summers of 1993 and 1994.

Workshop lectures/seminars that are open to the public are indicated below in the sections LM95 Guest Participants and Guest Speakers at LM95.

This year there will be a total of 24 "full-time" participants all of whom are listed below.

Language Modeling Workshop 1995 CLSP/JHU
Workshop Chair: Frederick Jelinek CLSP/JHU

Senior Participants

Additional Participants:

Government Participants

Student Assistants

LM95 Guest Participants

In addition to the full participants, there will be eight guest participants who will visit for short periods from the fields of Linguistics and Information Theory. When it is decided what public lectures/seminars these guest participants will give, this information will be posted.

The week of July 31 - August 4, 1995 will be devoted to Linguistics, when four visiting scholars in the area will be present. These individuals are: A. Kroch (University of Pennsylvania), M. MacDonald (University of Southern California), M. Seidenberg (University of Southern California), and J. Groenendijk (University of Amsterdam).

The week of August 7-10 will be devoted to Information Theory, when four visiting scholars in the area will be present. These individuals are: T. Cover (Stanford University), I. Csizsar (Budapest, Hungary), P. Narayan (University of Maryland), J. Ziv (Technion-Israel).

Guest Speakers at LM95

LM95 will invite guest speakers twice a week. The guest lectures are planned for every Tuesday and Thursday at 12:00 NOON in Maryland Hall 110. All lectures given by these visitors are open to the public. When the Tuesday/Thursday lecture series is finalized a complete schedule will be posted.

Workshop Location

The workshop site is at Barton Hall on the Hopkins Homewood Campus. The laboratory is Barton 123.

Parking

General visitor parking is available at the Visitor Parking lot (fee charged) or at any of the meters on campus (quarters only).

Communications

The number where messages can be left for participants is (410) 516-8584 or (410) 516-4237. The fax number is (410) 516-5050. The address where mail may be sent is:

LM95
3400 N. Charles Street/Barton Hall
Johns Hopkins University
Baltimore MD 21218-2686

Smoking Policy. Johns Hopkins University is a smoke-free workplace. There is no smoking in the University's laboratories, offices, classrooms and cafeterias.

LM95 Details

Three teams will work on conversational English and one on Spanish. Success will be measured by speech recognition error rate achieved in English on the Switchboard corpus (the corpus for Spanish is not yet determined). It is intended that LM95 be provided with recognizer output lattices so that no complete recognition experiments will need to be run during the workshop.

As stated above, the participants will be formed into 4 teams, each with a distinct project goal. The goals, project leaders and teams are as follows:

Succinctly put, the LM95 Workshop will have four goals:

  1. To use language modeling to get better recognition results on conversational English speech over the telephone.
  2. Start work on Spanish with the same eventual goals as above.
  3. Develop a methodology for describing the content of conversations and for measuring the accuracy of the description.
  4. Reach out to linguists and information theorists to engage them in work of interest to the Human Language Technology (HLT) community.


Directions

Directions to the Johns Hopkins University Homewood Campus, Charles and 34th Streets, Baltimore, MD 21218-2686. Center for Language and Speech Processing (410) 516-4237

By CAR from the SOUTH (Ft. Meade, MD; Washington, DC) USING the BALTIMORE-WASHINGTON PARKWAY


By CAR from the SOUTH (Washington, DC) USING I-95:


FINAL APPROACH --ENTRY into CAMPUS from WYMAN PARK DRIVE

Updated 7 July 1995 - Mark L. Chang