A -- EFFECTIVE, AFFORDABLE, REUSABLE SPEECH-TO-TEXT (EARS)

Notice Date

November 7, 2001

Contracting Office

Other Defense Agencies, Defense Advanced Research Projects Agency, Contracts Management Office, 3701 North Fairfax Drive, Arlington, VA, 22203-1714

ZIP Code

22203-1714

Solicitation Number

BAA02-06

Response Due

November 8, 2002

Point of Contact

Charles Wayne, Program Manager, DARPA Information Technology Office (ITO), Phone none, Fax 703-522-7161, Email none

Description

The Defense Advanced Research Projects Agency (DARPA) seeks strong, responsive proposals from well-qualified sources for a new speech recognition research program known as EARS (Effective, Affordable, Reusable Speech-to-Text). GOAL: The goal of the EARS program is to produce powerful new speech-to-text (automatic transcription) technology whose outputs are substantially richer and much more accurate than currently possible. The program will focus on natural, unconstrained human-human speech from broadcasts and telephone conversations in a number of languages. The intent is to create core enabling technology suitable for a wide range of advanced applications, not to develop those applications. Inputs and outputs will be in the same language. OBJECTIVES: Specific objectives of the EARS program include: Word error rates in the 5-10% range (reached within 36 months for broadcasts and within 60 months for conversations). Useful metadata of various kinds extracted from the signal. Readable transcripts (with standard punctuation, formatting, etc). Domain, topic, speaker, style, and vocabulary independent technology. Robust with respect to variations in channel and speaking environment. Multiple source languages (including varieties of English, Chinese, and Arabic). Portable to an arbitrary language with one week of concerted effort. Executable in real time (or faster) with standard hardware (one processor). Commercialization of successful technology. SCOPE: To achieve those objectives, the EARS program will encompass: Wide-ranging, multidisciplinary research. Quantitative comparative evaluations of algorithm accuracy. Quantitative evaluations of algorithm utility (for people and machines). Technology demonstration prototypes. Collaboration among program participants. DURATION: The EARS program is expected to last five years. Vigorous algorithmic research will occur throughout. During the first 36 months, a wide range of ideas will be tried and tested and initial technology demonstration prototypes will be developed. During the final 24 months, the most promising research avenues will be extended, additional prototypes will be produced, and run-time efficiency and portability to new languages will be given increased attention. PROPOSALS: This BAA seeks four types of proposals: Rich Transcription. Novel Approaches. Interfaces. Linguistic Data. Offerors wishing to do more than one type of work must submit separate proposals. TASKS: Within a proposal, offerors must divide the proposed work into distinct, coherent, separately priced tasks. In assembling a balanced, high quality program, the government may choose to fund some tasks and not others. To enable sensible decisions, offerors should try to define proposed tasks so that the annual budget for an individual task is in the $100,000 to $1,000,000 range. EMPHASIS: Because of the very difficult technical challenges to be overcome, research to develop effective algorithms will constitute the vast majority of the EARS program. The bulk of that research will be on general-purpose speech-to-text algorithms that can be adapted to different media and different languages. DARPA plans to invest both in substantial improvements to state-of-the-art approaches and in radically new approaches. DARPA will also invest, but less heavily, in research to extract useful metadata and in the creation of prototype speech-to-text systems and user interfaces. SOURCE DATA: The source data will be natural human-human speech in any of several languages from broadcasts and telephone conversations with various environments and channels. CAPABILITIES: Offerors are expected to possess substantial experience with the technical problems being addressed plus state-of-the-art technology upon which to build. Prospective research teams should have demonstrated competence in building speech-to-text technology for human-human speech. This can best be proved via participation in a NIST-sponsored benchmark test (Hub 4 or Hub 5). Individuals or organizations that could make a valuable contribution to EARS research but who have not run and cannot run a test in time are encouraged to join a team with a track record. In view of the multidisciplinary character of the teams and the research, outstanding individuals from other fields are encouraged to participate in this way. SCHEDULE: The initial round of awards will be based on proposals received by the 14 January 2002 deadline specified below. If all goes well, offerors will learn the status of their proposals at the end of February 2002 and receive contracts in April 2002. The EARS kickoff meeting is scheduled for 9-10 May 2002 in the Baltimore-Washington area. GENERAL INFORMATION: This solicitation is not related to any specific application, system, or hardware procurement. Rather, it is for efforts that will significantly advance the state of the art. This notice published in the Governmentwide point of entry and the Commerce Business Daily, in conjunction with the Proposer Information Pamphlet (PIP) and all references, constitutes the total BAA. No additional information is available, nor will a formal Request for Proposal (RFP) or other solicitation regarding this announcement be issued. Requests for same will be disregarded. Proposers must obtain the BAA 02-06 Proposer Information Pamphlet (PIP), which provides further information on the areas of interest, submission, evaluation, funding processes, and full proposal formats. This pamphlet may be obtained by fax, electronic mail, or mail request to the administrative contact address given below, or at URL address http://www.darpa.mil/ito/Solicitations.html. Proposals not meeting the format described in this pamphlet may not be reviewed. All responsible sources capable of satisfying the Government's needs may submit a proposal that shall be considered by DARPA. Historically Black Colleges and Universities (HBCUs) and Minority Institutions (MIs) are encouraged to submit proposals and join others in submitting proposals. However, no portion of this BAA will be set aside for HBCU and MI participation due to the impracticality of reserving discrete or severable areas of this research for exclusive competition among these entities. Restrictive notices notwithstanding: Proposals may be handled, for administrative purposes only, by a support contractor. This support contractor is prohibited from competition in DARPA technical research and is bound by appropriate non-disclosure requirements. PROPOSAL SUBMISSION PROCESS: The Defense Advanced Research Projects Agency/Information Technology Office (DARPA/ITO) requires completion of a Broad Agency Announcement (BAA) Cover Sheet Submission for each Proposal, by accessing the URL below: http://www.dyncorp-is.com/BAA/indexcw.asp?BAAid=3D02-06. After doing so, the proposer must print the BAA Confirmation Sheet that will automatically appear on the web page and submit it attached to the front of each copy of the proposal. Failure to comply with these submission procedures may result in the submission not being evaluated. The full proposal (original and designated number of hard and electronic copies) must be submitted to the administrative address for this BAA in time to reach DARPA by 12:00 PM Noon (ET), Monday, 14 January 2002, in order to be considered during the initial evaluation phase. Proposals must not be submitted by fax or e-mail; any so sent will be disregarded. The BAA will remain open until 12:00 Noon (ET), Friday, 8 November 2002. Proposals submitted after the 14 January deadline will be evaluated by the Government, but the likelihood of their being funded is less than for proposals submitted in accordance with the initial evaluation and award schedule. EVALUATION AND FUNDING PROCESSES: Each proposal will be evaluated through a scientific review against the following criteria, which are listed in descending order of relative importance: (1) Overall Scientific and Technical Merit; (2) Innovative Technical Solution to the Problem; (3) Potential Contribution and Relevance to DARPA Mission; (4) Offeror's Capabilities and Related Experience; (5) Plans and Capability to Accomplish Technology Transition; and (6) Cost Realism. Proposals will not be evaluated against each other, since they are not submitted in accordance with a common work statement. When the proposal evaluation is completed, the proposer will be notified of selectability or non-selectability. Selectable proposals will be considered for funding; non-selectable proposals will be destroyed (but copies may be retained for filing purposes). Not all proposals deemed selectable will be funded. Decisions to fund selectable proposals will be based on the funds available and on how the proposal taken together with others meets DARPA's needs. DARPA may retain selectable proposals for a period of up to one year, in order to reconsider those proposals for funding. Submitters of those retained proposals will receive notification to that effect. The Government reserves the right to select for award all, some, or none of the proposals received. Proposals identified for funding may result in a contract, grant, cooperative agreement, or other transaction depending upon the nature of the work proposed, the required degree of interaction between parties, and other factors. If warranted, portions of resulting awards may be segregated into pre-priced options. All administrative correspondence and questions on this solicitation, including requests for information on how to submit a proposal to this BAA, must be received at one of the administrative addresses below by 4:00 PM (ET) Friday, November 1, 2002; e-mail is preferred. DARPA intends to use electronic mail and fax for some of the correspondence regarding BAA 02-06. Proposals MUST NOT be submitted by fax or e-mail; any so sent will be disregarded. ADDRESSES: The administrative addresses for this BAA are: Electronic Mail: BAA02-06@darpa.mil. Electronic File Retrieval: http://www.darpa.mil/ito/Solicitations.html. Fax: 703-522-7161 addressed to: DARPA/ITO, BAA 02-06. Mail to: DARPA/ITO, ATTN: BAA 02-06, 3701 N. Fairfax Drive, Arlington, VA 22203-1714.

Web Link

Visit this URL for the latest information about this (http://www.eps.gov/spg/ODA/DARPA/CMO/BAA02-06/listing.html)

Record

Loren Data Corp. 20011109/ASOL007.HTM (D-311 SN5125F0)