Vous êtes sur la page 1sur 14

Audio and Broadcast White Paper

www.autonomy.com

Autonomy Audio and Broadcast White Paper


Table of Contents
1 Autonomy fundamentals .......................................................................................1 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 1.2 Autonomy infrastructure technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .1 1.2.1 1.2.2 1.2.3 Automated content operations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2 Automatic classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 Automatic personalization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3

2 Audio broadcast: Autonomy and multimedia ...........................................................4 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 2.2 Features and benefits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 2.3 Application of Autonomy VoiceSuite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5 2.3.1 2.3.2 Broadcast monitor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6 Enterprise Knowledge Management . . . . . . . . . . . . . . . . . . . . . . . . . . . .6

2.4 Technical components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7 2.4.1 2.4.2 2.4.3 Audio streams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 Audio aggregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 Audio files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8

2.5 Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .8 2.5.1 2.5.2 2.5.3 Large vocabulary recognition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9 Inter-speaker independence (Variation between speakers) . . . . . . . . . . . .10 Non-dictated speech . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10

2.6 Performance and scalability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10 2.7 Deployment platforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11 2.8 Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .11

Table Of Figures
Figure 1: Figure 2: VoiceSuite integration within an Autonomy system ......................................7 Time-first hypothosis extension..................................................................9

Autonomy Audio and Broadcast White Paper

1 Autonomy fundamentals

1.1 Introduction
Autonomy employs a fundamentally different and unique combination of technologies to enable computers to form an understanding of a page of text, web pages, emails, voice, documents and people. Autonomy's solution is therefore able to power any application dependent upon unstructured information within every market sector, including e-commerce, customer relationship management, knowledge management, enterprise information portals and online publishing applications. This evidence is supported by the significant penetration of the technology in a diversity of vertical markets and has been achieved principally because every market sector needs to manage and leverage the benefits of unstructured information. This combination also provides unique advantages: automatic accurate computational efficiency language independence format agnosticism Already, Autonomy has become the standard for managing and processing unstructured information across every business in every industry. Autonomy technology can be rapidly implemented either as a complete out-of-the-box solution or as an integrated component of an existing software application. For many Fortune 500 organizations, as well as OEMs partners, Autonomy software is playing a key part in their success.

1.2 Autonomy infrastructure technology


Built upon a unique combination of powerful concept-matching algorithms, Autonomy delivers the Intelligent Data Operating Layer - IDOL, an intelligent infrastructure technology, which makes it possible for organizations to automatically process digital information. Sitting above the sea of unstructured information within the enterprise, IDOL automatically identifies the subject matter of each piece of information by extracting the document's "digital essence." Many critical processes and tasks traditionally performed manually within the enterprise can now be automated by Autonomy's technology.

page one

1.2.1

Automated content operations

Whether connecting people to content, content to content or even people to people, Autonomy provides a complete modular range of IDOL functionality that enables organizations to integrate the latest in personalization, collaboration, classification, retrieval and proactive information delivery features that solve real business issues. The strength of Autonomy's technology is that it powers a wide range of operations automatically, reducing costs and adding value in real time. Because of its modular architecture, the enterprise can rapidly tailor the technology's functionality to meet their business objectives. IDOL functionality includes:

1.2.1 Automated content operations


Concept matching & retrieval IDOL offers higher degrees of accuracy and sophistication, using a scalable technology that recognizes concepts. This unique Autonomy differentiator provides powerful retrieval features, including natural language, conceptual search, refine by example, crosslanguage search and query by example. Autonomy also supports legacy retrieval mechanisms, such as keyword, Boolean, Proximity, Exact Phrase, Soundex and many others Active matching Proactively link users with relevant information they require, accurately, in context and in real-time without the user being needlessly diverted from their work in progress to perform a search or retrieval operation Automatic hyperlinking Completely removing the requirement to manually insert hyperlinks into content, Autonomy IDOL generates hyperlinks in real-time to all types of data, ensuring they are immediately up-to-date. Automatic summarization Autonomy also intelligently returns an automatic summary of the information containing the most salient concepts of the content. Summaries can be generated that relate to the context of the original inquiry - allowing the most applicable dynamic abstract to be provided for a given operation. etc.

page two

Autonomy Audio and Broadcast White Paper

1.2.2 Automatic classification


Automatic categorization The flexibility of Autonomy's categorization feature allows you to precisely derive categories using concepts found within unstructured text. This ensures that all data is classified in the correct context with the utmost accuracy. Taxonomy generation Autonomy's automatic taxonomy generation eradicates the necessity for human intervention and builds taxonomies based on the meaning of the information itself. Clustering Autonomy's automatic clustering capabilities can take large sets of document data or even user-profile information and automatically identify the main set of information clusters (themes) inherent within your information assets.

1.2.3 Automatic personalization


Automatic profiling Automatic profiling provides the organization with a real time tool to accurately understand individuals' interests based on browsing, content consumption or content contribution. Generating a multi-faceted and concurrent conceptual profile of each user based on both explicit profiles (agents) and implicit profiles (click thru and submission), automatic profiling avoids the need for explicit input of any form from the user and delivers options for identifying and managing expertise and collaboration. Community & collaboration IDOL automatically stores a concurrent, accurate and multifaceted understanding of every user based on powerful profiling operations. This ability enables IDOL to automatically match users with similar interests and drive collaboration through discovering communities of knowledge. Expertise identification Automatically gaining an understanding of every individual in the community IDOL facilitates the recognition of highly focused experts enabling users to engage in proactive collaboration ventures.

page three

Audio broadcast: Autonomy and multimedia

2 Audio broadcast: Autonomy and multimedia

2.1 Introduction Harnessing unstructured information affects virtually any business. Analysts claim that up to 80% of all business critical information is in unstructured form and in large organizations, it's doubling every three months. This proliferation of web pages, word processing documents, spreadsheets, emails, PDFs etc is a great resource there to be exploited using a technology that can make sense of it. Already, Autonomy has become the standard for managing and processing unstructured information across every business in every industry. Autonomy's technology provides an infrastructure that seamlessly brings together unstructured, semi-structured and structured information, allowing companies to leverage critical information irrespective of where it resides. SoftSound, a renowned speech recognition company that forms part of the Autonomy group, further leverages Autonomy powered infrastructures by providing the ability to handle multimedia content, such as videos, broadcasts, audio archives, news feed streams, etc, which becomes largely available in organizations, thanks to the bandwidth and storage improvements. Autonomy's technology understands multimedia content by transcribing the audio content into text, then identifies and ranks the main concepts within it, and automatically personalizes and delivers that information to those who need it any way they want - across the Internet, the extended enterprise or using other digital channels, such as mobile phones, PDAs, etc. This White Paper describes Autonomy VoiceSuite, a component of Autonomy's infrastructure solution, allowing the automated aggregation of multimedia content.

page four

Autonomy Audio and Broadcast White Paper

2.2 Features and benefits Autonomy VoiceSuite provides the following features and benefits: Automatically aggregates multimedia content wherever it is stored Automatically cross-references archived and live data, audio and human readable text, in real time Automatically creates metadata Automatically generates textual conceptual summaries of the audio content Increases accessibility to multimedia information allowing greater reuse of and access to archived material Improves the workflow around multimedia content leading to a reduction in both duplication of stored material and cost of maintaining that content Allows users to retrieve multimedia files, using natural language query. The most important implication of this technology is in providing the aggregation of a vast wealth of knowledge and information that was previously unavailable to the business due to the enormous requirements in terms of computation power, effort and storage needed to implement such a system.

2.3 Application of Autonomy VoiceSuite Autonomy VoiceSuite plugs into any Autonomy product, allowing businesses to implement intelligent, scalable and automated applications that can handle multimedia. Using VoiceSuite, businesses can extend applications such as ecommerce, customer relationship management, knowledge management, enterprise information portals and online publishing. This section describes the use of Autonomy VoiceSuite in applications such as: Broadcast monitor Enterprise Knowledge Management

page five

2.3.1

Broadcast Monitor

2.3.1 Broadcast monitor Using real-time audio/video feeds of broadcast material, users can be alerted to the development of news stories, financial events or even the latest sports results. Users receive links, sound bites or video footage of the latest events based upon their personal interests rather than what is deemed important by the broadcasters. This massively reduces the information overload to the end-user and allows them to have personalized content delivered to their WAP/i-Mode telephone or their portal site. At the backend, no manual input is required, as Autonomy automatically processes any audio/video feed. 2.3.2 Enterprise Knowledge Management An increasing number of companies are using multimedia services such as voice or video mail within the corporate networks. Businesses and end users now also have access to software and devices that enable them to create multimedia presentations and messages to add extra depth to their communications. Autonomy VoiceSuite technology provides a powerful component for integrating such multimedia data within a company's knowledge management system and extranets.

page six

Autonomy Audio and Broadcast White Paper

2.4 Technical components


The Autonomy VoiceSuite module can be used as a standalone audio-processing component or as a seamlessly integrated part of a wider Autonomy system.

Figure 1: VoiceSuite integration within an Autonomy system

Integration with other Autonomy modules is accomplished through use of Autonomy Application Builder (ACI API) and TCP/IP communication. Autonomy VoiceSuite is mainly oriented around the aggregation of multimedia content as a major component of the Autonomy automated Content Infrastructure (ACI). The technology represents a powerful broadcast aggregation system for handling multimedia data either fed into the system directly from real-time audio streams or through monitoring of a file system for the appearance of suitable audio files.
pagr seven

2.4.1

Audio Streams

2.4.1 Audio streams Real-time audio can be streamed into the Autonomy VoiceSuite module for processing. This enables broadcasts such as radio or Internet streams to form the audio-input to the Autonomy Voice 2.4.2 Audio aggregation Autonomy's Audio Connector allows for real-time audio data to be received and stored on a computer. Using hardware such as Digital Audio Broadcast (DAB) receivers the Audio Connector can record broadcast audio programs which can then form the input to the Autonomy VoiceSuite module. 2.4.3 Audio files Multimedia files such as presentations, recording of meetings can be stored on a computer system so as to be accessible by the Autonomy VoiceSuite module. They can then be processed and included within the ACI.

2.5 Technology VoiceSuite utilizes acoustic, pronunciation and language models which are combined with a patented search technique to provide fast and accurate speech recognition in a wide range of acoustic environments. The acoustic models comprise both recurrent neural networks and hidden Markov models to obtain maximum speed and accuracy. These statistical models are made of every phoneme in the context of the preceding and following phoneme (e.g. the "a" sound is slightly different in "cat" than "nan"), so that they are capturing coarticulation effects. The pronunciation and language models can be derived automatically from IDOL serverTM thus proving unique flexibility for the automatic adaptation of the system to new domains. Not only are new words automatically learned but the context in which these words is captured, and appropriate pronunciations are derived. All of the models are integrated with a proprietary pattern matching search to identify the most likely spoken phrases. Unlike conventional time synchronous (Viterbi) search the time-first search concentrates on exploring the most likely options first. Not only is this a much faster technique but it has the additional advantage of requiring less memory.

page eight

Autonomy Audio and Broadcast White Paper

In addition to the transcription and timings, the recognition module provides alternative transcriptions, word confidences, phone identification and timings. All of this information is passed to IDOL serverTM in order to provide higher accuracy solutions. The major benefits of the approach to speech recognition taken by the Autonomy VoiceSuite technology are described below:

Figure 2: Time-first hypothosis extension

2.5.1 Large vocabulary recognition Using patented predictive technology the VoiceSuite module is able to provide the benefits of a large vocabulary speech recognition system without the overhead of a vast search space when considering sample audio.

page nine

2.6

Performance and Scalability

2.5.2 Inter-s speaker independence (Variation between speakers) Recognition of speakers requires no initial training on their part. The system complements Autonomy's core technology of IDOL server by approaching any data as "found" data. "Found" data is data whose quality or construction cannot be relied upon and must therefore be regarded as unstructured at the best of times. 2.5.3 Non-d dictated speech Information feeds such as news broadcast are intrinsically difficult to recognize and transcribe due to the varied number of speakers and the fact that the broadcast was not intended for such processing and so the speakers may not always speak clearly and continuously. VoiceSuite is able to disregard these acoustic conditions and provide recognition of data in many situations. In combination with Autonomy Content Infrastructure layer, users are able to monitor and be alerted to information that best matches their interests.

2.6 Performance and scalability The VoiceSuite module can be tailored to many different applications, from small to full scale broadcast channel monitoring. A system for the monitoring of a multiple real-time audio broadcasts is possible using standard entry-level hardware with the VoiceSuite software being able to process incoming audio feeds with a maximum latency of 30 seconds. When working in RAID and extended storage provider environments, such as EMC, Autonomy leverages disk optimization and caching features to ensure the optimal performance. Communication between the VoiceSuite module and other Autonomy components within a larger system is consistent with all other Autonomy modules, using HTTP over TCP/IP allowing for a highly scalable distributed system to be constructed easily.

page ten

Autonomy Audio and Broadcast White Paper

2.7 Deployment platforms Solution Type Broadcast monitor Enterprise solution Supporting Real-time audio feeds producing 100 MB of audio data per hour (uncompressed audio sampled rate at 16 kHz). Recordings of corporate memos, voice-mail and meetings (wav files sampled at 16 kHz).

2.8 Further reading Tony Robinson and James Christie "Time-first search for large vocabulary speech recognition", International Conference on Acoustics Speech and Signal Processing, 1998 Frederick Jelinek "Statistical Methods for Speech Recognition", MIT Press, 1999. Herve Bourlard and Nelson Morgan "Connectionist Speech Recognition: A Hybrid Approach", Kluwer Academic Publishing, 1994. Steve Renals and Tony Robinson, editors "Accessing information in spoken audio", Speech Communication, Elsevier Science, Vol.32, Nos. 1-2, September 2000. Soft Sound web site http://www.softsound.com

page eleven

Headquarters
Autonomy Inc 301 Howard Street 22nd Floor San Francisco CA 94105 Tel: (415) 243 9955 Fax: (415) 243 9984 Email: info@aungate.com Autonomy Corporation Cambridge Business Park Cowley Rd Cambridge CB4 0WZ Tel: +44 (0) 1223 448000 Fax: +44 (0) 1223 448001 Email: info@aungate.com

Autonomy Germany - Hamburg Valentinskamp 24 D-20354 Hamburg Germany Tel: 49 (40) 31 112 - 308 Fax: 49 (40)31 112 - 641 Email: germany@autonomy.com Autonomy Germany - Munich Leopoldstrasse 244 D-80807 Munich Germany Tel: +49 (0) 89 244 45 2027 Fax: +49 (0) 89 244 45 5056 Email: germany@autonomy.com Autonomy Italy Largo Richini, 6 20122 Milano Italy Tel: +39 02 5821 5510 Fax: +39 02 5821 5400 Email: italy@autonomy.com Autonomy Italy Via di Vigna Murata, 40 00143 Rome Italy Tel: +39 06 5483 2028 Fax: +39 06 5483 4000 Email: italy@autonomy.com Autonomy Netherlands Teleport Towers Kingsfordweg 151 1043 GR Amsterdam Postbus 57674 1040 BN Amsterdam Nederland Tel: +31 (0) 20 491 96 80 Fax: +31 (0) 20 491 73 66 E-mail: netherlands@autonomy.com Autonomy Spain C/ Maudes 51 8a Planta 28003 Madrid Spain Tel: +34 91 3956325 Fax: +34 91 3956396 Email: spain@autonomy.com Scandinavia Autonomy Nordic AS Fridjof Nansensplass 4 P.O.Box 35 Sentrum 0101 Oslo Norway Tel: +47 23 100 727 Fax: +47 23 100 701 Email: info@autonomy.no Autonomy Sweden AB Stockholm Stureplan Stureplan 4c, 4th floor Stockholm 114 35 Sweden Phone: +46 8 545 273 70 Fax: +46 8 545 273 89 Email: sweden@autonomy.com Asia-P Pacific Autonomy Asia-P Pacific Level 14 33 Berry Street North Sydney NSW 2060 Australia Tel: 61 (2) 9959 1951 Fax: 61 (2) 9959 1035 Email: asiapac@autonomy.com Autonomy Systems Singapore 3 Temasek Ave Level 34 Centennial Tower Singapore 039190 Tel: +65 6549 7848 Fax: +65 6549 7584 Email: asiapac@autonomy.com

Regional Offices
North America Autonomy Federal Office Autonomy, Inc. 8102 Greensboro Drive Suite 601 McLean, VA 22102 Phone: 1 703 821 1600 Fax: 1 703 821 1662 Autonomy has additional offices in: Boston, MA Dallas, TX, New York, NY, Chicago, IL and Washington, DC. Continental Europe Autonomy Belgium Bessenveldstraat 25 1831 Diegem Belgium Tel: +32 (2) 716 40 05 or +32 (2) 716 40 57 Fax: +32 (2) 716 41 92 Email: belgium@autonomy.com Autonomy France 112, avenue Klber 75116 Paris France Tel: +33 (0) 1 47 55 74 02 Fax: +33 (0) 1 47 55 74 21 Email: france@autonomy.com

(Autonomy Inc. and Autonomy Systems Limited are both subsidiaries of Autonomy Corporation plc) Copyright 2003 Autonomy Systems Ltd. All rights reserved. Other trademarks are registered trademarks and the properties of their respective owners. [WP AUD/BRO] 10.03

The information contained in this document represents the current view of Autonomy Systems Ltd on the issues discussed as of the date of publication. Because Autonomy must respond to changing market conditions, it should not be interpreted to be a commitment on the part of Autonomy, and Autonomy cannot attest to the accuracy of any information presented after the date of publication. This document is for informational purposes only; Autonomy is not making warranties, express or implied, in this document.

www.autonomy.com

Vous aimerez peut-être aussi