Seek within audio and video content.

Technology


Our Approach

The core of our approach to search is based on automated speech recognition (ASR). Speech recognition enables us to generate an index file for each audio or video file that specifies where each word occurs in time within the file. We recognize that state-of-the-art ASR can not provide a complete solution - large vocabulary speech recogntion (the class of speech rec applicable to our case) simply does not have a high enough accuracy level. Given this challenge, we take several steps to improve the quality of our audio/video search service as described below.

  • Improved Speech Recognition
    We leverage any available text associated with a given audio/video file to simplify the index generation process. This not only improves our accuracy but supports useful text post-processing of our data.

    As with all ASR engines, training is necessary so that the engine can "learn" an accurate model of speech. We place special emphasis on the quality and type of data used in our training process to ensure that the accuracy of our engine continues to improve.

  • Text Processing
    In addition to ASR, we use text processing techniques to ensure a highly accurate index file and facilitate a rich array of searches. In addition to a variety of features such as Boolean queries, classification and stemming, our text processing techniques also enable advanced features such as speaker-based search.

  • Software as a Service
    We use a Software as a Service (SaaS) model to deliver our solutions. Given the complexity of speech recognition, SaaS alleviates the challenges of deployment and maintenance and enables a versatile server model in which we can optionally share the responsibilities of data hosting when desired by our customers. In addition to reducing total cost of ownership, SaaS facilitates continual updates as we continually improves our services. Most importantly, SaaS enables our customers to remain focused on their core competencies while we provide unparalleled audio/video search capabilities.

Please contact us with any questions about our technology or to learn how LingFling can provide a solution for your company.