 |
 |
 |
 |
Speech Analytics has a core set of deliverables governing the success of the technology related to Total Cost of Ownership (TCO): Accuracy, Scalability, Speed and Relevancy. A truly viable speech analytics solution is a mixture of the best approaches in all four of these areas. It is this combination that gives Nexidia Enterprise Speech Intelligence the competitive advantage over all other solutions.
Affordable TCO is measured in hardware processing power and speed of producing actionable information. Using Nexidia ESI, the ingest process renders your audio instantly searchable, as well as delivers timely reports based on your first level search. ESI provides an unparalleled ingest speed of 83 times real time, which means that the audio is made searchable and reported on sixty-three times faster than it is spoken. The Phonetic Search Engine (abbreviated as PSE, trademark pending) is an open-vocabulary retrieval system, which greatly reduces the time, and increases the accuracy of searches against large collections of recorded speech. Searches can be conducted at speeds up to 548,000 times faster than real-time playback of the recordings. Due to the extremely fast nature of the technology, Nexidia can render audio searchable with significantly less hardware than other technologies. For example, if you record 2,500 hours of audio a day within your organization, all of that audio can be processed using just a single dual-processor server (i.e. 2 CPUs). As you can see, the amount of hardware necessary is dependent upon how much audio you will be processing, but due to the nature of the product, the hardware requirements are feasible – leading to the viable TCO. With ESI, you can begin searching your audio immediately, even using proper names, brand names, acronyms, and slang. Compared to traditional speech-to-text solutions, ESI requires typically 1/30th the hardware processing power, therefore, delivering the only true commercially viable speech analysis solution.
Accuracy is the ability to generate actionable and accurate results. The Nexidia ESI Phonetic Search Engine is an open-systems vocabulary retrieval system, which greatly reduces the time, and increases the accuracy of searches against large collections of recorded speech. During the ingest process, audio files are marked in its smallest component, phonemes, the smallest unit of human speech. ESI ensures accurate results by providing language models trained on a wide variety of accents and dialects. Compared to traditional speech-to-text models that are heavily dependent on dictionaries, Nexidia ESI's phonetic approach provides a dramatically more accurate solution on all recorded audio.
Speed is measured in two categories – the rendering of searchable audio and the speed at which you can search the processed audio. Nexidia ESI ingest process not only renders audio searchable, but also performs the first level search and analysis of the audio at 83 times real time, a fraction of the time required by speech-to-text technologies. At time point, you are able to further investigate the initial findings by immediately drilling down and listening to the actual audio files containing relevant results. ESI provides a highly scalable architecture and database search performance that generates results at speeds up to 548,000 times faster than real time playback.
Relevancy is the ability to generate search results from audio within context; just getting data from data will not produce relevant information. Nexidia ESI utilizes advanced audio mining techniques to increase the accuracy and relevancy of search results. Relevant results are produced by understanding the context in which words and phrases are said. The search and query functionality available through ESI, not only finds words and phrases, but to ensure relevancy, finds them in proximity to other content, thus generating relevant results. The audio can immediately be played and the result can be listened to within the context of the original file. |
