SAIL LABS Technology, one of the top providers of speech technology solutions, is proud to announce the availability of Media Mining Indexer (MMI) 6.0, the new major release of its flagship product for real-time, multi-lingual, multi-media indexing.
The Media Mining Indexer is built on SAIL LABS’ award winning speech recognition system, provides a suite of best-of-breed technologies such as large vocabulary automatic speech recognition, speaker identification, detection of named entities and topics, and transforms audio-/and video data into indexed and searchable information in real-time.
“Developed in close co-operation with users and a worldwide network of partners, we have refined our flagship product, the MMI, to enhance customisation, user flexibility, stability, performance and standards compliance. Thus, this new release represents a quantum leap towards customised media intelligence and advances the state of the art in speech technology,” said Gerhard Backfried, Head of Research at SAIL LABS.
Standard Features of the Media Mining Indexer include:
- Automatic Speech Recognition enabling accurate conversion of speech to text, even inthe presence of high levels of background noise
- Speaker Change Detection segregating audio segments based on speaker and audio environment for focused information retrieval
- Speaker Identification attributing statements to certain speakers (“who” has been saying “what” in the news?)
- Topic Classification helping to locate stories relating to user-specific interests
- Named Entity Detection tagging words within the text output of automatic speech recognition with categories of entities such as persons, locations, organizations.
Some of the exciting new features include:
- Online acoustic adaptation
- Support of multiple language models
- SOAP Web-Service interface
- Extended Named Entity Detection (NED):
– Flexible named entities (user editable named entity tables)
– Flexible named entity types (base-types as provided by Sail Labs may be extended by end-users to meet their particular needs)
– Multiple named entity files (allowing sharing of named entities between Media Mining Indexer and Media Mining Server)
- Enhanced modular design through extended licensing mechanism (e.g. Speaker Identification (SID) can be turned on/off; speech recognition can be turned on/off allowing for text technologies to be active only)
- Improved Gender Identification (GID)
- Speed improvements in Automatic Speech Recognition (ASR)
- Extended API and enhanced compatibility:
– Speaker Identification (SID): SID models can be loaded (set) and reset via the API now. This allows for flexible setups, e.g. SID per model, per channel, per program, etc.
– Keywords can be set and queried via the API
– API support for C++ and plain C
– Compatibility with current operating systems (Windows 7, Windows 2008R2, etc.)
Currently Arabic, American & British English, French, German, Greek, Hebrew, Mandarin Chinese, Norwegian, Polish, Russian and Spanish are supported. Italian and Farsi will be available soon.
Media Mining Indexer 6.0 is available immediately from SAIL LABS and its global network of partners to new customers and to registered customers with support & maintenance contracts including updates.
Share on Social Media