Texts of better quality faster than before
Whether it is a matter of finding or creating information, it is worth transferring routine tasks from people to machines. Costs are reduced, time is saved and work becomes more meaningful, while the number of human errors is reduced. Speech recognition increases productivity and efficiency. It reduces the time spent on recording entries, leaving more time for actual expert work and customer service. At the same time, recorded information is available more quickly to other experts and customers, which makes the service smoother and improves customer satisfaction.
"Launching the speech recognition service provided by Lingsoft was easy, it was effortless to use and recognition results were almost error-free."
"By far the best speech recognition service I have ever tried."
"I think I can finish my tasks a lot faster than by typing."
“It recognizes even very difficult words and
inflected forms, such as the Finnish word
"My own experience is that speech recognition works brilliantly and it has been a great help. Hopefully (our use) will continue even after the trial period."
Why choose Lingsoft?
Accurate recognition results
Structural analysis of human language is at work behind Lingsoft's service and ensures that the system is genuinely familiar with the intricacies of the Finnish language. This core element for solutions developed by Lingsoft itself, combined with high-quality and versatile language models, ensures the most accurate recognition results available at today’s market. Furthermore, we are able to customise language models to recognise special terminology of highly specialised professions.
Our solution is speaker-independent and does not need to be separately trained. The service is a true turnkey solution.
Ease of use
Rapid deployment is supported by the intuitive user interface of Lingsoft Speech Service. Also worth noting is that the user can use voice commands to control dictation during dictation. A phraseological function is also available to enable the dictating person to create and use repetitive text fragments that he or she has defined.
The browser-based service is available 24 hours a day. This is a zero-installation solution which works everywhere and every hour of the day. In addition, Lingsoft's voice recognition solution can be integrated with other systems through application protocol interfaces (APIs).
Quality assurance tools
Real-time speech recognition is accompanied by a text quality-check feature which helps ensure high quality results effortlessly. The dictating person is also assisted by other tools which facilitate text editing.
Extensive range of solutions
The customer can utilise speech recognition in real time, so that the dictating person can check and use the text immediately. The Speech Recognition Service also recognises separate, existing audio files. Recognition can also be used as background support for the dictation process to save word processors’ time and facilitate their work.
Lingsoft has long-standing experience with customers handling sensitive data, and we are committed to strictly complying with all data protection regulations. In all our operations, we employ technical and organisational safeguards to ensure that all personal information is processed safely. For us, it is of utmost importance that our customers can trust us.
We develop our voice recognition solution further continuously and rapidly. All the latest features become available to our customers via automatic updates.
Speech recognition can be utilised wherever you want to convert speech into text or use speech for controlling devices or software. Our solutions are being utilised by health care sector, public administration and by the media.
Read about voice recognition solutions provided by us.
The Minutes Office of the Parliament draws up the minutes of Parliament's plenary sessions. There is a lot of work, the required level of precision is high and the quality and reliability of the end result must be ensured in every situation. Lingsoft's speech recognition service, together with its quality-check feature, helps in this work.
Lingsoft developed a solution for SVT for subtitling live weather broadcasts with the help of voice recognition. The project was part of the accessibility project of the Swedish post and telecommunications agency and it yielded significant linguistic capital for the Swedish national repository for language resources.
MeMAD is an EU research and innovation initiative where technologies under development can be used to convert videos into text and metadata on an industrial scale. In this project, Lingsoft focuses especially on speech recognition and keyword selection.
Lingsoft provided its speech recognition service as part of the forensic medicine dictation process used by the Finnish institute for health and welfare (THL). Thanks to speech recognition, document processing times have been reduced without compromising the quality of documentation.
What is speech recognition quality made of?
Technologies developed for English are often not the best starting point for Finnish speech recognition. The reason for this is obvious. If all inflected forms of nouns, adjectives and verbs were actually used, the Finnish language corpus would have the same amount of entries as the whole universe is expected to have stars. It would be utterly impossible to gather such a vast, comprehensive material base and teach it to machines. The sheer amount of work would be daunting.
At the core of Lingsoft's language technology solutions lies language structure analysis which, to put it simply, means teaching the rules of inflecting and forming of a word to a machine. The word can then be identified in all its inflected forms in free text of the background material. In other words, teaching new words to the speech recognition system does not require all forms of inflection of a single word.
It is also relevant what kind of linguistic material the dictating person is using. In professionally specialised language, vocabulary is more limited than in general language, so the accuracy of recognition becomes better as the specificity of language increases. In addition to general language settings, Lingsoft also offers customised language models for recognising specialised terminology of professionally specific language settings, such as health care sector. A solution that genuinely understands the Finnish language, together with a correctly selected language model, ensures a speech recognition result where the dictating person or word processor is left with little, if at all, content to be further stylised.
Speech recognition processes the way you need it
There are different ways to produce text using speech. Different recognition methods do not exclude each other, and the same solution can support multiple use cases.
Foreground recognition – speech to text in real time
Foreground recognition means converting speech into text almost in real time. In this case, the speech recognition result is displayed to the dictating person, so that they can verify the text and edit it as needed instantly. Editing a high-quality speech recognition result is quick and the desired information is immediately captured.
Background recognition - speech recognition as an integrated part of production process
As its name implies, background recognition takes place in the background of the dictation process. Background recognition often improves the efficiency of centralised word processing either within the organisation or when outsourced. Lingsoft utilises background recognition as part of its dictation decoding, transcription and programme subtitling services.
Command recognition - controlling user interface with speech
Command recognition can be used to control the functions of a device or software. The functions are predefined and the speech recognition model is optimised to recognise words or phrases used as commands. In addition, the commands are used as part of solutions based on background and foreground recognition, such as dictation decoding for healthcare sector or programme subtitling solution. In addition to continuous speech recognition, commands can be used, for example, to insert a line-break, delete words, or introduce a predefined phrase.