Forensic medicine professionals spend a considerable portion of their daily routines entering data. Lingsoft provided its speech recognition service as part of the forensic medicine dictation process used by the National Institute for Health and Welfare. Thanks to speech recognition, document processing times
have been reduced without compromising the quality of documentation.
The National Institute for Health and Welfare (THL) is an independent expert agency and research institute, which produces data, tools and solutions to support decision-making in the health care and welfare sector. Its Forensic Medicine Unit is responsible for performing forensic autopsies and it guides and supervises cause-of-death investigations in Finland.
THL’s forensic medical examiners perform approximately 8,000 autopsies a year, each of which involve the drafting of numerous different documents. The documentation process can take several weeks, from when the forensic medical examiner has dictated the results of an autopsy to when a text processor has reviewed and approved the dictation transcription. In the autumn of 2019, the THL Forensic Medicine Unit began using Lingsoft’s speech recognition service. The goal is to replace the use of conventional dictation with a more streamlined process, reduce the probability of error, and shorten document through-put times.
Efficient dictation transcription - without compromising quality
Thanks to the Lingsoft Speech Service used by THL, dictation by the forensic medical examiner is recognised in real time. Painstaking, time-consuming transcription and text reviewing are no longer necessary - cases can be entered into the system considerably more quickly. The speech recognition service supports specialised pathology and medical terminology as well as general Finnish and Swedish vocabulary. This ensures high quality results for different types of documents and makes text production fast and seamless. Lingsoft’s own language technologies ensure that the system properly recognises the Finnish language. For example, the conjugation of words, contraction of numbers and correct abbreviations for units of measurement combined with the precise speech recognition result significantly reduce the need for any further revision of the text. Based on the most positive user experiences, “sometimes, there’s no need to make any corrections at all”.
“Why wasn’t this service available before?”
Feedback received from forensic medical examiners using the service has been positive, and the number of dictations taken with speech recognition is constantly increasing. Right from the start, THL has been using Lingsoft’s cloud service, which is accessed through a browser user interface. Adoption of the service has therefore not required the client to launch any major production transition projects, perform server maintenance or make user-specific desktop installations. The fact that the Lingsoft service does not need to learn each user’s voice individually also made for a seamless adoption. Another indication of THL’s satisfaction with the service is its desire to directly integrate Lingsoft’s service as part of THL’s forensic medicine information system. The solution in use at THL is constantly being improved in co-operation with the client.