Speech Technology and Forensic Phonetics

Speech Technology

• Speech and language technology (SLT) uses computational methods to process and understand human speech and written language.
• SLT comprises numerous aspects of human-computer interaction, including speech & voice recognition, predictive text, voice-command interfaces, spell & grammar checkers, document summarization, and text-to-speech synthesis.
• SLT relies on automated parsing and analysis of human language.
• SLT comprises two subfields: speech processing and natural language processing (NLP).

Forensic Phonetics

• Forensic phonetics is the application of knowledge, theories, and methods of general phonetics to practical tasks related to police work or legal proceedings.
• It includes developing new forensic-phonetic knowledge, theories, and methods.
• Experts in forensic phonetics often use a combination of software, expertise, and statistical approaches in their analyses.
• Computer scientists have developed technologies to automate linguistic and phonetic analyses using Artificial Intelligence (AI).

Multimodal Technologies

• Multimodal technologies integrate multiple types of input and output data, such as speech, text, images, and gestures to enhance human-computer interaction.
• Aim to create more natural, accurate, and usable systems.
• Examples of multimodal systems:
  • Virtual assistants like Siri and Alexa.
  • Interactive language tutoring platforms.
  • Speech recognition systems incorporating visual context.
• These technologies are key advancements that create more engaging and natural communication systems.

Speech Processing

• Speech processing is the science of how speech communication works: production by the speaker and understanding by the listener.
• It's also about analyzing and modeling these processes and using those models for technologies that produce and understand speech (synthetic voices, speech recognizers).
• Speech technology is vital for understanding and remediating disordered speech.
• Speech technology intersects various disciplines, particularly linguistics, psychology, acoustics, and engineering.

Core Speech Technologies

• Automatic Speech Recognition (ASR) transcribes speech to text.
• Text-to-Speech (TTS) transforms written text into speech.
• Language Generation translates concepts into words.
• Spoken language understanding (similar to written language parsing).
• Speaker verification and voice print technologies are also fundamental.

Language Processing

• Language processing focuses on computational theories of grammar and meaning.
• It provides access to fundamentals of linguistics as a science and an engineering discipline.
• It's concerned with language as a medium for thought and communication.
• It's used in tools like predictive text, automated personal assistants, web search, and sentiment analysis.

Key Technologies in Speech and Language Technology

• ASR: covers its workings and major systems like Google Speech and DeepSpeech.
• TTS: outlines its pipeline and popular engines like Amazon Polly and Festival.

Real-World Use Cases

• Speech assistants (Siri, Alexa, Google Assistant).
• Real-time transcription (Otter.ai).
• Machine translation (DeepL, Google Translate).
• Healthcare and accessibility applications.

Course Content

• Course topics:
  • Introduction to Forensic Phonetics and Linguistics Applications.
  • Speaker Identification and Verification, Voice print, and the Document Examiner (using Audacity software).
  • Challenges of Spoken Arabic Processing.
  • Speech Synthesis (using Praat).
  • Arabic Text-to-Speech.
  • Speech Recognition System (using Google Cloud).
  • Data in Forensic Phonetics.

Pre-Reading for Next Lecture

• Read and summarize the chapter: Sinha.S (2015): Forensic Linguistics and Forensic Phonetics: An Introduction, International Journal of Interdisciplinary and Multidisciplinary Studies (IJIMS), 2015, Vol 2, No.6, 153-157.