FAQs – Frequently Asked Questions
- What kinds of cases do Language Intelligence work on?
- What programs do you use in speech analysis?
- What is the process for speech analysis?
- How long does a typical voice comparison take?
- What is the process with determining authorship, regarding textual cases?
- Who have Language Intelligence worked with?
- What’s interesting about the work Language Intelligence has completed thus far?
- What steps should someone aspiring to work in forensic linguistics take?
- What about handwriting analysis?
- Do Language Intelligence work on music and song plagiarism?
- Do Language Intelligence work with only Australian agencies?
What kinds of cases do Language Intelligence work on?
One of the services we provide is voice analysis. The most common type of voice analysis we provide is speaker identification. Sometimes it’s with disputed identities for crimes, other times it’s for voice impersonation. We also do something that’s called ‘speaker profiling.’
We work alongside state and federal agencies to identify suspects using both acoustic details, and sociolinguistic features, such as word choice.
We are not limited to voice analysis. Language Intelligence has worked on textual analysis dealing with authorship in cases of plagiarism and defamation.
We also compile and enrich data for machine translation projects that utilise corpora.
What programs do you use in speech analysis? Video
We have several different software at our disposal, like Praat, Wavesurfer, and Audacity for speech analysis. We use the program R for statistical analysis. It’s not special software that makes the difference, but professional experience and know-how that matters.
What is the process for speech analysis? Video
Language Intelligence takes a four-step approach to speech analysis.
STEP 1: We listen to all the speech samples provided by the client requesting our advice.
STEP 2: We transcribe the content of the speech samples in the International Phonetic Alphabet and give detailed annotations of the transcriptions. This is done by both Dr Mark Donohue and Dr Paul Sidwell separately, to ensure objectivity.
STEP 3: The notes are compared for both agreements and differences and analysed accordingly.
STEP 3a: For auditory analysis (with the ear alone), matching features in the questioned and identified speech samples are correlated.
STEP 3b: For acoustic analysis (with measurements using computer), voices are measured for fundamental frequency and other features (as appropriate), and modeled for statistical comparison.
STEP 4: Reports are drawn up detailing findings. This can include quantification of similarity, and potentially the likelihood that the questioned speech was produced by the same person and the known speech sample.
How long does a typical analysis take? Video
It takes a couple days to deal with the analysis of a short audio recording with a reference sample. And we normally require a couple weeks’ notice to schedule work.
What is the process with determining authorship, regarding textual cases? Video
The process for determining authorship is much the same in principle as in speaker identification. We read and annotate the text(s), and correlate features: this means identifying and matching style markers (i.e., defining characteristics of an individual’s writing style), of which there are over 200 are recognized in the forensic linguistics literature. These can be analysed statistically to model the author habits and quantify similarity across reference texts.
Who have Language Intelligence worked with? Video
Language Intelligence has worked with a variety of different state and federal agencies. These include New South Wales Police, Western Australia Police, South Australia Police, Victoria Police, Federal Police, Defense Science and Technology Group, Counter Terrorism and Special Tactics Command. We have also assisted private counsel in legal cases.
What’s interesting about the work Language Intelligence has completed thus far? Video
One of the most interesting parts about this line of work is the diversity of events and the conduct of participants observed in the recordings. The most positive experience is when we can help solve a crime. Two downsides of this line of work are that 1) we often deal with recordings of distressing or traumatic events, and 2) occasionally we are unable to get a positive result.
What steps should someone aspiring to work in forensic linguistics take? Video
Anyone aspiring to work in Forensic Linguistics should first study linguistics at the university level, it is also recommended that you do courses in criminology and general forensic science courses. There are published handbooks and textbooks in forensic linguistics, and the International Journal of Speech, Language and the Law. You should familiarize yourself with the literature and see what interests you.
What about handwriting analysis? Video
Handwriting isn’t such a big area in the era of digital communication, but it still features in suicide notes, threatening letters, and such. As long a text has features that can be counted and analysed statistically, we can quantify its similarity to known texts.
Do Language Intelligence work on music and song plagiarism?
Potentially. Dr Sidwell has been approached on multiple occasions but has yet to pursue specific cases. Music and song plagiarism is a problematic area in that the information structure of music and song is not as complex as other media, and a certain amount of copying of style/content is acceptable. Language Intelligence is willing to work on cases pertaining to this topic on a case-by-case basis.
Do Language Intelligence work with only Australian agencies? Video
Language Intelligence is open to working with agencies and individuals internationally. To date, the principle business has been within Australia, although we have done some limited international work.