Category Archives: Uncategorised

Recap Semester 1 2022/2023

Since September 2022, our research group has been conversing in rich discussion about various research projects occurring both within and outside our P&P team. Recent talks and presentations which have been discussed this semester include:

  • SLS Seminar- Leveraging the Adaptable Speech Perception System for Dysarthria Remediation (Stephanie Borrie):
    Carol-Ann McConnellogue gave a run through of Stephanie Borrie’s seminar for those members who could not make her presentation. Her work stresses the importance of taking the focus off speakers with speech disorders impacting their intelligibility and encouraging listeners to take more responsibility in communication. Stephanie found that listeners increased their perceptive skills of disordered speech through listening, transcribing, and imitating the disordered speech.
  • Working Memory and Speech Perception (Kai Alter):
    Kai presented a study about working memory and lexical access. The study stressed the importance of memory capacity as (a) it correlates with language comprehension in children and geriatric populations; (b) it relates to performance on school achievement tasks; and (c) it offers opportunities for intervention in SLI and DLD, with the causes of DLD still under debate. This study used high-frequency, mono-syllabic concrete nouns recorded and presented at different paces to test participants’ maximal number of items to be recalled. Unlike previous research, it found that the medical number of memory capacity was 4 instead of 7 plus or minus 2.
  • Retainability and Sustainability of Phonetic Discrimination Ability of an English Vowel Contrast by German infants (Hiromasa Kotera):
    Hiromasa presented his preliminary PhD research to the group. More information on his presentation can be found HERE.
  • Tianjin Mandarin Prosody (Cong Zhang):
    Cong came to the group with some questions and ideas regarding a presentation she planned to present at talks in Cambridge and Edinburgh. Topics she brought to the group included: (a) Is there intonation in Chinese?; (b) tone systems of Standard Mandarin vs. Tianjin Mandarin; (c) floating boundary tone; (d) functional Principal Component analysis (fPCA); (e) chanted call tune in Tianjin Mandarin; and (f) how she can turn data from recorded TV shows into a scientific study. The group provided Cong with feedback and advice.
  • Nonword Learning Project (Rory Turnbull):
    Rory presented his nonword learning project to the group so that he could receive general feedback on the overall idea, specific feedback on the pilot methodology, and recommendations of any literature and related topics. His main question posed to the group focused on why languages are the way they are, with a secondary idea questioning what are the things that influence language structure. More information on his presentation can be found HERE.
  • Correlates of Stress in Palestinian Arabic (Niamh Kelly):
    Niamh discussed her study which is investigating the main indicators of lexical stress in Palestinian Arabic through acoustical analysis. After her presentation she opened the floor to members of the group for feedback and ideas on how to improve her methodology and analysis. More information on his presentation can be found HERE.
  • Pathological Speech Analysis (Shufei Duan):
    Shufei is a visiting scholar of scalable group. She is currently Associate Professor in Taiyuan University of Technology in China, Shanxi Province. Shufei spoke to us about her project focusing on the classification of imbalanced data of Chinese dysarthria based on affective articulation. The project forms an emotional Chinese pronunciation dataset and an emotional pathological speech dataset and conducts research on the data processing of affective articulatory movement for Chinese dysarthria. Her aim is to develop an AI-assisted medical treatment and diagnosis so that diagnoses can be made remotely. Shufei’s work was of great interest to many group members and gave rise to several question-and-answer sessions. She has since attended some of our sessions, alongside her host Elly, in Semester 2 to learn about the research occurring within our group.

A new subgroup with the lovely name PIG (Prosody Interest Group) has been set up this year for our members who have an interest in prosody and intonation. This group will involve practice talks, feedback on research ideas, and discussion of relevant papers.

Some members of our group have submitted abstracts to the International Congress of Phonetic Sciences (ICPhS). We wish them the best of luck in their submissions.  

We would also like to welcome three new members to our research group: Dr Cong Zhang, Niamh Kelly, and Hajar Moussa.

Stefano Coretta’s Workshop on Electroglottography (EGG)

Date: 06/27/2021

Stefano Coretta from the University of Edinburgh has a wide range of knowledge in statistical analyses and research methods in linguistics research. Today we had him walk us through how to operate the EGG machine, apply it in the practical study and process the data.


We firstly started with the anatomy of larynx. 

  1. When making voiced sounds, the vocal cords would contact and dispart over time. 
  2. The frequency of vocal-fold vibration varies between sounds and speakers. For the same sound, if produced in certain distinctive voices, like in creaky or breathy voices, the vocal folds would vibrate differently, which can be measured by the EGG. 


We learned about the principles of the EGG and got the chance to see how to use the machine.

  1. The principle of EGG is that there are two metal signal-receiving tabs, which would be put against the neck skin. When the two larynges contact each other, the current is allowed to pass, which would be recorded as waveforms. 
  2. The maximum of the EGG signals corresponds to the close phase of the glottis, and the minimum corresponds to the open phase of the glottis.
  3. The EGG can be used to measure
  • Robust estimation of fundamental frequency (f0).
  • Detection of voiced/voiceless intervals.
  • Voice quality via Contact Quotient (CQ): higher CQ = creakier voice, lower CQ = breathier voice.
  • Vocal Folds configuration via complex methods like wavegram.

4. The limitation of the EGG is that it only works when there is vocal fold vibration and shows no information when there is a voiceless interval.

Analysing the EGG signal

There are various ways to analyse the EGG data, which can be processed on different platforms, including Praat, R, Matlab, Python and so on. To process the EGG signal, what usually needs to be done are:

  • Pre-processing

       – Filtering

       – Calculate the first derivative (dEGG)

  • Extract measurements

       – Contact Quotient (ratio between closed phase and glottal cycle)

       – Wavegram (Herbst 2010)

  • Analyse measurements

       – WaveGAMs (Coretta 2019)

Calculate measurements

We can get various measurements with the use of different tools. For example:

In Praat:

  1. Get maxima and minima of the dEGG.
  2. Calculate the glottal period.

In R:

  1. Calculate the Contact Quotient.
  2. Calculate f0.

Reflections from our Research Group

  • Our group has long been exploring different quantitative research methods and data analyses in linguistics. This workshop is very relevant to what we are pursuing.
  • It gave us a clear understanding of using the EGG.
  • This gave us the confidence to look at glottal movements in our research.

Three-minute thesis final

Date: 16th June, 16:00 – 18:00 
Location: Baddiley Clark Seminar Room
Book Drinks reception and canapés from 16:00  Presentations will begin at 17:00

The finals of the 3-minute thesis are taking place on the 16th of June. One of our members, Carol-Ann McConnellogue, has made it to the final. Carol-Ann is developing an individualised speech therapy programme for children with cerebral palsy and is doing her PhD jointly with ECLS and FMS.

The Three Minute Thesis (3MT) competition asks doctoral students to explain their research in just three minutes using only one slide. The explanation should be easily understood by a non-specialist. Originally developed by the University of Queensland, Australia it has been taken up by Universities across the world. The competition offers training then the opportunity to compete in a University final in front of the public. The winner of this final will go forward to compete in the national Vitae 3MT competition in September.

It’s a great opportunity to listen to students from different disciplines talk about their PhD topics in a succinct and non-technical way.

Recap Semester 2 2021/2022

From January 2022 to May 2022, our research group has continued to be well-engaged in projects from Semester 1 and arranged new workshops to discuss topics of interest amongst our team members. The following is a summary of what we covered this semester:

  • Accent and Social Justice

Since the beginning of this academic year, we have focused on the theme ‘Accent and Social Justice’, reviewed several related articles, and had Melissa Baese-Berk from the University of Oregon share her and her colleagues’ recent research with us. We organised and held an interdisciplinary workshop on accent, communication, and social justice in March of this Semester which was very successful. We were honoured to have presenters from both within our research group and outside of the group share their research and opinions. More information can be found in this blog post.

  • Many Speech Analyses

One of our main discussion topics of Semester 2 has been the Many Speech Analyses project we signed up for at the end of last semester. This project aims to compare what approaches different researchers take to answer the same research question using the same dataset. The general research question is: ‘Do speakers phonetically modulate utterances to signal atypical word combinations?’. We scheduled fortnightly meetings for this project. We started by reviewing other studies to help us plan a suitable analysis and decided to measure the timing of utterances to answer the research question. We imported the sound files to the MFA (Montreal Forced Aligner) for the forced alignment, and the results were distributed to the members for the crossed hand-correcting. Rory Turnbull, our project leader who is also a member of the P&P research group, guided us in extracting the timing of articulation of certain vowels. After analysing the dataset and submitting our report, we took a few weeks to review the reports from other researchers/research groups. Some peer analyses involved certain research methods or related tools unfamiliar to us, allowing us to expand our knowledge outside our expertise. These included:

  • Forced alignment and inter-rater reliability in Praat

During a couple of weekly meetings, we had Caitlin Halfacre and Rory run Forced alignment in the Montreal Forced Aligner and demonstrate how to hand-correct it in Praat, such as tier setting, labelling and calibration of the initial phone etc. Group members teamed up separately to help each other and shoot problems together. Bruce Wang coded in Praat to sample and measure the agreement of each text grid from the crossed hand-correct. The inter-rater reliability of our group members turned out to be quite strong.

  • Praat Phonetic Analysis 

After checking the correctness and reliability of phone alignment, Rory led two sessions demonstrating how to extract specific labels and measure the timing of utterances by coding in Praat.

  • Digital Signal Processing (DSP)

When we reviewed other researchers’ reports, we found certain research methods which were unfamiliar to us, such as DSP. We used a session as an introduction to these techniques. During this session, we watched a video to recap the anatomy of sound perception, discussed the anatomy of the cochlea, and talked about the acoustic versus auditory differences between two tones that are 100 Hz apart and gammatone filter back. However, without a well-established background in neurolinguistics, it’s still difficult for us to fully understand what the results of one peer reviewed report meant.

To conclude, we successfully ran the ‘Accent and Social Justice’ workshop and completed the Many Speech Analyses project together this semester and learnt much research knowledge and relevant expertise from this experience. We expect to explore more exciting topics and themes in the future and keep updating and publicising our work here.

LabVanced Workshops and the Lab Tour

We are excited to announce that our linguistics laboratory, which was under refurbishment last year, has been reopened. The Lab has many world-class pieces of equipment. It has two small sound-treated isolation booths and one sizeable sound-proof booth to provide an ideal experimental environment. It is also equipped with high-quality phonetic and speech recording tools, as well as ultrasound tongue imaging. The Lab can support a wide range of articulatory, phonetic, psycholinguistic, and corpus research, and you can read here for more information.

The Lab is licensed to use LabVanced, an online experimental designing and launching system. On the 19th and 25th May, Fengting Du and Andreas Krug, who are members of our research group, demonstrated how they used LabVanced to benefit their research. The questions covered were: (1) how to start up a new study; (2) how to import your stimuli; (3) how to set variables and record the data you want; (4) how to randomize the order of your questions or stimuli; and (5) how to launch your study and export your data. We also shot some problems about data analysis together during the workshops.

After the workshop, the research group had a demonstration of ultrasound tongue imaging from the Lab administrators. If you are interested in the Lab equipment and want to know more, you can call into the Lab or book the Lab here.

Please leave a comment if you have questions on LanVanced or would be interested in another training session. We look forward to arranging further workshops like this.