Amr Nour-Eldin, is the Vice President of Know-how at LXT. Amr is a Ph.D. analysis scientist with over 16 years {of professional} expertise within the fields of speech/audio processing and machine studying within the context of Automated Speech Recognition (ASR), with a selected focus and hands-on expertise lately on deep studying methods for streaming end-to-end speech recognition.
LXT is an rising chief in AI coaching knowledge to energy clever know-how for world organizations. In partnership with a global community of contributors, LXT collects and annotates knowledge throughout a number of modalities with the velocity, scale and agility required by the enterprise. Their world experience spans greater than 145 international locations and over 1000 language locales.
You pursued a PhD in Sign Processing from McGill College, what initially you on this subject?
I all the time needed to check engineering, and actually appreciated pure sciences normally, however was drawn extra particularly to math and physics. I discovered myself all the time attempting to determine how nature works and how you can apply that understanding to create know-how. After highschool, I had the chance to enter drugs and different professions, however particularly selected engineering because it represented the proper mixture for my part of each principle and software within the two fields closest to my coronary heart: math and physics. After which as soon as I had chosen it, there have been many potential paths – mechanical, civil, and so forth. However I particularly selected electrical engineering as a result of it is the closest, and the hardest for my part, to the kind of math and physics issues which I all the time discovered difficult and therefore, loved extra, in addition to being the inspiration of recent know-how which has all the time pushed me.
Inside electrical engineering, there are numerous specializations to select from, which usually fall beneath two umbrellas: telecommunications and sign processing, and that of energy and electrical engineering. When the time got here to decide on between these two, I selected telecom and sign processing as a result of it is nearer to how we describe nature by physics and equations. You are speaking about alerts, whether or not it is audio, photographs or video; understanding how we talk and what our senses understand, and how you can mathematically signify that info in a method that permits us to leverage that information to create and enhance know-how.
May you talk about your analysis at McGill College on the information-theoretic side of synthetic Bandwidth extension (BWE)?
After I completed my bachelor’s diploma, I needed to maintain pursuing the Sign Processing subject academically. After one 12 months of finding out Photonics as a part of a Grasp’s diploma in Physics, I made a decision to modify again to Engineering to pursue my grasp’s in Audio and Speech sign processing, specializing in speech recognition. When it got here time to do my PhD, I needed to broaden my subject a bit bit into basic audio and speech processing in addition to the closely-related fields of Machine Studying and Data Principle, reasonably than simply specializing in the speech recognition software.
The car for my PhD was the bandwidth extension of narrowband speech. Narrowband speech refers to traditional telephony speech. The frequency content material of speech extends to round 20 kilohertz, however the majority of the knowledge content material is concentrated as much as simply 4 kilohertz. Bandwidth extension refers to artificially extending speech content material from 3.4 kilohertz, which is the higher frequency certain in standard telephony, to above that, as much as eight kilohertz or extra. To higher reconstruct that lacking larger frequency content material given solely the out there slim band content material, one has to first quantify the mutual info between speech content material within the two frequency bands, then use that info to coach a mannequin that learns that shared info; a mannequin that, as soon as skilled, can then be used to generate highband content material given solely narrowband speech and what the mannequin realized concerning the relationship between that out there narrowband speech and the lacking highband content material. Quantifying and representing that shared “mutual info” is the place info principle is available in. Data principle is the research of quantifying and representing info in any sign. So my analysis was about incorporating info principle to enhance the factitious bandwidth extension of speech. As such, my PhD was extra of an interdisciplinary analysis exercise the place I mixed sign processing with info principle and machine studying.
You have been a Principal Speech Scientist at Nuance Communications, now part of Microsoft, for over 16 years, what have been a few of your key takeaways from this expertise?
From my perspective, an important profit was that I used to be all the time engaged on state-of-the-art, cutting-edge methods in sign processing and machine studying and making use of that know-how to real-world functions. I acquired the prospect to use these methods to Conversational AI merchandise throughout a number of domains. These domains ranged from enterprise, to healthcare, automotive, and mobility, amongst others. Among the particular functions included digital assistants, interactive voice response, voicemail to textual content, and others the place correct illustration and transcription is crucial, reminiscent of in healthcare with physician/affected person interactions. All through these 16 years, I used to be lucky to witness firsthand and be a part of the evolution of conversational AI, from the times of statistical modeling utilizing Hidden Markov Fashions, by the gradual takeover of Deep Studying, to now the place deep studying proliferates and dominates virtually all points of AI, together with Generative AI in addition to conventional predictive or discriminative AI. One other key takeaway from that have is the essential function that knowledge performs, by amount and high quality, as a key driver of AI mannequin capabilities and efficiency.
You’ve revealed a dozen papers together with in such acclaimed publications as IEEE. In your opinion, what’s the most groundbreaking paper that you simply revealed and why was it necessary?
Essentially the most impactful one, by variety of citations in accordance with Google Scholar, can be a 2008 paper titled “Mel-Frequency Cepstral Coefficient-Based mostly Bandwidth Extension of Narrowband Speech”. At a excessive degree, the main focus of this paper is about how you can reconstruct speech content material utilizing a function illustration that’s extensively used within the subject of computerized speech recognition (ASR), mel-frequency cepstral coefficients.
Nonetheless, the extra modern paper for my part, is a paper with the second-most citations, a 2011 paper titled “Reminiscence-Based mostly Approximation of the Gaussian Combination Mannequin Framework for Bandwidth Extension of Narrowband Speech“. In that work, I proposed a brand new statistical modeling approach that comes with temporal info in speech. The benefit of that approach is that it permits modeling long-term info in speech with minimal further complexity and in a vogue that also additionally permits the technology of wideband speech in a streaming or real-time vogue.
In June 2023 you have been recruited as Vice President of Know-how at LXT, what attracted you to this place?
All through my educational {and professional} expertise previous to LXT, I’ve all the time labored instantly with knowledge. In actual fact, as I famous earlier, one key takeaway for me from my work with speech science and machine studying was the essential function knowledge performed within the AI mannequin life cycle. Having sufficient high quality knowledge in the suitable format was, and continues to be, very important to the success of state-of-the-art deep-learning-based AI. As such, after I occurred to be at a stage of my profession the place I used to be looking for a startup-like atmosphere the place I might be taught, broaden my expertise, in addition to leverage my speech and AI expertise to have probably the most impression, I used to be lucky to have the chance to hitch LXT. It was the proper match. Not solely is LXT an AI knowledge supplier that’s rising at a powerful and constant tempo, however I additionally noticed it as on the excellent stage by way of development in AI know-how in addition to in consumer dimension and variety, and therefore in AI and AI knowledge varieties. I relished the chance to hitch and assist in its development journey; to have a huge impact by bringing the attitude of an information finish consumer after having been an AI knowledge scientist consumer for all these years.
What does your common day at LXT seem like?
My common day begins with trying into the newest analysis on one subject or one other, which has recently centered round generative AI, and the way we are able to apply that to our clients’ wants. Fortunately, I’ve a wonderful group that may be very adept at creating and tailoring options to our shoppers’ often-specialized AI knowledge wants. So, I work carefully with them to set that agenda.
There’s additionally, in fact, strategic annual and quarterly planning, and breaking down strategic aims into particular person group objectives and protecting in control with developments alongside these plans. As for the function growth we’re doing, we usually have two know-how tracks. One is to verify we’ve got the suitable items in place to ship the very best outcomes on our present and new incoming tasks. The opposite monitor is bettering and increasing our know-how capabilities, with a deal with incorporating machine studying into them.
May you talk about the varieties of machine studying algorithms that you simply work on at LXT?
Synthetic intelligence options are remodeling companies throughout all industries, and we at LXT are honored to supply the high-quality knowledge to coach the machine studying algorithms that energy them. Our clients are engaged on a variety of functions, together with augmented and digital actuality, laptop imaginative and prescient, conversational AI, generative AI, search relevance and speech and pure language processing (NLP), amongst others. We’re devoted to powering the machine studying algorithms and applied sciences of the long run by knowledge technology and enhancement throughout each language, tradition and modality.
Internally, we’re additionally incorporating machine studying to enhance and optimize our inner processes, starting from automating our knowledge high quality validation, to enabling a human-in-the-loop labeling mannequin throughout all knowledge modalities we work on.
Speech and audio processing is quickly approaching close to perfection with regards to English and particularly white males. How lengthy do you anticipate it will likely be till it’s a fair taking part in subject throughout all languages, genders, and ethnicities?
This can be a sophisticated query, and is dependent upon a variety of components, together with the financial, political, social and technological, amongst others. However what is evident is that the prevalence of the English language is what drove AI to the place we are actually. So to get to a spot the place it is a degree taking part in subject actually is dependent upon the velocity at which the illustration of information from completely different ethnicities and populations grows on-line, and the tempo at which it grows is what is going to decide after we get there.
Nonetheless, LXT and related firms can have a giant hand in driving us towards a extra degree taking part in subject. So long as the info for much less well-represented languages, genders and ethnicities is tough to entry or just not out there, that change will come extra slowly. However we try to do our half. With protection for over 1,000 language locales and expertise in 145 international locations, LXT helps to make entry to extra language knowledge doable.
What’s your imaginative and prescient for the way LXT can speed up AI efforts for various shoppers?
Our objective at LXT is to supply the info options that allow environment friendly, correct, and sooner AI growth. By way of our 12 years of expertise within the AI knowledge area, not solely have we gathered in depth know-how about shoppers’ wants by way of all points referring to knowledge, however we’ve got additionally repeatedly fine-tuned our processes in an effort to ship the very best high quality knowledge on the quickest tempo and greatest value factors. Consequently, because of our steadfast dedication to offering our shoppers the optimum mixture of AI knowledge high quality, effectivity, and pricing, we’ve got grow to be a trusted AI knowledge associate as evident by our repeat shoppers who preserve coming again to LXT for his or her ever-growing and evolving AI knowledge wants. My imaginative and prescient is to cement, enhance and increase that LXT “MO” to all of the modalities of information we work on in addition to to all varieties of AI growth we now serve, together with generative AI. Attaining this objective revolves round strategically increasing our personal machine studying and knowledge science capabilities, each by way of know-how in addition to assets.
Thanks for the nice interview, readers who want to be taught extra ought to go to LXT.
