DEV Community: Kelsey Foster

Best 5 PII Redaction APIs for 2022-2023

Kelsey Foster — Tue, 06 Sep 2022 22:51:15 +0000

Any company that handles customer data must meet internal privacy requirements or external compliance regulations like GDPR and HIPAA. With digital data, manually finding and removing this sensitive information at scale can be nearly impossible.

Some companies are looking to create features or products that can remove or redact this confidential data automatically. State-of-the-Art AI models that power APIs for PII (Personally Identifiable Information) Redaction can help.

Built using cutting-edge Machine Learning research, PII Redaction APIs can automatically identify PII in bodies of text and remove, redact, or sensor this sensitive content–at high accuracy. Some PII Redaction APIs also work in conjunction with Speech-to-Text APIs to redact confidential information in transcriptions or to remove utterances from audio/video streams.

If you’re considering building tools with a PII Redaction API, there are several top choices to consider. This article examines the best APIs on the market for performing PII Redaction for 2022/2023:

1. AssemblyAI

AssemblyAI is an API platform for State-of-the-Art, production-ready AI models for ASR, NLP, and NLU applications. Developers and enterprises transcribe audio/video streams with its Speech-to-Text API and then create powerful, intelligent tools on top of that transcription data using its AI models such as PII Redaction, Content Moderation, Summarization, Sentiment Analysis, and more.

AssemblyAI’s PII Redaction API detects and replaces confidential content like social security numbers, credit card numbers, and addresses with a series of # for each redacted character. The API can also beep out spoken PII in the audio file as well.

Pricing for AssemblyAI’s PII Redaction API starts at $.000583 per second on top of its core transcription pricing.

2. Private AI

Private AI is another top tier PII Redaction API that lets users identify, replace, or redact PII in large text documents, audio/video files, or even images. The API also supports redaction across multiple languages.

For text files, users can choose to replace redacted data with a series of # or to replace the data with synthetic data if security issues are a concern. For images, the API will blur out the needed PII.

Pricing for Private AI is broken into three tiers, depending on usage: Starter, Scale, and Pro.

3. Amazon Transcribe

Amazon Transcribe also offers its own PII Redaction API for text and live or asynchronous audio/video streams, though only for English.

Amazon Transcribe’s PII Redaction API can identify and redact PII such as bank account numbers, bank routing numbers, email addresses, credit card numbers, credit card CVV codes, and more. However, its documentation states that its PII Redaction API does not meet the requirements to meet privacy laws such as HIPAA.

Pricing for Amazon Transcribe and its PII Redaction API can be a bit hard to decipher but interested users can calculate estimated pricing based on usage needs here.

4. Azure

Azure has another top-rated API for PII Redaction. The API can identify, remove, or redact sensitive entities such as a person’s name, job type, medical information, IP address, account numbers, SWIFT codes, and more. Redaction is only available for files with text in English.

Users must have an Azure account and Visual Studio IDE to use its PII Redaction API. To get started, follow Azure’s quickstart guide here.

5. Super.ai

Finally, Super.ai is another great PII Redaction API for removing/redacting confidential information in text, video, and images. Its PII Redaction API is also compliant with stringent international data privacy regulations such as GDPR, CCPA, PIPL, and PIPA.

The API returns each redacted file with the redacted PII airbrushed out of the document/image or replaced with pseudonyms.

Those interested in learning more about Super.ai or its pricing structure can sign up for a free demo here.

5 Best APIs for Performing Text Summarization for NLP

Kelsey Foster — Wed, 18 May 2022 15:03:32 +0000

Text Summarization for NLP refers to the process of shortening papers, podcasts, documents, videos, and other large bodies of texts into their most important parts. This is done by leveraging Deep Learning and Machine learning models.

Typically, Text Summarization APIs summarize either pre-existing, or static, bodies of texts, like a legal document, or to audio streams, like a YouTube video, that is transcribed via ASR-technology like a top Speech-to-Text API.

Developers and enterprises alike use Text Summarization APIs to:

Summarize large documents and flag sections for follow-up
Create chapters for YouTube videos or videos shared via an educational course
Automatically providing a Table of Contents for podcasts
Summarizing calls in a cloud-based contact center to increase agent occupancy

And more.

This article compares APIs that perform Text Summarization on either category, or both, with top-rated accuracy.

Here are the 5 Best APIs for Performing Text Summarization for NLP:

1. AssemblyAI’s Auto Chapters API

AssemblyAI is a leading speech recognition provider with industry-best accuracy. In addition to its core transcription API, it offers a host of Audio Intelligence APIs that let users build high ROI features on top of their audio data. One of these Audio Intelligence APIs is Auto Chapters, its Text Summarization API. When users run an audio file through its Auto Chapters API, the API will return both a one paragraph summary and single headline for each “chapter,” or points where the audio naturally changes topics.
Pricing for its Auto Chapters API begins at $0.000583/second. Users can sign-up to test the API for free here.

2. NLP Cloud Summarization API

NLP Cloud’s Text Summarization API offers Summarization for pre-existing bodies of texts at decent accuracy, in addition to other NLP/NLU APIs. Its online community also offers AI models that its community of members can use to train, fine-tune, and deploy their own models into production. Users can access NLP Cloud’s APIs for free up to a certain usage each month, with plans then ranging up to $499/month.

3. Microsoft Azure Text Summarization

Microsoft Azure is another top name in the ASR and NLP world. Its Text Summarization API is offered as part of its Text Analytics suite and is available for static texts. Pricing for Azure varies significantly depending on usage and if other APIs are also needed, with a pay-as-you-go payment structure. You’ll also need an Azure subscription to get started.

4. plnia’s Text Summarization API

Plnia also offers a Text Summarization API for pre-existing bodies of text, in addition to other NLU APIs such as Abusive Language Check, Keyword Extractor, Sentiment Analysis, Article Extraction, Language Detection, and more. For those looking to test plnia, the company offers a 10-day free trial and then plans start at $19/month after that.

5. MeaningCloud’s Automatic Summarization

Finally, MeaningCloud offers an “Automatic Summarization” API for static bodies of text. Its model works by first extracting the most important sentences in the document and then using those sentences to build its summary. MeaningCloud’s API also supports a wide range of languages, so users can still use Text Summarization regardless of the text’s native language. Users looking to try MeaningCloud can sign-up for a free developer account; fees then range from $0-$999/month, depending on usage.

Top 9 ML, AI, and Data Science Internships in 2022

Kelsey Foster — Tue, 08 Mar 2022 17:49:29 +0000

Internships are a great way to get hands-on, practical experience in the fields of Machine Learning, Artificial Intelligence, and Data Science. Those looking for internships in these, and related fields, should explore the options below.

Be sure to note requirements–some internships have minimum lengths and only accept those pursuing higher educational degrees while others offer rolling acceptance for a range of experiences and backgrounds.

All of them, however, will provide interns an opportunity to apply their theoretical and practical skills to real word, tangible scenarios.

Here are the Top 9 ML, AI, and Data Science Internships to consider for 2022:

1. Data Science Intern, Meta

Facebook’s Meta is hiring for a group of data science interns interested in learning more about using Data Science for more effective advertising, marketing, and sales applications. Interns can expect to gain real world experience in designing and executing data science projects, as well as in performing strategic analysis over the twelve week internship. A Bachelor’s or Master’s Degree in a quantitative field, as well as experience with data querying and manipulation and data analysis in a scripting language, is required to apply.

Apply here.

2. Machine Learning Intern, Motional

Self-driving vehicle company Motional is looking for Machine Learning interns to join its team for Summer 2022. Those interested can expect to get hands-on experience in research and developing Machine Learning strategies that can solve complex problems. Interns will also collaborate with additional teams for greater exposure and learning opportunities. Candidates must have a Master’s Degree or Ph.D. in Computer Science, Robotics, or a related field to apply.

Apply here.

3. Deep Learning and more, Borealis AI

Borealis AI is an AI and Machine Learning Research Institute funded by the Royal Bank of Canada. It is currently hiring virtual internships for Summer 2022 across Deep Learning, Reinforcement Learning, Graphs and Optimization, Unsupervised and Semi-supervised Learning, Bayesian Optimization, Privacy and Fairness, Interpretability and Explainability, AutoML, Time Series Forecasting, Computer Vision, and Natural Language Processing areas such as ASR and Audio Intelligence. Candidates must hold a Master’s Degree or have experience working towards a Ph.D.

Apply here.

4. Machine Learning/AI Internship, Apple

Apple is looking to hire an AI/ML intern interested in Deep Learning, Computer Vision, Natural Language Processing, Optimization, and Reinforcement Learning. Those selected will research and develop cutting edge ML strategies and collaborate with broader teams across Apple. Candidates must be pursuing a Bachelor’s Degree, or greater, in a related field, have a strong publication record, excellent programming skills, and knowledge of common ML frameworks.

Apply here.

5. Research Intern, Applied Research - Natural Language Processing, Meta

Meta is also hiring for a Natural Language Processing research intern. Candidates should have deep interest in areas such as Deep Learning, Machine Learning, and NLP applications such as Sentiment Analysis and Speaker Diarization. Interns selected for the 12-week assignment will have the opportunity to make core algorithmic advances and submit research papers to top tier publications. Applicants who are currently pursuing a Ph.D. in a related field are preferred, though exceptional Master’s and Bachelor’s students may also be considered.

Apply here.

6. Autonomy - Deep Learning Infrastructure Intern, Tesla

Tesla is looking for a Deep Learning infrastructure intern to join its team for a minimum of 12 weeks, full time, starting in May or June 2022. If chosen, interns will be working on reinforcing, optimizing, and scaling Tesla’s Neural Network training infrastructure and building out its Machine Learning platform. Applicants must be currently enrolled in a university and have experience with Python, C++, PyTorch, and Machine Learning.

Apply here.

7. Artificial Intelligence Internship, Intel

Bachelor’s, Master’s, and Ph.D. candidates are welcome to apply to Intel’s Artificial Intelligence internship, where they will work to build large scale AI systems and build Machine Learning algorithms. Interns will work with mentors to gain real world Machine Learning and AI experience and present their findings in internal presentations. Applicants are welcome to apply at any point during the year for rolling acceptance.

Apply here.

8. Research Internship, Allen Institute for AI

Full-time, year-round internships are also available at the Allen Institute for AI for undergraduate and graduate students. Applicants need to be interested in AI-related fields such as computer vision, question answering, textual entailment, knowledge representation, Machine Learning, NLP, and/or semantics. Research internships are remote during the COVID-19 pandemic.

Apply here.

9. Internship, Center for Artificial Intelligence in Medicine and Imaging, Stanford University

The Center for Artificial Intelligence in Medicine and Imaging at Stanford University accepts post-doctoral, graduate, and a limited number of high school and undergraduate applicants to its research internship program. Internship foci vary based on availability. Post-doctoral scholars must have at least one first author publication to be considered.

Apply here.

Top Transcription APIs and Open Source Libraries in 2022

Kelsey Foster — Mon, 07 Mar 2022 17:18:01 +0000

The number of applications leveraging speech recognition and voice transcription technology has skyrocketed in the past decade. More people than ever before are using voice AI technology in their homes, cars, and places of business.

Advances in deep learning, machine learning, and AI research have powered this adoption, making speech recognition technology more accessible, affordable, and most importantly–accurate.

With this increase in interest and adoption, there’s also been a simultaneous increase in the number of speech transcription APIs and open source libraries available for users.

This article looks at some of the top transcription APIs and open source libraries available on the market today, as evaluated by accuracy, pricing, documentation, and additional features offered.

Top Transcription APIs

Three speech transcription APIs stand out in this category: AssemblyAI, Google Speech-to-Text, and AWS Transcribe.

1. AssemblyAI

AssemblyAI is a Speech-to-Text APIstartup with competitive accuracy and an easy-to-use interface. The API offers three free transcription hours per month, an affordable paid tier, and extensive documentation, making it a developer-favorite.**

As a startup, the API invests heavily in the latest deep learning research and is constantly shipping updates to improve its models. Most recently, the API released its suite of Audio Intelligence APIs that provide greater business value for its customers. These include sentiment analysis, content moderation, Entity Detection, PII Redaction, Summarization, and Automatic Transcript Highlights, with more expected to be released soon. **

Since it’s newer to the market, the API does lack a few of the features available from some of its more seasoned competitors.

2. Google Speech-to-Text

Google Speech-to-Text continuesto be a dominant player in the speech recognition market. With good accuracy, robust language support, and domain-specific models, it is a popular choice among other big-name companies.

Google’s name recognition comes with a higher price tag than other Speech-to-Text APIs, especially since the company only supports transcribing files in a Google Cloud Bucket. It can also be a bit complicated to use, as you must first sign up for a GCP account and project.

Still, those looking to test the API can do so with an initial 60 minutes of free transcription and $300 free for Google Cloud hosting.

3. AWS Transcribe

__AWS Transcribe __is another good option for larger companies. The API offers one hour of free transcription per month for the first twelve months of use. Accuracy, however, is somewhat lower than other APIs on the market today and documentation is not as regularly updated.

Like Google, getting started with AWS Transcribe can be a bit tricky and expensive, as it only supports files hosted in an Amazon S3 bucket.

Those looking for specialty transcription, such as the medical industry, should check out its Transcribe Medical API which is trained to perform accurately in this profession.

Top 3 Open Source Transcription Libraries

In addition to transcription APIs, there are a host of open-source transcription libraries available for public use. While free, open-source libraries require significantly more leg work than APIs in order to perform at high accuracy and utility.

However, if you’re willing to put in the effort, and have a basic understanding of speech recognition, these are the top three options to consider:

1. Wav2Letter

Wav2Letter, Facebook AI Research’s Automatic Speech Recognition (ASR)toolkit, is designed for research and developers to use for speech transcription.

With pre-trained models for the Librispeech dataset, it’s a good open source library to get started with quickly.

Wav2Letter boasts decent accuracy and is written in C++.

2. DeepSpeech

Built using the end-to-end model architecture pioneered by Baidu, DeepSpeechis a great open-source speech transcription option.

DeepSpeech is easy to work with, especially since it’s designed to work with a range of devices, from a Raspberry Pi 4 to a high-powered GPU.

It also has good out-of-the-box accuracy for an open-source library.

3. Kaldi

Finally, Kaldiis another very popular open-source speech recognition library.

Because of its popularity, there are an abundance of free tutorials to help you get started with training your own speech recognition models and customize your experience.**

Like DeepSpeech, Kaldi also has good out-of-the-box speech recognition accuracy and is designed to get developers started using it quickly.

What Is Data Bias and How to Avoid It

Kelsey Foster — Tue, 01 Mar 2022 19:24:53 +0000

Data bias can have significant implications for research and practical applications. Think back to last year’s Facebook scandal, where its AI asked users if they wanted to “keep seeing videos about primates” after watching a video featuring Black men. Viewers were understandably appalled, and it sparked a heated conversation about the limitations–and potential dangers–of AI-backed software.

While this may seem like an extreme example, it serves as a powerful reminder of the real-life implications of AI models trained with biased data.

What exactly is data bias? Data bias refers to data sets that are not representative of the population in study. Models trained on biased data may contain inherent prejudice against particular subjects or subpopulations.

Unfortunately, data bias is a common problem in AI and Machine Learning applications, often occurring unintentionally. That’s why AI Researchers and Data Scientists must be constantly vigilant to ensure that they train models free, or as free of as possible, of any type of bias.

This article will look at three ways to limit data bias: collecting data from a variety of sources, ensuring data is diverse, and monitoring real-world performance.

1. Collect data from a variety of sources

To avoid data bias, it’s imperative that data is collected from a wide variety of sources. Here are the most common avenues for collecting training data:

Paying for data sets
Using public data sets
Sourcing open source content
Using in-person or field-collected data sets

The best training data would be sourced from a combination of all four.

If your model involves predictions relating to speech, you’ll need to make sure that your overall data set is robust to all environments and background noise. This will help guarantee that your models can make just as (or nearly as) accurate predictions with noisy audio as they could with studio-quality audio.

2. Make sure data is diverse

In addition to collecting data from myriad sources, you’ll also want to make sure that the data itself is diverse. This means that speakers in the audio or video files encompass a wide range of characteristics, such as locations, dialects, genders, sex, race, nationality, and more.

Unfortunately, sourcing such diverse data may prove difficult, especially if you rely solely on open-source data. That’s why it’s important that these first two recommendations go hand in hand–a variety of sources and diverse data within each source.

3. Monitor real-world performance

Now that you’ve ensured that your initial data set is diverse, you can be confident that your model will be unbiased, right? While the chances are minimized, unfortunately, this isn’t a guarantee.

To make sure, it’s important that you monitor your model’s real-world performance, looking for any areas where bias may have crept in. Does your model better predict female speech over males? Midwestern speech over southern? If so, take time to retrain with new datasets to weed out any problem areas.

Training, testing, and retraining will be an iterative process over the model’s lifetime. For example, leading Speech-to-Text APIs are often trained on text with billions of words that help boost their accuracy. Then, researchers constantly strive to improve the model by looking for areas of deficiencies, sourcing new training data, and retraining.

By focusing on these three main steps–collecting data from many sources, ensuring a diverse data set, and monitoring model performance–you can be confident that your models will perform well in real-world scenarios and prevent any embarrassing missteps like Facebook’s AI disaster.

Top 12 Twitter Feeds to Learn About AI

Kelsey Foster — Tue, 01 Mar 2022 19:16:08 +0000

Twitter can be a great place to stay abreast of the latest AI research, keep connected with new AI companies and products, and get insight into the thinking of some of today’s top AI minds.

For those interested in AI, make sure these twelve Twitter accounts are on your daily feed. From the developer of scikit-learn to professors at top name universities to staff scientists at Google Brain, these accounts share AI research papers, industry events, tutorials, videos, blog posts, and more insightful AI commentary that you won’t want to miss.

Here are Top Twitter Feeds for AI:

1. Andreas Miller

Twitter Handle: @amuellerml

Andreas Miller helped develop scikit-learn, Python’s machine learning library, and is now a principal researcher at Azure Data. His Twitter feed is filled mostly with retweets of other top names and organizations in the AI and ML fields, giving you a great overview of current events and public commentary.

Follow here.

2. Lisha.eth

Twitter Handle: @lishali88

Lisha is the founder and CEO of Rosebud AI, a generative AI for web3 creators. She holds a Ph.D. in AI and mathematics from UC Berkeley. Her tweets include retweets and responses to the latest AI research as well as relevant commentary on the field.

Follow here.

3. KDnuggets

Twitter Handle: @kdnuggets

KDnuggets is the official Twitter account of the AI, Machine Learning, and Data Science website, KDnuggets.com. While the account mostly promotes content on their website, it’s a great way to easily access AI and ML tutorials, guides, and blog posts.

Follow here.

4. AssemblyAI

Twitter Handle: @AssemblyAI

@AssemblyAI is the official Twitter account for an industry-leading Speech-to-Text API and Audio Intelligence API making a name for itself in the speech recognition startup field. Developers and AI enthusiasts will find links to YouTube deep dives and tutorials, in-depth blog posts, and other content related to Python, Natural Language Processing (NLP), AI, and more.

Follow here.

5. OpenAI

Twitter Handle: @OpenAI

OpenAI is a research and deployment company dedicated to ensuring AI research equality and equity. On the company’s Twitter feed, you’ll find research announcements, demos, informational videos, and more.

Follow here.

6. Fei-Fei Li

Twitter Handle: @drfeifei

Dr. Fei-Fei Li is a professor at Stanford University, Co-Director of Stanford HAI, Co-Director and Chair of the non-profit AI4ALL, and top researcher in AI, computer vision, and machine learning. Her tweets include updates of her AI research, company events and news, and field commentary.

Follow here.

7. Data Science Central

Twitter Handle: @Datasciencecectrl

The Data Science Central Twitter feed shares new publications and research related to data science, AI, machine learning, IoT, deep learning, and more. It’s a great account to follow to stay on top of general news and discoveries throughout this field.

Follow here.

8. Andrew Ng

Twitter Handle: @AndrewYNg

Andrew Ng is the co-founder of popular online learning platform Coursera, a Stanford Computer Science faculty member, and the former head of Baidu AI Group and Google Brain. While there is promotional content throughout the feed, Twitter users will find his perspective on current AI research and trends worthwhile.

Follow here.

9. Jana Eggers

Twitter Handle: @jeggers

Jana Eggers is the CEO of Nara Logics, a Synaptic Intelligence platform, and self-described “applied ai nerd.” Eggers reshares AI-related news stories with her own added commentary as well as intriguing AI research, commentary, and more.

Follow here.

10. Demis Hassabis

Twitter Handle: @demishassabis

Demis Hassabis is the founder and CEO of DeepMind and Isomorphic Labs. He typically retweets AI-related research papers, publications, and AI business content.

Follow here.

11. Thang Luong

Twitter Handle: @lmthang

Thang Luong is a staff research scientist at Google Brain and co-founder of the non-profit Viet AI. Luong’s Twitter feed is filled with his take on new AI research papers, making him a great person to follow for more technical and applied content. He also often includes Natural Language Processing, or NLP, content.

Follow here.

12. Thomas Wolf

Twitter Handle: @Thom_Wolf

Thomas Wolf is the co-founder and Chief Science Officer at HuggingFace. On Wolf’s account, you’ll find AI-related musings, research papers, tutorials, and Natural Language Processing-related content like Sentiment Analysis, speech transcription, and more.

Follow here.

8 Best AI Conferences to Attend in 2022

Kelsey Foster — Tue, 01 Mar 2022 18:46:03 +0000

While many Artificial Intelligence (AI) and Machine Learning conferences were paused in 2021, those looking to attend one in 2022 are in luck -- in-person conferences are back! However, many conferences will also continue to offer virtual options as well for those not wishing to venture back out quite yet or for those who like the ease of participating from their living room.

AI conferences for 2022 range from the most technical to business-focused to academic. Many cover the full spectrum, with specific tracks and workshops designed for those looking to specialize.

Topics will include general AI, strategy, privacy, Deep Learning, Machine Learning, data analytics, data science, big data, natural language processing (NLP) like Speech-to-Text, robotics, risk management, surveillance, and more.

Here’s the full list of top AI conferences to attend in 2022:

1. Gartner Data & Analytics Summit

The Gartner Data & Analytics Summit takes place in March in Orlando, Florida. Top leaders in data analytics and AI will gather to discuss business strategy, technology, and leadership in an attempt to discern how to deliver value in an ‘uncertain’ world. The conferences’ seven tracks ensure that there are sessions for all interests.

When: March 17-22, 2022
Where: Orlando, FL
Pricing: $3,250+
Notable Speakers:

Pieter den Hamer, Sr. Director Analyst, Gartner

Donald Fienberg, Distinguished VP Analyst, Gartner

Gareth Hershel, VP Analyst, Gartner

Sally Parker, Sr. Director Analyst, Gartner

2. World Artificial Intelligence Cannes Festival (WAICF)

The World Artificial Intelligence Cannes Festival, or WAICF, is a hybrid conference and exhibition with additional workshops and demonstrations taking place in February in Cannes, France. The conference aims to explore AI topics related to society, technology, strategy, API privacy, and applications. Attend workshops run by big names like Meta and IBM or listen to world class speakers in the numerous conference sessions.

When: February 10-12, 2022
Where: Cannes, France and Digital
Pricing: 440 Euros+ (Full Pass), also free tier
Notable Speakers:

Darlington Akogo, Founder/Executive Director, Minohealth AI Labs

Beena Ammanath, Executive Director, Global Deloitte AI Institute

Fabian Aufrechter, Head, Havas Sovereign Technologies

Elias Baltassis, Partner/Director, Data & Analytics, Boston Consulting Group

3. Re-work Deep Learning Summit

The Deep Learning Summit, sponsored by Re-Work, brings together industry experts and academic researchers to explore the latest advancements in the fields of Deep Learning and Artificial Intelligence. Though held in person in San Francisco, the conference also has a fully virtual option for those not wishing to travel.

When: February 17–18, 2022
Where: San Francisco, CA (or virtual)
Pricing: $795-$1,695 ($199 for virtual ticket)
Notable Speakers:

Maithra Raghu, Sr. Research Scientist, Google Brain

Lex Fridman, AI Researcher, MIT

Richard Socher, Founder, you.com

Ilya Echstein, ML/AI Researcher, Google

4. Data Innovations Summit

The Data Innovations Summit in Dubai brings together experts in the fields of data management, advanced analytics, and AI. Sessions cater to both the technical and business minds alike in an attempt to explore innovation in the field of AI. If in-person attendance is out of the question, consider the virtual option.

When: March 7th, 2022
Where: Habtoor Grand Resort, Autograph Collection, Dubai, or virtual
Pricing: 189-399 Euros
Notable Speakers:

Henrik Gothberg, Interim Global Insight and Data Backbone Owner, Scania Financial Services and Founding SEO, Scania and Dairdux

Abel Aboh, Data Management Lead, Bank of England

Max Metral, Senior Analytics Manager

F1 David Dadoun, CDO, BRP

5. Big Data & Analytics Summit Canada 2022

The 2022 Big Data & Analytics Summit will offer a hybrid in-person and virtual experience with exhibit booths, on-demand speaking sessions, virtual networking, and more. Attend to forge relationships and discuss innovation in the field of data analytics.

When: April 5-6, 2022
Where: Toronto, Canada
Pricing: TBD
Notable Speakers (Past):

Dr. Satyam Priyadarshy, Chief Data Scientist, Halliburton

Chapin Flynn, SVP, Senior Data Advisor

Advait Bopardikar, Canadian ML Practice Lead, Google

Nastaran Bisheban, Chief Technology Officer, KFC

6. AI in RegTech Summit

Sponsored by Re-work, the AI in RegTech Summit offers three tracks with multiple stages, round table discussions, Q&As, 1:1 meetings, technical labs, practical workshops, and more. Topics to be covered include Deep Learning, Automated Speech Recognition (ASR), regulation, big data, data security, KYC initiatives, risk management, regulatory reporting, market surveillance, and more.

When: April 21-22, 2022
Where: New York
Pricing: $795+ ($149 for virtual)
Notable Speakers:

Kishore Karra, Executive Director, Model Review Group, J.P. Morgan

Glenn Fung, Chief Machine Learning Research and Innovation Scientist, American Insurance Family

Ruchi Sharma, VP, Group Strategic Analytics, Deutsche Bank

Tyler Mey, Director, Data Science and Artificial Intelligence, Mastercard

7. World Summit AI Americas

In its fourth iteration, the World Summit AI Americas conference combines the full AI ecosystem--startups, academics, investors, business leaders, and tech companies in an interactive conference format. The conference will feature more than 150 speakers via two content tracks over its two day span.

When: May 4-5, 2022
Where: Montreal, Canada
Pricing: $449 CAD+ ($199 student)
Notable Speakers (Past):

Yoshua Bengio, Professor of Computer Science and Operations Research, University of Montreal

Professor Daphne Koller, Founder and CEO, Insitro, Co-Founder, Coursera

Marc G Bellemare, Research Scientist, Google Brain Hassan

Sawaf, Director, Artificial Intelligence, Amazon Web Services

8. The AI Summit London

The 2022 AI Summit in London will feature 500 hours of content, 300+ expert speakers, 100+ solutions providers, and 13 stages. Attendees can choose from tracks across AI finance, AI healthcare, CX, Marketing and eCommerce, NLP topics like Sentiment Analysis, Quantum Computing, AI Manufacturing, Cyber Security, and Computer Vision.

When: June 15-16, 2022
Where: London, UK
Pricing: TBD
Notable Speakers: TBD

5 Best Open Source Libraries and APIs for Speaker Diarization

Kelsey Foster — Thu, 10 Feb 2022 16:33:06 +0000

In Automatic Speech Recognition, or ASR, Speaker Diarization refers to labeling speech segments in an audio or video file transcription with corresponding speaker identities. It's also sometimes referred to as Speaker Labels, and at its most basic form, helps answer the question: who spoke when?

In order to accurately predict a speaker, a Speaker Diarization model must perform two actions:

Determine the number of speakers that can be found in an audio or video file.
Attribute each speaker to their appropriate speech segment.

Why is Speaker Diarization useful? When reading a transcription text with multiple speakers, your mind automatically tries to sort the text by speaker–Speaker Diarization does this automatically to make a transcript much more readable.

In addition, organizations like call centers might use Speaker Diarization to automatically label an “agent” or a “customer” in a transcription text for a help hotline. Medical professionals might use Speaker Diarization to automatically label “doctor” and “patient” in the transcription text for a virtual appointment and attach this transcript to a patient file.

While this can seem like a complicated task, today’s best open source libraries and APIs for Speaker Diarization are trained using the latest Deep Learning and Machine learning research, making the process much simpler than it was in the past.

This article looks at the five best open source libraries and APIs available today to perform Speaker Diarization:

1. Kaldi

Kaldi ASR is a well-known open source Speech Recognition platform. To use its Speaker Diarization library, you’ll need to either download their PLDA backend or pre-trained X-Vectors, or train your own models.

Familiar with Kaldi but need help getting Speaker Diarization set up? This tutorial can help. If you’ve never used Kaldi ASR before, this Kaldi Speech Recognition Tutorial for Beginners is a great jumping off point.

2. AssemblyAI

AssemblyAI’s Speech-to-Text and Audio Intelligence APIs offer accurate Speech Recognition without the need to pre-train a model. To perform Speaker Diarization, you can sign up for a free account for its Core Transcription API, though note that there is a limit to how much you can transcribe per month before having to upgrade your account to a paid option.

The API’s detailed, easy-to-follow documentation library can help you get started as well.

3. PyAnnote

Similar to Kaldi ASR, PyAnnote is another open source Speaker Diarization toolkit, written in Python and built based on the PyTorch Machine Learning framework.

For optimal use, you will need to train PyAnnote’s end-to-end neural building blocks to tailor your Speaker Diarization model, though some pre-trained models are also available.

4. Google Speech-to-Text

Google Speech-to-Text is a popular Speech Recognition API that also offers Speaker Diarization. The API has good accuracy and language support, though using it to transcribe a large volume of files can be quite pricey.

You’ll need to enable Speaker Diarization when you are transcribing an audio or video stream either via a file or the Google Cloud Storage Bucket. This documentation can walk you through the necessary steps.

5. AWS Transcribe

AWS Transcribe also offers Speaker Diarization for either batch or real-time transcription. Though AWS Transcribe can also be expensive, the API does offer one hour free per month, so it can be a good option for low volume transcription.

To enable Speaker Diarization, you’ll need to login to your Amazon Transcribe console and create a transcription job. This documentation page can show you how.

Top Use Cases for Sentiment Analysis in NLP

Kelsey Foster — Thu, 20 Jan 2022 22:11:16 +0000

Recent advances in Deep Learning and Artificial Intelligence have pushed forward significant text analytic applications. One of these applications is Sentiment Analysis.

Sentiment Analysis is a subfield of Natural Language Processing, or NLP, and is often used to understand natural language, whether it is written text or transcription text via a Speech-to-Text API.

This post will first explain what Sentiment Analysis is before examining top use cases for Sentiment Analysis in NLP.

What is Sentiment Analysis?

What is Sentiment Analysis? In NLP, Sentiment Analysis is the process of detecting emotions in a specific speech segment from a transcribed audio or video file. It is also sometimes referred to as sentiment mining or extraction.

Deep Learning models are trained to perform Sentiment Analysis using open source data found in Twitter feeds, review sites (like IMDB or Rotten Tomatoes), and talk show or news transcripts.

The most common approach to train a Sentiment Analysis model is by using Sentiment Polarity. For each specified input, the model will output a number on a scale from -1 to 1, where:

-1 = negative
0 = neutral
1 = positive

For example, the model might output the number .89. Since this is closest to 1, the speech segment would be labeled “positive.”

Another training method is multiclass classification, where instead of a sliding scale, the model would output a probability that a speech segment is positive, neutral, or negative. Then, the sentiment with the highest probability would be the one that is classified.

Currently, more research needs to be done in order to expand these limited scales to a more human range of emotions, though some models do already attempt this with varying degrees of success. Once higher accuracy can be attained on this imitation-human scale of emotions, it would of course expand the utility of Sentiment Analysis further.

Sentiment Analysis Use Cases

Still, Sentiment Analysis can be a significant analytical tool for companies to implement.

Let’s look at how some industries are already applying Sentiment Analysis successfully.

Call Centers

Call centers use Sentiment Analysis to extract sentiments in agent and customer interactions across specific products, locations, times of day, and agents. Then, they can analyze trends and make adjustments to agent training, product descriptions, customer relations, and more.

Video Meetings

Video meeting platforms use Sentiment Analysis to extract sentiments in conversations by meeting topic, length, or format.

Podcasts and Broadcast Media

Podcasts and Broadcast Media use Sentiment Analysis to find trends in how listeners react to particular content. Do more positive oriented podcasts produce more positive commentary? Do podcasts or news segments with more negative commentary incite negative commentary, infighting, or hate speech? Advertisers can also use Sentiment Analysis to help better determine which podcasts would be the best fit for sponsorship.

Telemedicine

Telemedicine uses Sentiment Analysis to audit doctor-patient conversations and interactions to ensure positive outcomes. Then, the analyzed data can be used to retrain doctors, rethink delivery methods, or identify other trends.

Top AI Startups to Watch in 2022

Kelsey Foster — Fri, 14 Jan 2022 17:33:38 +0000

Artificial Intelligence research has made massive strides in the past few years. From intelligent assistants to self-driving vehicles to curated show recommendations, AI-backed technology is everywhere.

However, this list showcases the top AI startups that are going beyond everyday applications to do something truly groundbreaking.

Often backed by big name investors and run by impressive teams of Ph.Ds, MDs, engineers, deep learning experts, and data scientists, these startups are capitalizing on advances in AI to design products that make all aspects of human life more efficient–and intelligent.

Together, they make up an impressive range of new AI companies–and are showing us all just what can be accomplished when innovation and technology meet.

Here are the Top AI Startups to Watch:

1. Private AI

Industry: Privacy
Location: Toronto, Canada

Private AI is leading the way in privacy and security software. Its service helps companies redact Personally Identifiable Information and replace them with contextual categories. For example, a social security number is replaced with the term ‘SSN’. No third party processing is required, meaning your sensitive data is safe. Other key features include detecting and redacting 50+ entities at 99%+ accuracy, fast processing time, and 100% private and secure data processing.

Try Private AI for free here.

2. AssemblyAI

Industry: Speech Recognition
Location: Fully Remote

AssemblyAI is pioneering accurate and accessible speech recognition powered by cutting edge Deep Learning, Machine Learning, and AI research. Its Speech-to-Text API transcribes audio and video files and live audio streams with industry-best accuracy. In addition, the company offers Audio Intelligence APIs that secure higher ROI for users, including Sentiment Analysis, Topic Detection, Content Moderation, Auto Chapters, Entity Detection, and more. AssemblyAI focuses on delivering a world class developer experience through complete documentation of all features in its docs and a dedicated team of developers to answer any questions about the API.

Try AssemblyAI for free here.

3. LabelBox

Industry: Data Management
Location: Fully Remote

Labelbox is a training data platform that optimizes the training data iteration loop to make it more efficient. With Labelbox, you can annotate data, diagnose model errors and better understand performance, and prioritize your data. It also helps fully remote teams work more seamlessly when working with training data, facilitating faster progress and collaboration.

Try LabelBox for free here.

4. Deep Genomics

Industry: Biotechnology
Location: Toronto, Canada

Deep Genomics offers a groundbreaking approach to drug discovery and disease management. Its proprietary AI Workbench helps facilitate easier drug development by untangling RNA biology complexity, identifying novel targets, and determining the best therapies to modulate diseases. Its AI Workbench 3.0 is set to be released soon, which will support target and therapy identification for complex, multi-gene diseases.

Read more about Deep Genomics and the AI Workbench here.

5. DarwinAI

Industry: **Intelligent Manufacturing
**Location: Toronto, Canada

Darwin AI provides end-to-end visual quality inspection for top manufacturers like Honeywell, Audi, BMW, and Lockheed Martin. By analyzing a company’s manufacturing process, Darwin AI can recommend hardware and build AI models that identify product defects with greater than human accuracy, reducing overhead and improving KPIs.

Contact Darwin AI here.

6. Hugging Face

Industry: Natural Language Processing
*Location: * New York, New York

Hugging Face is a Natural Language Processing (NLP) startup aiming to democratize AI with an open source community. Find new state-of-the-art models on Hugging Face’s model hub, run large scale NLP models directly from Hugging Face infrastructure, or browse the NLP library, Transformers. More than 5,000 companies already use Hugging Face’s service, including AWS, Facebook AI, Microsoft, Google AI, and more.

Try Hugging Face for free here.

7. Arize AI

Industry: MLOps
Location: Fully Remote

Arize is a machine learning observability platform that helps users improve their models by automatically discovering issues and diagnosing problems. Key features include a simple integration for any model regardless of platform or environment, pre-launch validation, an automatic AI monitoring system, dynamic troubleshooting, and more.

Request a trial of Arize here.

8. Cresta

Industry: Conversational AI
Location: San Francisco, CA

Cresta’s Expertise AI uses AI to provide real-time agent assistance, automated coaching, and business insights for customer-agent conversations. Its solution improves customer experience, drives better relationship building, and increases sales. Each customer-agent interaction is analyzed to ensure top performance from every agent. Cresta can also integrate with most CRMs and knowledge bases.

Request a demo of Cresta here.

DEV Community: Kelsey Foster

Best 5 PII Redaction APIs for 2022-2023

1. AssemblyAI

2. Private AI

3. Amazon Transcribe

4. Azure

5. Super.ai

5 Best APIs for Performing Text Summarization for NLP

Here are the 5 Best APIs for Performing Text Summarization for NLP:

1. AssemblyAI’s Auto Chapters API

2. NLP Cloud Summarization API

3. Microsoft Azure Text Summarization

4. plnia’s Text Summarization API

5. MeaningCloud’s Automatic Summarization

Top 5 Resources to Read and Watch about DALL-E 2

1. How Does DALL-E 2 Actually Work? YouTube Video

2. How DALL-E 2 Actually Works Blog Post

3. How DALL-E 2 Could Solve Major Computer Vision Challenges

4. OpenAI DALL-E: Creating Images from Text YouTube Video

5. How is It so Good? (DALL-E Explained Pt. 2)

Top 9 ML, AI, and Data Science Internships in 2022

Here are the Top 9 ML, AI, and Data Science Internships to consider for 2022:

1. Data Science Intern, Meta

2. Machine Learning Intern, Motional

3. Deep Learning and more, Borealis AI

4. Machine Learning/AI Internship, Apple

5. Research Intern, Applied Research - Natural Language Processing, Meta

6. Autonomy - Deep Learning Infrastructure Intern, Tesla

7. Artificial Intelligence Internship, Intel

8. Research Internship, Allen Institute for AI

9. Internship, Center for Artificial Intelligence in Medicine and Imaging, Stanford University

Top Transcription APIs and Open Source Libraries in 2022

Top Transcription APIs

1. AssemblyAI

2. Google Speech-to-Text

3. AWS Transcribe

Top 3 Open Source Transcription Libraries

1. Wav2Letter

2. DeepSpeech

3. Kaldi

What Is Data Bias and How to Avoid It

1. Collect data from a variety of sources

2. Make sure data is diverse

3. Monitor real-world performance

Top 12 Twitter Feeds to Learn About AI

Here are Top Twitter Feeds for AI:

1. Andreas Miller

2. Lisha.eth

3. KDnuggets

4. AssemblyAI

5. OpenAI

6. Fei-Fei Li

7. Data Science Central

8. Andrew Ng

9. Jana Eggers

10. Demis Hassabis

11. Thang Luong

12. Thomas Wolf

Top 8 Machine Learning Content Creators on YouTube

1. So you want to be a data scientist?

2. Python Engineer

3. Sentdex

4. AssemblyAI

5. Two Minute Papers

6. Tech with Tim

7. Machine Learning with Phil

8. Lex Fridman

8 Best AI Conferences to Attend in 2022

Here’s the full list of top AI conferences to attend in 2022:

1. Gartner Data & Analytics Summit

2. World Artificial Intelligence Cannes Festival (WAICF)

3. Re-work Deep Learning Summit

4. Data Innovations Summit

5. Big Data & Analytics Summit Canada 2022

6. AI in RegTech Summit

7. World Summit AI Americas

8. The AI Summit London

5 Best Open Source Libraries and APIs for Speaker Diarization

1. Kaldi

2. AssemblyAI