USC & Amazon ‘SLADE’ Self-Training Framework Uses Unlabelled Data to Improve Information …

USC & Amazon ‘SLADE’ Self-Training Framework Uses Unlabelled Data to Improve Information …

Let’s face it: the advanced AI systems of the future are unlikely to be built with data that’s been hand-labelled by humans. Finding ways to eliminate the time-consuming data-labelling process has been a challenge in the machine learning (ML) community for some time. A new paper from researchers at the University of Southern California and Amazon approaches the problem with a self-training framework that can improve information retrieval performance using unlabelled data.

Recent methods in self-supervised learning and self-training have shown promising results using unlabelled data. Existing methods for self-supervised learning or self-training however mainly focus on classification but not retrieval — the process of identifying and obtaining relevant information system resources such as texts, images, etc. The researchers’ proposed SeLf-trAining framework for Distance mEtric learning (SLADE) framework combines self-supervised learning and distance metric learning methods to improve information retrieval performance.

Distance metric learning is a research area where the basic objective is to push similar samples closer to each other and different samples away from each other. It can automatically construct task-specific distance metrics from (weakly) supervised data, and the learned distance metrics can then be used to perform various tasks.

The researchers first trained a teacher model on labelled data and used self-supervised representation learning to initialize the teacher model. Once the teacher model was pretrained and fine-tuned, they used it to generate pseudo labels for unlabelled data. They then trained a student model on both labels and pseudo labels to generate final feature embeddings.

To deal with the noisy pseudo labels generated by the teacher network, the researchers designed a new feature basis learning component for the student network. This comprises a new representation layer added after the embedding layer, which is only used for learning basis functions for the feature representation in unlabelled data.

The researchers say their learned basis vectors better measure the pairwise similarity for unlabelled data and can select high-confident samples for training the student network, which can then be used to extract embeddings of query images for retrieval.

The researchers evaluated SLADE on several standard retrieval benchmarks including CUB-200, Cars-196 and In-shop. The results show that the proposed approach can outperform SOTA methods on CUB-200 and Cars-196, is competitive on In-shop, and significantly boosts the performance of fully-supervised methods.

The paper SLADE: A Self-Training Framework For Distance Metric Learning is on arXiv.

Reporter: Yuan Yuan | Editor: Michael Sarazen

Synced Report | A Survey of China’s Artificial Intelligence Solutions in Response to the COVID-19 Pandemic — 87 Case Studies from 700+ AI Vendors

This report offers a look at how China has leveraged artificial intelligence technologies in the battle against COVID-19. It is also available on Amazon Kindle. Along with this report, we also introduced a database covering additional 1428 artificial intelligence solutions from 12 pandemic scenarios.

Click here to find more reports from us.

We know you don’t want to miss any news or research breakthroughs. Subscribe to our popular newsletter Synced Global AI Weekly to get weekly AI updates.

Published at Wed, 25 Nov 2020 20:15:00 +0000

Technological Innovation in Law: The Recent Trends in Legal Firm Adoption

Technology is advancing in every industry on a global scale even hitting law firms. Whether its digitalisation or even the adoption of artificial intelligence, while some embrace innovation with open arms, others struggle to make the change.

Someone who knows all about this is Sascha Mehlhase is the vice president of products and innovation at ABC Legal Services. Sascha oversees ABC Legal’s growing product, marketing and customer experience teams in transforming ABC Legal into the best-in-class technology and service industry leader, while simultaneously finding new avenues to scale. With nearly 20 years of product and marketing experience in software and technology, Sascha has advanced product strategies and led global teams in a variety of industries. 

Here he discusses the ongoing impact technological adoption is having on the legal industry.

There’s often some resistance towards technological innovation at law firms from partners that are set in their ways. It’s been especially difficult to convince attorneys that didn’t grow up with digital technologies to work without hard copies, much less to consider trusting their clients’ cases to artificial intelligence. The Covid-19 pandemic has forced legal professionals to adopt technology at a rapid pace, making it more palatable to even the most entrenched Luddites.

Resistance to Technology from Law Firms and Courts

From the hallowed chambers of British criminal courts to the overcrowded marble pillars and stately staircases of the New York County Supreme Court, technology has taken a back seat to tradition. Enhancements such as e-filing have been phased in at a snail’s pace, but requirements for in-person appearance and hard copies of documents remained a fixture in many courtrooms. Some partners who are more resistant to new automation tools consider technology an expensive superfluous tool that halts operations for installation and reduces productivity with the required learning curve. The resistance by this group of legal professionals has made it difficult for the broader industry to embrace the benefits legal technologies can provide.

 Covid-19 Pandemic Ushers in The Immediate Need to Implement Technology

By the end of March 2020, law firms and courthouses around the world were put on hold. Even the most successful law firms had to alter operations and had glimpses of the unknown for the future. It was the firms that had the foresight to implement paperless files, e-conferencing, and e-signing technology up to speed that were able to return to operations the quickest. Judges, clerks, and other court personnel received crash courses in audio and video conferencing tools and operating their caseloads remotely.

Law firm partners and employees quickly learned how to consult with clients via Zoom from home. Further, with children home from school, many court and law firm personnel were unable to leave their homes, even when safety precautions could be implemented at their workplaces. Embracing legal technology was not an option for these individuals if they wanted to safely resume work.

Law Firm Managers Now Face Less Resistance to Implementing Technology

Now that established lawyers have been somewhat forced to learn new approaches to traditional legal processes, senior partners are much less resistant to adopt new technologies into their permanent workflows. Law firm managers are finding that they now have the attention and credibility when they suggest implementing artificial intelligence (AI), natural language processing (NLP), machine learning (ML), as well as e-discovery at their firm. Many law firms are now looking to be leaders in adopting new technology into existing processes, rather than risking once again being caught on their heels.

Law Firms and Artificial Intelligence, Machine Learning and Natural Language Processing

AI can be extremely helpful behind the scenes improving the efficiency of most large corporate firms, but certainly not as a substitute for human beings. Instead, legal professionals are leveraging AI to supplement existing processes and automate repetitive tasks such as drafting NDAs.

ML enables AI through its ability to identify and analyse large amounts of information from court records and cases. The technology then recognises patterns and flags them to attorneys and paralegals, reducing the need to read through hundreds of pages. This can positively impact the client’s experience because their questions can be answered by legal professionals more quickly and accurately. This level of automation and added intelligence also helps legal teams by increasing efficiency and productivity, allowing more time for other billable tasks.

Despite the success of these technologies, there is still extreme value in human interactions and support, especially as more of the industry becomes comfortable with new automated processes. For example, even the most technologically advanced firms have maintained their administrative staff during the global pandemic to answer and route calls from customers and clients to the appropriate firm personnel. Firms can still adopt intelligent tools, while maintaining meaningful connections with their networks.

Advances in E-Filing and the Courts

Most courts have begun implementing some level of automation through e-filing; however, other courts have had types of cases that were not included in digital advancements or had some functions that still required in-person visits or hard copies to be completed. Despite past hesitations to implement e-filling and unique case requirements aside, the global pandemic accelerated the widespread need and desire to embrace digital transformation and improve ease of use by completing court cases through automated processes.

As demand increases, these tools will also see improvements. E-filing for example, continues to evolve to be able to automate cases not previously acceptable for transformation. For example, matrimonial cases that were not subject to e-filing due to privacy concerns, are now being automated with the use of encryption. These types of e-filling applications reduce the need for in-person appearances at the courthouse to file and obtain papers.

After seeing the impacts automation tools have delivered in just several months, courts around the world are now prioritising widespread adoption of these technologies to improve efficiency and experiences for all parties involved. The implementation of technology to power legal process automation is just getting started.

Published at Wed, 25 Nov 2020 20:03:45 +0000