Natural language processing (NLP) techniques have been widely applied in the requirements engineering (RE) field to support tasks such as classification and ambiguity detection. Although RE research is rooted in empirical investigation, it has paid limited attention to replicating NLP for RE (NLP4RE) studies. The rapidly advancing realm of NLP is creating new opportunities for efficient, machine-assisted workflows, which can bring new perspectives and results to the forefront. Thus, we replicate and extend a previous NLP4RE study (baseline), "Classifying User Requirements from Online Feedback in Small Dataset Environments using Deep Learning", which evaluated different deep learning models for requirement classification from user reviews. We reproduced the original results using publicly released source code, thereby helping to strengthen the external validity of the baseline study. We then extended the setup by evaluating model performance on an external dataset and comparing results to a GPT-4o zero-shot classifier. Furthermore, we prepared the replication study ID-card for the baseline study, important for evaluating replication readiness. Results showed diverse reproducibility levels across different models, with Naive Bayes demonstrating perfect reproducibility. In contrast, BERT and other models showed mixed results. Our findings revealed that baseline deep learning models, BERT and ELMo, exhibited good generalization capabilities on an external dataset, and GPT-4o showed performance comparable to traditional baseline machine learning models. Additionally, our assessment confirmed the baseline study's replication readiness; however missing environment setup files would have further enhanced readiness. We include this missing information in our replication package and provide the replication study ID-card for our study to further encourage and support the replication of our study.

Today's work in artificial intelligence is amazing. We've taught computers to beat the most advanced players in the most complex games. We've taught them to drive cars and create photo-realistic videos and images of people. They can re-create works of fine-art and emulate the best writers. Yet I know that many businesses still need people to, e.g., read PDF documents about an office building and write down the sizes of the leasable units contained therein.

In early 2017, Alex Borek of Volkswagen convinced me that "this time, machine learning is real" and that data quality was a real problem. So I dug in--researched what we knew, talked to a lot of people, and thought through the various ways that bad data could do harm. And it struck me--this is really scary. As I use that metaphor, I find that practically everyone agrees! The next steps were to sort out what to do and write a straightforward article.

Like Google and Amazon before it, Apple has been caught sending voice assistant recordings to contractors, who listen to snippets of your requests and conversations, without telling anyone. In response to the privacy concerns that raises, Apple says it will eventually give users control over whether their Siri data gets sent to third-party eavesdroppers, but it's unclear whether that consent will be opt-in or opt-out. Letting people opt out of data collection is better than not giving them any choice at all. But for decades, that's been the extent of the conversation. It gives too many giant tech companies plausible deniability for the rampant hoovering of your personal information, and allows them to implicitly blame the victim when they overreach: Don't get angry at us, you could have opted out this whole time.

CyberPoint has seen great success in using supervised machine learning for malware detection. A while back, however, some colleagues and I set out to investigate whether we could make any interesting discoveries by applying unsupervised learning to CyberPoint's malware dataset. In supervised learning, one has a set of samples, each with an assigned label. In the field of malware analysis, a sample would typically be a file, and its label might be either benign or the malware family to which it belongs. The goal is: given a new sample, correctly predict its label.

Poor data quality is enemy number one to the widespread, profitable use of machine learning. While the caustic observation, "garbage-in, garbage-out" has plagued analytics and decision-making for generations, it carries a special warning for machine learning. The quality demands of machine learning are steep, and bad data can rear its ugly head twice -- first in the historical data used to train the predictive model and second in the new data used by that model to make future decisions. To properly train a predictive model, historical data must meet exceptionally broad and high quality standards. First, the data must be right: It must be correct, properly labeled, de-deduped, and so forth.

It's a widely accepted fact--at least in my circle of recent college grads--that Tinder, and the world of online dating writ large, is a wasteland. The lure of carefully curated profiles, relative anonymity, and endless swiping seem to bring out the worst behavior--catfishing, terrible bios, and misogyny abound. And even when motives are pure, the nagging idea that there's a better match one swipe away tends to make Tinder interactions feel like placeholders. This, combined with the fact that men swipe right on anything that moves (meaning women are forced to second-guess every single one of their matches), makes the entire situation feel like a huge time suck with little chance of romantic reward. An Oxford study revealed that about half of conversations between matches were one-sided, meaning the other person never responded.

There are conflicting opinions about the impact emerging technologies will have on the future job market. The reality is this: many of the jobs that will exist in 2030 do not exist today. Our jobs are not at risk of disappearing, but rather evolving and ultimately enabling us to maximize the impact of the human characteristics that make us so valuable: empathy, emotion and sociability. There is no debate over whether artificial intelligence will change the workplace – it already is – but whether or not companies will have the capacity to successfully integrate AI in a way that empowers humans to work faster, better and smarter. New tech opens up countless exciting possibilities, as long as we are able to anticipate them and adapt accordingly.

Hanson Lenyoun is the head of health at Mark One. Data is hot right now. We generate tons of it, but most of it sits there, latent, unused and useless. This is particularly pronounced when it comes to health and fitness data, where we strap on our fitness trackers and expect the pounds to melt away with each step passively logged. But we haven't seen a dramatic improvement in our nation's health with the emergence of the "quantified self movement" and the pervasiveness of wearables. We still live in a country where two-thirds of us are overweight or obese and 80 percent of adults do not get the recommended amount of exercise.

The law firm BakerHostetler officially hired the first artificial intelligence lawyer. IBM makes the robot lawyer, and apparently, it will work in the firm's bankruptcy practice -- because what better way to test out a robot lawyer than by putting people with their lives crumbling around them in its virtual hands. Look, once I watched the entirety of "I, Robot" without sound on another guy's screen on a flight, so, yeah, you could say I already know a lot about artificial intelligence. But not even I, with my extensive experience in the subject, could predict robotics would have already developed so far, a law firm would actually hire an artificially intelligent lawyer. IBM named this lawyer robot "ROSS" because, even though he is a robot, he is first and foremost a lawyer and, therefore, very boring.