Automatic document organization, topic extraction, information retrieval and filtering all have one thing in common. They require text clustering (sometimes also known as document clustering) to be done quickly and accurately. If you've never heard of text clustering, this post will explain what it is, what it does, and how its currently being used to aid businesses. We'll also briefly discuss how a business could employ text clustering too! First, let's define text clustering.

Pseudo-relevance feedback (PRF) is commonly used to boost the performance of traditional information retrieval (IR) models by using top-ranked documents to identify and weight new query terms, thereby reducing the effect of query-document vocabulary mismatches. While neural retrieval models have recently demonstrated strong results for ad-hoc retrieval, combining them with PRF is not straightforward due to incompatibilities between existing PRF approaches and neural architectures. To bridge this gap, we propose an end-to-end neural PRF framework that can be used with existing neural IR models by embedding different neural models as building blocks. Extensive experiments on two standard test collections confirm the effectiveness of the proposed NPRF framework in improving the performance of two state-of-the-art neural IR models.

Usually considered as a classification problem, entity resolution can be very challenging on real data due to the prevalence of dirty values. The state-of-the-art solutions for ER were built on a variety of learning models (most notably deep neural networks), which require lots of accurately labeled training data. Unfortunately, high-quality labeled data usually require expensive manual work, and are therefore not readily available in many real scenarios. In this paper, we propose a novel learning paradigm for ER, called gradual machine learning, which aims to enable effective machine learning without the requirement for manual labeling effort. It begins with some easy instances in a task, which can be automatically labeled by the machine with high accuracy, and then gradually labels more challenging instances based on iterative factor graph inference. In gradual machine learning, the hard instances in a task are gradually labeled in small stages based on the estimated evidential certainty provided by the labeled easier instances. Our extensive experiments on real data have shown that the proposed approach performs considerably better than its unsupervised alternatives, and it is highly competitive with the state-of-the-art supervised techniques. Using ER as a test case, we demonstrate that gradual machine learning is a promising paradigm potentially applicable to other challenging classification tasks requiring extensive labeling effort.

Finding someone in a surveillance video could soon be as easy as Googling them. Descriptions of people of interest, such as a suspect or a missing person, are normally given in terms of their height, gender or clothing. But using this information to find a short woman wearing a red jacket in a video, say, often requires scanning hours of footage manually, which is no easy task. But a new search tool can do it automatically.

If you want to build your blog audience, you're going to have to get smarter with your content. One of the biggest challenges that bloggers and content marketers face is writing content that's optimized for search engines, yet will also appeal to people. According to Copyblogger, SEO is the most misunderstood topic online. But, SEO content isn't complicated, once you understand that people come first, before search algorithms. SEO firms make their money understanding these simple concepts. Thriving in your online business means that you must go beyond simply "writing content." Your content needs to accomplish two goals: first, appeal to the end-user (customers, clients, prospects, readers, etc.) and second, solve a particular problem. But, how do you create content that meets those goals? How do you create content that ranks well with Google and also persuades people? Don't worry if you can't afford an expensive SEO copywriter. You can do this following simple rules. And, that's what you're going to learn in this article. We all know what happens when you type a search query into a search engine and hit "enter": You get a list of search results that are relevant to your search term. Those results pages appear as a result of search engine optimization (SEO).

With Google's use of AI and machine-learning helping it pinpoint, more clearly than ever, the specific factors that satisfy search queries in different niches and contexts, brands and retailers are being encouraged to tightly tailor their search strategies. A new study, "Searchmetrics Google Ranking Factors 2018", reveals, for example, that high-ranking Google results for searches related to the'weight loss' niche are 4x more likely to have a video on the page than results for'financial planning' or'credit' niches. And that eCommerce sites in the'furniture' niche can get away with displaying nearly 28 images on a page (more than most other niches) and still rank highly despite the fact that more images can sometimes make pages load slower. According to Jordan Koene, Chief Evangelist, VP Professional Services, Google's use of sophisticated AI and machine-learning techniques, such as its RankBrain system, help it to better understand the real intention behind the words that searchers enter in the search box – and learn what types of web pages will satisfy individual searches. "Google now recognizes much more clearly if someone's searching online to buy a table, for instance, or needs personal finance advice or wants to learn weight loss exercises. And by tracking user signals such as how often certain results are clicked and how long people spend there, the search engine learns what factors – such as more or less images or text, or whether a site uses encryption to protect personal information entered by visitors are appropriate for satisfying searchers in individual niches."

Neyshabur and Srebro proposed Simple-LSH, which is the state-of-the-art hashing method for maximum inner product search (MIPS) with performance guarantee. We found that the performance of Simple-LSH, in both theory and practice, suffers from long tails in the 2-norm distribution of real datasets. We propose Norm-ranging LSH, which addresses the excessive normalization problem caused by long tails in Simple-LSH by partitioning a dataset into multiple sub-datasets and building a hash index for each sub-dataset independently. We prove that Norm-ranging LSH has lower query time complexity than Simple-LSH. We also show that the idea of partitioning the dataset can improve other hashing based methods for MIPS. To support efficient query processing on the hash indexes of the sub-datasets, a novel similarity metric is formulated. Experiments show that Norm-ranging LSH achieves an order of magnitude speedup over Simple-LSH for the same recall, thus significantly benefiting applications that involve MIPS.

Google is ending a controversial practice in Europe where it requires smartphone makers seeking to pre-install Google's app store to also add other Google apps, such as search and Chrome. Instead, Google will allow device manufacturers to pre-install the Google Play Store on a stand-alone basis, and offer the option to pre-install Google's other proprietary apps for an extra, unspecified fee. The company's announcement Tuesday came ahead of an Oct. 29 deadline to comply with a European Union antitrust decision, which saw regulators slap the company with a $5 billion fine for bundling its apps in an allegedly anticompetitive manner. Google is fighting the order but is working to meet its terms, because not doing so by the deadline could risk further penalties. In making their decision, antitrust officials in Europe had said that Google's practice of tying the apps together could harm competition by giving Google a built-in advantage over new apps struggling to attract an audience.

Google on Monday finally confirmed a secretive project that's been fueling an employee-led backlash for weeks at the company: an effort to build a version of its search engine that complies with China's online censorship regime. The project, code-named Dragonfly, is not only real but is already performing to the satisfaction of top Google executives. And it could pave the way for Google to reenter China's online search market after nearly a decade. "If Google were to operate in China, what would it look like? What queries will we be able to serve?" chief executive Sundar Pichai said during an event hosted by Wired on Monday night.

Google CEO Sundar Pinchai has said a separate, censored version of its search engine for the Chinese market has undergone several successful internal tests. The comments are the first time Google has officially confirmed it is working on the search engine, dubbed Project Dragonfly, which has been criticised heavily by human rights organisations. Pinchai defended the decision of working on a search engine which will censor any results critical of the Chinese government by saying providing some information is better than providing no information at all. "We are compelled by our mission [to] provide information to everyone, and [China is] 20 percent of the world's population," the Google CEO said during the Wired25 conference, as reported by the organiser, Wired. "People don't understand fully, but you're always balancing a set of values," he continued, adding that the company will try to provide information in any market it enters.