We’ve been doing text analytics since the mid-2000s. What’s working and what’s not? Seth Grimes presents findings from a 2019 study that surveyed user experiences with text technologies, solutions, and providers. What capabilities are users looking for in text analytics products and services, and are they finding what they need? How has the market evolved—both demand and supply—and how should practitioners and solution providers stay on top of developments? This talk provides practical guidance from fellow users to help you extract the greatest value from text analytics.

At Indeed, we analyze petabytes of unstructured text in support of our mission to help people get jobs. We strive to go beyond keyword search and enable users to search by concepts relevant to their job interests, such as occupation, location, or salary range. Our talk focuses on entity extraction: how can it unlock meaning from big data and what are the best strategies for implementing it? Should you build an in-house tool, choose an off-the-shelf tool, or combine both of these approaches? We review lessons learned from implementing each of these strategies at Indeed. This presentation is targeted at practitioners who want to develop scalable information extraction systems and are interested in decision factors such as model performance, startup training costs, operator workflows, ongoing maintenance, and risk mitigating strategies.

Text analytics is a well-established discipline, yet many organizations don’t have dedicated text analyst roles. When they do, they arise from ad hoc needs and are embedded within specific functional units, often doing work which could be repurposed for other processes. One part call to action, one part proposal, hear ideas about who a text analyst is, where this role should sit in the organization, and what actions to take to create a text analyst role in your organization.

Increasingly, knowledge managers and product developers are looking beyond in-house proprietary data to external sources to provide context and critical information. For companies focused on innovation and research, important sources include scientific journal publications, specialized databases, and a growing body of open access data. This talk looks closely at two different customer use cases that take text and data mining (TDM) projects from start to finish. The first is a knowledge manager seeking to enhance content discovery and linking key research findings to in-house data. The second is a company focused on leveraging research content for product development. In both cases, TDM is applied to drive an efficient research and discovery process, saving time and money, and positively impacting revenue.

Text Analytics has emerged as a defining technology for enterprise transformation. Its promise is no less than a radical rethinking of how businesses organize their workflows and take decisions. When cognitive processes can be automated and scaled, the impact is faster processes and better insights, enabling professionals to focus on the highest added-value parts of their mission. Join this session to discover today’s “art of the possible” in NLP, based on examples of leading-edge analytics and process automation applications

A panel of four text analytics experts answer questions that have been gathered before the conference, during the conference and some additional questions from the program chair. This was one of our most popular features last year, so come prepared with your favorite questions and be ready to learn!

Efforts to counter human trafficking internationally must assess data from a variety of sources to determine where best to devote limited resources. How can analysts effectively tap all the relevant data to best inform decisions to counter human trafficking? This presentation showcases a framework supporting AI for exploring all data related to counter human trafficking initiatives internationally. The framework incorporates rule-based and machine learning (ML) text analytics results not available in the original datasets. As a focal point, we demonstrate how to apply rule-based text extraction of trafficking victims to generate training data for subsequent ML and deep learning models. We ultimately show how this framework provides decision makers with capabilities for countering human trafficking internationally, and how it is extensible as new AI techniques and sources of information become available.

As organizations shift focus from data-generating to data-powered, the ability to incorporate all information—structured and unstructured—is key to delivering insight that is trusted by the business. What does an organization need to bring together all information and successfully manage its data quality? Semantic AI provides the context and meaning that transforms textual information into trusted data. It uncovers the insights and relationships using NLP, machine learning, and AI strategies so you can expose new information to the business and answer questions you couldn’t answer before. Using the real-life success story of one of the world’s largest auto manufacturers that uses Semantic AI to harmonize, extract, and enrich data for vehicle safety analysis, this session covers how the use of Semantic AI cleans and calibrates IoT, text-based data, and unstructured content to improve data quality, analytics, and the fidelity of business decisions.

How can we classify audio files of music with very sparsely available text? A large commercial music publisher with a faceted classification system (for 650,000 tracks!) realized that its search problems were caused by incomplete, missing, or misapplied metadata. Additionally, the text associated with each track (or album) was spotty at best. In this talk, Kasenchak describes the variety of text analytical (and other) approaches used to try to solve the problem: adding or correcting metadata to improve search.

Vocabularies and natural language processing (NLP) often work handin- hand to provide text analytics solutions. This talk explores this partnership in detail in the context of a specific knowledge domain: biomedicine. Standard assumptions made by NLP engines regarding how words are stemmed, tokenized, assembled to form phrases and sentences can be challenged by a specialized domain such as biomedicine, which has its own terminology and knowledge models. Biomedical vocabularies and ontologies play an essential role to make biomedical text is being analyzed appropriately. This talk goes into more detail about how the two capabilities—domain-specific vocabularies and ontologies and NLP engines—can learn from each other to deliver better biomedical text analytics solutions.

In some contexts, such as e-discovery, achieving high recall when retrieving documents is critical. Over the past year, Dimm has challenged audiences at several conferences to construct keyword searches that perform better than supervised machine learning. This talk summarizes the results and explains why it is so hard for humans to beat machine learning when seeking high recall.

It's “How the Future Was” in 2004: Microsoft and enterprise search vendors showed semantic search demos that were "so close" to being better than keywords. Fifteen years on, we're still "so close," and most information retrieval searches are still keyword lookups in hash tables. What has happened is that semantic search matured in other directions. We explore multiple use cases and specific applications to government missions and commercial business problems, where semantic search has established these and a few other niches. We give special emphasis to the “analytic refutation problem,” in which both keyword search and much of current AI serves only to assist people find even more content that reinforces their biases and mistaken conceptions. Here semantic search has found its deepest niche: helping human analysts triage otherwise intractable quantities of textual information, maintaining a healthy bias against their working hypotheses.

Combining search with text analytics creates a powerful tool called SAS Cognitive Search to elevate the intelligence of information retrieval. Search features a flexible query syntax to fit various business needs and help uncover insights hidden in data. Text analytics can extract entities from user text data and enrich the raw data with category and sentiment information. This session presents an easy-to-use interface that leverages SAS Cognitive Search to perform search on temporal and spatial data, enriched with NLP features. With this interface, the user can analyze customer reviews for a product category to create a timeline and deduce trends. Other features include an interactive map facilitating geographic data search and filtering and a facet-based view for query results aggregation. Understanding your customers and what they think of your products has never been easier!

Key words and hundreds, if not thousands, of rules are no longer enough to keep up with Amazon and recapture lost market share. Amazon has set the new standard that organizations must meet to effectively compete for their customers’ attention. Join this presentation and learn how to capture and aggregate valuable customer interactions like queries, clicks, and cart behavior in real-time so every customer gets a customized experience that is continuously being refined; easily incorporate regional trends and seasonality to deliver relevant results; make every customer experience personal; and run A/B testing and experiments so the shopping and purchase flow is constantly fine-tuned and optimized. Go from running a few experiments that take months to get out the door to dozens running live in production, without having to bother your data scientists or the engineers.

Similar to many enterprises, the Inter-American Development Bank (IDB) has multiple information sources which are isolated in different systems. There is no link between all these information resources that can make them accessible outside of their native systems. It is not possible to relate distinct kinds of resources that share some characteristics, e.g., to find a course that is about the same topic as a publication. To achieve this objective, IDB implemented a system that can automatically extract entities and concepts from its systems, including structured and unstructured data. Further, it semantically enhanced the data and made it accessible in a Knowledge Graph. Hernandez and Marino share lessons learned from this project that can help interested attendees start with a baseline of best practices for their own projects, saving valuable time and money.

Products like Amazon Alexa and Google Home are changing the expectations as to how search should work. Searchers now expect voice-driven search solutions that provide answers and not just a list of links. Part of this talk shares how knowledge graphs enable a natural language search and how text analytics along with machine learning can be used to populate these powerful constructs. We explain how to architect these solutions and provide real world examples as to how many of our clients have taken advantage of these powerful tools.