During SpeechTEK, which was held in late April in Washington, D.C., hundreds of customer care professionals gathered to wrestle with the shifting sand that is customer care, particularly the challenges and opportunities presented by automation in the form of virtual assistants or chatbots. The conference provided insights that included answers, objections, and more questions. Companies spend massively on customer care and automation will reduce costs. In Tuesday's keynote panel, Cognitive Code's Chief Operating Officer (COO) Brian Garr said industry observers estimate customer care spending to be nearly $500 billion. Everest Group estimated in 2016 that call center spending was $300 to $320 billion annually.

The adage "everyone's a critic" has never been more true than it is today. Long gone are the days of needing to rely solely on surveys to figure out how consumers feel about your product or business. Now -- whether it's through social media, online reviews, customer service emails, call centers or chat bots -- there is an overflow of customer feedback available for every type of business, content, or product out there. This is a problem technology helped to create, where we are asked to give everything a review, but thankfully, technology is now helping businesses sift through and understand all of that data floating around. According to a report last year by IDC, organizations that analyze all relevant data and deliver actionable information will achieve an extra $430 billion in productivity gains over their less analytically-oriented peers by 2020. Additionally, Forrester's Artificial Intelligence predictions for 2017 estimated that companies that are truly "insights-driven businesses" will steal $1.2 trillion per annum from their less-informed peers by 2020.

Then we add some world knowledge, in the form of commonsense statements, to help in the text understanding. The result combines this knowledge to form a multidimensional space where concepts, people, groups, and projects are all represented as vectors. From that space we retrieve information relevant to lab visitors--dynamically creating their presence in the vector space by creating a vector from the projects they have chosen as favorites. We then use the vector space to determine the relevance of objects in the space to each other--determining which projects are similar, which projects would be good fits for a lab visitor, and which projects fit which lab themes. Additionally, we have designed a user interface that makes this system easy and social to interact with. The following subections discuss our approach to interface design, our methods for extracting semantic information from the text base, and for assessing similarity of user interests with that knowledge.

We propose and give a preliminary test of a new metric for the quality of the commonsense knowledge and reasoning of large AI databases: Using the same measurement as is used for a four-year-old, namely, an IQ test for young children. We report on results obtained us- ing test questions we wrote in the spirit of the questions of the Wechsler Preschool and Primary Scale of Intelligence, Third Edition (WPPSI-III) on the ConceptNet system, which were, on the whole, quite strong.

Web 2.0 has changed the ways people communicate, collaborate, and express their opinions and sentiments. But despite social data on the Web being perfectly suitable for human consumption, they remain hardly accessible to machines. To bridge the cognitive and affective gap between word-level natural language data and the concept-level sentiments conveyed by them, we developed SenticNet 2, a publicly available semantic and affective resource for opinion mining and sentiment analysis. SenticNet 2 is built by means of sentic computing, a new paradigm that exploits both AI and Semantic Web techniques to better recognize, interpret, and process natural language opinions. By providing the semantics and sentics (that is, the cognitive and affective information) associated with over 14,000 concepts, SenticNet 2 represents one of the most comprehensive semantic resources for the development of affect-sensitive applications in fields such as social data mining, multimodal affective HCI, and social media marketing.

Most organizations have a wealth of knowledge about themselves available online, but little for a visitor to interact with on-site. At the MIT Media Lab, we have designed and deployed a novel intelligent signage system, the Glass Infrastructure (GI) that enables small groups of users to physically interact with this data and to discover the latent connections between people, projects, and ideas. The displays are built on an adaptive, unsupervised model of the organization developed using dimensionality reduction and common sense knowledge which automatically classifies and organizes the information. The GI is currently in daily use at the lab. We discuss the AI model’s development, the integration of AI into an HCI interface, and the use of the GI during the lab’s peak visitor periods. We show that the GI is used repeatedly by lab visitors and provides a window into the workings of the organization.

The fact that birds have feathers and ice is cold seems trivially true. Yet, most machine-readable sources of knowledge either lack such common sense facts entirely or have only limited coverage. Prior work on automated knowledge base construction has largely focused on relations between named entities and on taxonomic knowledge, while disregarding common sense properties. In this paper, we show how to gather large amounts of common sense facts from Web n-gram data, using seeds from the ConceptNet collection. Our novel contributions include scalable methods for tapping onto Web-scale data and a new scoring model to determine which patterns and facts are most reliable. The experimental results show that this approach extends ConceptNet by many orders of magnitude at comparable levels of precision.

Coarse word sense disambiguation (WSD) is an NLP task that is both important and practical: it aims to distinguish senses of a word that have very different meanings, while avoiding the complexity that comes from trying to finely distinguish every possible word sense. Reasoning techniques that make use of common sense information can help to solve the WSD problem by taking word meaning and context into account. We have created a system for coarse word sense disambiguation using blending, a common sense reasoning technique, to combine information from SemCor, WordNet, ConceptNet and Extended WordNet. Within that space, a correct sense is suggested based on the similarity of the ambiguous word to each of its possible word senses. The general blending-based system performed well at the task, achieving an f-score of 80.8\% on the 2007 SemEval Coarse Word Sense Disambiguation task.

Today millions of web-users express their opinions about many topics through blogs, wikis, fora, chats and social networks. For sectors such as e-commerce and e-tourism, it is very useful to automatically analyze the huge amount of social information available on the Web, but the extremely unstructured nature of these contents makes it a difficult task. SenticNet is a publicly available resource for opinion mining built exploiting AI and Semantic Web techniques. It uses dimensionality reduction to infer the polarity of common sense concepts and hence provide a public resource for mining opinions from natural language text at a semantic, rather than just syntactic, level.

Reasoning about Commonsense knowledge poses many problems that traditional logical inference doesn't handle well. Among these is cross-domain inference: how to draw on multiple independently produced knowledge bases. Since knowledge bases may not have the same vocabulary, level of detail, or accuracy, that inference should be "scruffy." The AnalogySpace technique showed that a factored inference approach is useful for approximate reasoning over noisy knowledge bases like ConceptNet. A straightforward extension of factored inference to multiple datasets, called Blending, has seen productive use for commonsense reasoning. We show that Blending is a kind of Collective Matrix Factorization (CMF): the factorization spreads out the prediction loss between each dataset. We then show that blending additional data causes the singular vectors to rotate between the two domains, which enables cross-domain inference. We show, in a simplified example, that the maximum interaction occurs when the magnitudes (as defined by the largest singular values) of the two matrices are equal, confirming previous empirical conclusions. Finally, we describe and mathematically justify Bridge Blending, which facilitates inference between datasets by specifically adding knowledge that "bridges" between the two, in terms of CMF.