In recent years, transformer-based language models have achieved state of the art performance in various NLP benchmarks. These models are able to extract mostly distributional information with some semantics from unstructured text, however it has proven challenging to integrate structured information, such as knowledge graphs into these models. We examine a variety of approaches to integrate structured knowledge into current language models and determine challenges, and possible opportunities to leverage both structured and unstructured information sources. From our survey, we find that there are still opportunities at exploiting adapter-based injections and that it may be possible to further combine various of the explored approaches into one system.

Retrofitting is a technique used to move word vectors closer together or further apart in their space to reflect their relationships in a Knowledge Base (KB). However, retrofitting only works on concepts that are present in that KB. RetroGAN uses a pair of Generative Adversarial Networks (GANs) to learn a one-to-one mapping between concepts and their retrofitted counterparts. It applies that mapping (post-specializes) to handle concepts that do not appear in the original KB in a manner similar to how some natural language systems handle out-of-vocabulary entries. We test our system on three word-similarity benchmarks and a downstream sentence simplification task and achieve the state of the art (CARD-660). Altogether, our results demonstrate our system's effectiveness for out-of-knowledge and rare word generalization.

Crowdsourcing common sense training data was born twenty years ago. It began with the idea to "harness the power of bored people on the Internet" to collect "what everyone knows but no one writes down". This was an era when we were all just starting to learn how to search the web, before people learned the dismal art of keywords, they tried typing their wants and needs. Search engines were woefully unequipped for these kinds of queries, and it was in that climate that we started ConceptNet, which we originally called OpenMind CommonSense (OMCS). Over the years, the effort and its methods evolved to address new applications. Today, we explore how structured common sense is becoming more relevant in NLP and the role it can play in helping solve problems of explainability, scalability, low-resourced languages, domain transfer, and AI bias. We look at how structured common sense and transformers are complementary and how we can combine them to keep the best of both worlds. Additionally, how does common sense need to evolve as we tackle larger and more complex and cross domain problems?

Machine learning about language can be improved by supplying it with specific knowledge and sources of external information. We present here a new version of the linked open data resource ConceptNet that is particularly well suited to be used with modern NLP techniques such as word embeddings. ConceptNet is a knowledge graph that connects words and phrases of natural language with labeled edges. Its knowledge is collected from many sources that include expert-created resources, crowd-sourcing, and games with a purpose. It is designed to represent the general knowledge involved in understanding language, improving natural language applications by allowing the application to better understand the meanings behind the words people use. When ConceptNet is combined with word embeddings acquired from distributional semantics (such as word2vec), it provides applications with understanding that they would not acquire from distributional semantics alone, nor from narrower resources such as WordNet or DBPedia. We demonstrate this with state-of-the-art results on intrinsic evaluations of word relatedness that translate into improvements on applications of word vectors, including solving SAT-style analogies.

The guest editors introduce novel statistical approaches to concept-level sentiment analysis that go beyond a mere syntactic-driven analysis of text and provide semantic-based methods. Such approaches allow a more efficient passage from (unstructured) textual information to (structured) machine-processable data, in potentially any domain.

The Web holds valuable, vast, and unstructured information about public opinion. Here, the history, current use, and future of opinion mining and sentiment analysis are discussed, along with relevant techniques and tools.

The Narratarium Colorizer device receives either keyboard input or speech recognition input and uses natural language processing to extract key terms. The terms are queried for in a knowledge base of words and associated colors, created by leveraging the Open Mind Common Sense database and ConceptNet. The system outputs a continually changing color display, which is projected uniformly throughout the room using a custom designed curved mirror projection system.

The guest editors introduce novel approaches to opinion mining and sentiment analysis that go beyond a mere word-level analysis of text and provide concept-level methods. Such approaches allow a more efficient passage from (unstructured) textual information to (structured) machine-processable data, in potentially any domain.

Languages differ in how they package the components of an event into words to form sentences. For example, while some languages typically encode the manner of motion in the verb (e.g., running), others more often use verbs that encode the path (e.g., ascending). Prior research has demonstrated that children and adults have lexicalization biases; that is, they assume that novel motion verbs will reflect the dominant pattern of their own language. These experiments explored the plasticity of these biases. In Experiments 1 and 2 we taught English-speaking adults motion verbs, varying the proportion of manner and path verbs in the training set; their interpretation of subsequent verbs closely reflected the probabilistic variation in the input. In Experiments 3 and 4, 5-year-old children also systematically shifted their lexicalization biases to reflect the verbs that they were taught. We conclude that lexicalization biases are adaptive inferences about verb meaning that are updated on the basis of experience.

ConceptNet is a knowledge representation project, providing a large semantic graph that describes general human knowledge and how it is expressed in natural language. Here we present the latest iteration, ConceptNet 5, with a focus on its fundamental design decisions and ways to interoperate with it.

ConceptNet is a knowledge representation project, providing a large semantic graph that describes general human knowledge and how it is expressed in natural language. This paper presents the latest iteration, ConceptNet 5, including its fundamental design decisions, ways to use it, and evaluations of its coverage and accuracy.

This editorial introduction describes the aims and scope of the special issue on Common Sense for Interactive Systems of the ACM Transactions on Interactive Intelligent Systems. It explains why the common sense knowledge problem is crucial for both artificial intelligence and human-computer interaction, and it shows how the four articles selected for this issue fit into the theme.

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 interact physically through a touch‐screen display 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 and its relationships developed using dimensionality reduction and commonsense knowledge that 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 a human‐computer interaction (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.

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.

ConceptNet is a knowledge representation project, providing a large semantic graph that describes general human knowledge and how it is expressed in natural language. The scope of ConceptNet includes words and common phrases in any written human language. It provides a large set of background knowledge that a computer application working with natural language text should know. These words and phrases are related through an open domain of predicates, describing not just how words are related by their lexical definitions, but also how they are related through common knowledge. Transforming ConceptNet into a vector space allows reasoning about similarity and relatedness of words with speed and broad coverage[1]. As ConceptNet currently has data in many written languages, we can build a vector space containing concepts from each of our languages enabling cross-language reasoning similar to parallel corpora in translation. Applications of this technology include free text analytics, sentiment analysis, intelligent search and machine reading. More information on ConceptNet including its fundamental design decisions, ways to use it, and evaluations of its coverage and accuracy can be found here[2]. Its content is searchable and downloadable under a Creative Commons license from http://conceptnet5.media.mit.edu. BODY ConceptNet 5 is released. It’s a free semantic network in many languages, enabling reasoning and machine learning about what words mean.

In a world in which millions of people express their opinions about commercial products in blogs, wikis, fora, chats and social networks, the distillation of knowledge from this huge amount of unstructured information can be a key factor for marketers who want to create an image or identity in the minds of their customers for their product, brand or organization. Opinion mining for product positioning, in fact, is getting a more and more popular research field but the extraction of useful information from social media is not a simple task. In this work we merge AI and Semantic Web techniques to extract, encode and represent this unstructured information. In particular, we use Sentic Computing, a multi-disciplinary approach to opinion mining and sentiment analysis, to semantically and affectively analyze text and encode results in a semantic aware format according to different web ontologies. Eventually we represent this information as an interconnected knowledge base which is browsable through a multi-faceted classification website.

Modeling Verb Lexicalization Biases using Hierarchical Bayesian Models Catherine Havasi (havasi@media.mit.edu) MIT Media Lab, 20 Ames Street Cambridge, MA 02139 USA Robert Speer (rspeer@mit.edu) MIT Media Lab, 20 Ames Street Cambridge, MA 02139 USA Abstract a single example (Gentner & Boroditsky, 2001), and they can even learn words for events they are unable to observe (Landau & Gleitman, 1985). Two faster and more noise-resistant strategies have been hypothesized by researchers. One is syntactic bootstrapping (Gleitman, 1990). In this theory, the syntactic frame of the verb is used to constrain hypotheses to those which makes sense in the given frame and are similar to known verbs with similar frames. In the manner/path example given earlier, you would be more likely to think the meaning of the novel verb was related to its motion if you had heard the semantically rich fame “Jesse gorped the frisbee to Edison.” Another hypothesis is that we are able to quickly learn words from few examples because we rely on our learned lexicalization biases about the meanings of words (Gentner & Boroditsky, 2001). Learners select word meanings that align with the features that are dominant in the learner’s native language (Naigles, 1990), indicating that language learners observe general features of the meanings of other words and apply them to new words as well. Modern evidence suggests that children use a combination of these strategies (Papafragou & Selimis, 2010). But how are these biases learned and regulated? In this paper, we explore the possibility that biases for certain components of meaning are associated with language and semantic frame. These biases represent examples of Bayesian overhypotheses about what a word is likely to mean, and these overhypotheses can themselves be learned from examples (Kemp, Perfors, & Tenenbaum, 2007). The overhypotheses can depend on observable features such as whether the referent is animate (Smith, Jones, Landau, Gershkoff-Stowe, & Samuelson, 2002), the syntactic patterns in which the word appears (Cifuentes- F´erez & Gentner, 2006), or known lexical relations to other words (Pustejovsky, 1998). Return momentarily to the analogous results for shape bi- ases in nouns — that early nouns that children learn tend to be easily clustered by the shape of their referents. In order to model this bias, it was postulated that children learn a “second- order generalization” that objects are often categorized by their shape (Samuelson & Smith, 1999). Smith et al. demonstrated this generalization by teaching 17-19 month old children a precocious shape bias (Smith et al., 2002). Kemp, Perfors, and Tenenbaum explained this kind of learning using a hierarchical Bayesian model, which could learn both base meanings and overhypotheses simultaneously (Kemp et al., 2007) and cases The expression of motion verbs differs between languages. The path of motion, such as crossing or entering, is more promi- nently featured in path-based languages such as Spanish than in manner-based languages such as English. Here, we revisit the data from a study on manner and path biases in verb lexi- calization (Havasi & Snedeker, 2004), and create a hierarchical Baysian computational model to further explore, verify, and define these biases. With this model, we can discover the large differences in subjects’ pre-existing manner and path biases that depend on the syntactic frame in which new verbs appear, as well as a difference in the learning rate between English speakers taking the experiment in English and bilingual Span- ish speakers taking the experiment in Spanish. We can also use the model to predict the responses of subjects in the experiment with more accuracy than before. Keywords: verb learning; bayesian modeling; hierarchical Bayes modeling; manner and path verbs Linguistic lexicalization biases People have the ability to intuit the meaning of a new verb after hearing it used to describe just a single event. In the case of a novel verb, there are many potential hypotheses of the verb’s meaning which may be consistent with the event witnessed. Suppose you hear a novel verb, such as “gorp”, being used to describe an event in which Jesse throws a frisbee across a field to , her dog. The verb could refer to Jesse throwing the frisbee, the frisbee’s motion as it glides across the field, the frisbee’s traverse of the field, or Edison’s act of catching the frisbee. To understand which aspect of the action the verb refers to, you must use situational clues and background knowledge. When one encounters a new object noun, one encounters the same ambiguity in meaning. In practice, languages sys- tematically favor a few different characteristics such as com- mon ancestry or base level category (Nelson, 1973) for noun meanings which is often indicated by shape. However, event categorization tends to be flexible across languages and even with a language (Talmy, 1975). A motion verb, for example, could easily refer to the manner, cause, or path of the motion with no universal preference across languages (Aske, 1989; Berman & Slobin, 1994; Jackendoff, 1990). Given the plethora of possible referents for a novel verb, how do children learn verb meanings? One solution would be to observe, over several examples, that certain semantic features seem to always be present and are thus associated with the verb’s meaning. However, this would require too much data to match the way that children learn words; children can often determine the relevant aspect of a word’s meaning from

The results of 2,256 neuroimaging experiments were an- alyzed using singular value decomposition (SVD) and non-negative matrix factorization (NMF) to extract pat- terns in the data. To evaluate the techniques’ efficacy at capturing regularities in the data, one positive and one negative result from each of 100 random experi- ments were treated as missing, and the values were it- eratively reconstructed using each technique for dimen- sionality reduction. Under the best conditions, preci- sion and recall of roughly 78% was achieved for each method. Weighting the domain matrix and area matrix to have equal first eigenvalues before combining them, a technique known as blending, significantly improved re- sults for both methods. While using unnormalized data appeared to produce a peak in results for 10-15 dimen- sions, normalizing to take into account variation in the popularity of experiment types removed the effect. The basis vectors produced by each method do not support the idea that current cognitive ontologies map well to individual brain areas.

The Association for the Advancement of Artificial Intelligence was pleased to present the 2010 Fall Symposium Series, held Thursday through Saturday, November 11–13, at the Westin Arlington Gateway in Arlington, Virginia. The titles of the eight symposia are as follows: (1) Cognitive and Metacognitive Educational Systems; (2) Commonsense Knowledge; (3) Complex Adaptive Systems: Resilience, Robustness, and Evolvability; (4) Computational Models of Narrative; (5) Dialog with Robots; (6) Manifold Learning and Its Applications; (7) Proactive Assistant Agents; and (8) Quantum Informatics for Cognitive, Social, and Semantic Processes. The highlights of each symposium are presented in this report.

We present Luminoso, a tool that helps researchers to visualize and understand a dimensionality-reduced semantic space based on textual information by exploring it interactively. It streamlines the process of creating such a space by taking input from a directory of text documents, and optionally including common-sense background information. This interface is useful for interactively discovering trends in a text corpus, such as free-text responses to a survey. We discuss a case study about restaurant reviews to show how Luminoso can be used for opinion mining.