Did you ever ask yourself about the wonder of natural language processing (NLP) as an innate part of human civilization? Do you wish to unlock some of the secrets of natural language understanding in minds and machines alike? Do you wish to understand some of the limitations of NLP and understanding for machines, such as those in the context of rising robots and artificial intelligence (AI)?

If you are a neuroscientist or interested in psycholinguistics, you may certainly have some answers to these questions, for instance, what the human association network of words may be and how this can be utilized in the brain for understanding, that is, capturing the meaning in natural language, spoken or written. If you are interested in computational linguistics and the semantics in NLP, you have certainly come across latent semantic analysis (LSA) and latent Dirichlet allocation (LDA) algorithms as two of the most prominent algorithms to capture the meaning of written text. However, it has rarely been the case that these disciplines learned from each other. In this context, this book is an excellent fusion of insights in both worlds, those shed by cognitive sciences and those shed by computer science.

However, it may be disappointing that this is not a textbook. It is an edited book--a collection of papers submitted for publication in a, quoting from the preface, “special issue dedicated to explore the relationship between natural language processing and cognitive science,” as one of the workshops in the Natural Language and Cognitive Science (NLPCS) series of international workshops launched in 2004. In that context, the book will certainly appeal to interdisciplinary research projects aiming at bringing computer scientists together with cognitive and linguistic researchers in order to advance natural language processing.

It is, therefore, no surprise that all the collected articles, chapters if you wish (ten in total), are attempts to highlight the limitations of machine-based NLP in regard to capturing the meaning in text from a cognitive science point of view. The first five articles, however, deal predominantly with human association networks (HANs), which are believed to be the main representation mechanisms of meaning of words in the mind. These HANs, which are attributed to the Church and Hunks [1] psycholinguistic and algorithmic approaches in the 1990s, are contrasted with what is being extracted as semantic dependencies, for instance, by LSA and LDA as main representatives of computerized NLP.

The fifth chapter, in particular, sheds some light into the hidden structure of a dictionary, as it highlights that the meaning of every word can be defined by some core word definitions. For a computer scientist like myself, having spent many years in NLP and semantic technology, this has been one of the most significant and exciting insights from the book.

The rest of the book is dedicated to challenges like automated production of coherent text (chapter 7), with the most interesting aspects being those of measuring coherence in text segments; disambiguating word senses by applying learning games (chapter 6); authorship detection and attribution (chapter 8), which is very close to named entity recognition in NLP, though this is not being mentioned explicitly as another example of the lack in communication between computerized NLP and cognitive science; a parallel cognition-oriented fundamental frequency estimation (chapter 9), mostly useful for speech recognition; and, finally, a chapter contrasting (actually, benchmarking) n-grams, recurrent neural networks (RNNs), and topic models in computer science, with “cloze completions” in neuroscience as a mechanism to see meaning in a context going beyond lexical meaning, for example, eye movement and electroencephalograms (EEGs).

All in all, for a computer scientist specializing in areas like computational linguistics, natural language processing, robotics, chatbots, navigation systems, speech recognition, and perhaps processing of sign languages, from an interdisciplinary point of view, this is an excellent starting point. Though not a textbook per se, it could be suggested to students at the master’s level or those embarking on related PhD studies.