Building a truly interactive customer service system like Nuance's requires solutions to each of the major challenges in natural-language processing: accurately transforming human speech into machine-readable text; analyzing the text’s vocabulary and structure to extract meaning; generating a sensible response; and replying in a human-sounding voice." - from Wade Roush's article, Computers That Speak Your Language

Introduction to NLG, a white paper from CoGenTex, Inc. "Natural Language Generation (NLG), also referred to as text generation, is a subfield of natural language processing (NLP; which includes computational linguistics). For those not familiar with these areas, this page provides a brief overview of what NLG is (and is not)...."

AI systems may blow weathermen away. New Scientist (September 24, 2005; Issue 2518: page 27). "Weather forecasters could find themselves pushed out of a job by an artificial intelligence system designed to write clearer, less ambiguous reports. Computer scientists at the University of Aberdeen, UK, were asked to generate an 'artificial weatherperson' by operators of offshore oil rigs, who wanted more clarity in their forecasts. ... To remove such uncertainties, the team programmed a natural language generation (NLG) software package to transform data on the forecast weather into an unambiguous written bulletin (Artificial Intelligence, vol 167, p 137)."

• Visit the Natural Language Generation Group at the University of Aberdeen.

Survey of the State of the Art in Human Language Technology (1996). Editorial Board: Ronald A. Cole, Editor in Chief, Joseph Mariani, Hans Uszkoreit, Annie Zaenen, Victor Zue; Managing Editors: Giovanni Battista Varile, Antonio Zampolli.

• Chapter 3 - Language Analysis and Understanding. 3.1 Overview. : By Annie Zaenen & Hans Uszkoreit. "We understand larger textual units by combining our understanding of smaller ones. The main aim of linguistic theory is to show how these larger units of meaning arise out of the combination of the smaller ones. This is modeled by means of a grammar. Computational linguistics then tries to implement this process in an efficient way. It is traditional to subdivide the task into syntax and semantics, where syntax describes how the different formal elements of a textual unit, most often the sentence, can be combined and semantics describes how the interpretation is calculated.
• Chapter 4: Language Generation. 4.1 Overview. By Eduard Hovy. "The area of study called natural language generation (NLG) investigates how computer programs can be made to produce high-quality natural language text from computer-internal representations of information."

Natural Language Generation overview from FLUIDS [Future Lines of User Interface Decision Support]. "The study of human language generation is a multidisciplinary enterprise, requiring expertise in areas of linguistics, psychology, engineering and computer science. One of the central goals is to investigate how computer programs can be made to produce high-quality natural language text from computer-internal representations of information."

Spoken Language Systems Group, MIT Computer Science and Artificial Intelligence Laboratory.

• About SLS: "User: Yes, I would like the weather forecast for London, England, please.JUPITER: In London in England Wednesday, partly cloudy skies with periods of sunshine. High 82 and low 63. Is there something else? ... SLS researchers make this kind of dialogue look easy by empowering the computer to perform five main functions in real time: speech recognition-- converting the user's speech to a text sentence of distinct words, language understanding -- breaking down the recognized sentence grammatically, and systematically representing its meaning, information retrieval -- obtaining targeted data, based on that meaning representation, from the appropriate online source, language generation -- building a text sentence that presents the retrieved data in the user's preferred language, and speech synthesis -- converting that text sentence into computer-generated speech. Throughout the conversation, the computer also remembers previous exchanges."
• Core Technology Development: "To support its research on spoken language systems for human/computer interaction, the SLS group has developed its own suite of core speech technologies. These technologies include: * speech recognition (SUMMIT) * natural language understanding (TINA) * dialogue modeling * language generation (GENESIS) * speech synthesis (ENVOICE)."
• Applications

Natural Language Generation. From Gerd Herzog at the German Research Center for Artificial Intelligence GmbH. A succinct definition plus links to additional information.

Computers That Speak Your Language - Voice recognition that finally holds up its end of a conversation is revolutionizing customer service. Now the goal is to make natural language the way to find any type of information, anywhere. By Wade Roush. Technology Review (June 2003). "Building a truly interactive customer service system like Nuance's requires solutions to each of the major challenges in natural-language processing: accurately transforming human speech into machine-readable text; analyzing the text’s vocabulary and structure to extract meaning; generating a sensible response; and replying in a human-sounding voice." And be sure to see the related illustration: Inside a Conversational Computer.

SHRDLU, one of the early AI programs. "SHRDLU is a program for understanding natural language, written by Terry Winograd at the M.I.T. Artificial Intelligence Laboratory in 1968-70. SHRDLU carried on a simple dialog (via teletype) with a user, about a small world of objects (the BLOCKS world) shown on an early display screen (DEC-340 attached to a PDP-6 computer)."

• And for fun, check out the the story of how SHRDLU got its name.

Google's Peter Norvig on managing the data deluge. Video of talk delivered on September 25, 2006 at UC Berkeley as part of the CITRIS Distinguished Speaker Series. "Researchers in computational linguistics and information retrieval now have a million times more data than was available 30 years ago. In this talk, Peter Norvig explores what this data can do for problems in language understanding, translation, information extraction, and inference, and extrapolates to what more data may bring in the future."

User Modeling and HCI Approaches in Natural Language Generation (Special Track). Proceedings of the Seventeenth International Florida Artificial Intelligence Research Society Conference,Valerie Barr and Zdravko Markov, eds. 2004. Menlo Park, Calif.: AAAI Press.

• Papers include: Generating Tailored, Comparative Descriptions in Spoken Dialogue. By Johanna Moore, Mary Ellen Foster, Oliver Lemon, and Michael White. Abstract: "We describe an approach to presenting information in spoken dialogues that for the first time brings together multi-attribute decision models, strategic content planning, state-of-the-art dialogue management, and realization which incorporates prosodic features. The system selects the most important subset of available options to mention and the attributes that are most relevant to choosing between them, based on the user model. It also determines how to organize and express descriptions of selected options and attributes, including determination of information structure and rhetorical structure at the level of content planning, resulting in descriptions which, we hypothesize, are both memorable and easy for users to understand."

Narrative Prose Generation. By Charles Callaway and James Lester. In Proceedings of the Seventeenth International Joint Conference on Artificial Intelligence, Seattle, WA, August 2001.

Natural Language Generation in Health Care. By Alison J. Cawsey, Bonnie L. Webber, and Ray B. Jones. Journal of the American Medical Informatics Association 4:473-482 (1997) "In health care, the evident need to translate between textual forms (human authored texts) and structured information has led to a large and continually growing body of research and development in natural language understanding. In this article we consider the reverse problem, how textual documents may be produced from structured data. In particular, we show how a range of current natural language generation techniques can be used to produce from the same data, many different documents with different content, terminology and style, and thereby help meet diverse information needs within health care."

At I.B.M., That Google Thing Is So Yesterday. By James Fallows. The New York Times (December 26, 2004; reg. req'd.). "Suddenly, the computer world is interesting again. ... The most attractive offerings are free, and they are concentrated in the newly sexy field of 'search.' ... [T]oday's subject is the virtually unpublicized search strategy of another industry heavyweight: I.B.M. ... I.B.M. says that its tools will make possible a further search approach, that of 'discovery systems' that will extract the underlying meaning from stored material no matter how it is structured (databases, e-mail files, audio recordings, pictures or video files) or even what language it is in. The specific means for doing so involve steps that will raise suspicions among many computer veterans. These include 'natural language processing,' computerized translation of foreign languages and other efforts that have broken the hearts of artificial-intelligence researchers through the years. But the combination of ever-faster computers and ever-evolving programming allowed the systems I saw to succeed at tasks that have beaten their predecessors. ... ... Jennifer Chu-Carroll of I.B.M. demonstrated a system called Piquant, which analyzed the semantic structure of a passage and therefore exposed 'knowledge' that wasn't explicitly there. After scanning a news article about Canadian politics, the system responded correctly to the question, 'Who is Canada's prime minister?' even though those exact words didn't appear in the article."

• Check out IBM's Natural Language Processing projects and the IBM Human Language Technologies Group's page about natural language understanding.
• Also see their Natural Language Processing research overview.

Natural Language Generation in Spoken and Written Dialogue: Papers from the 2003 Spring Symposium, ed. Reva Freedman and Charles Callaway. Technical Report SS-03-06. American Association for Artificial Intelligence, Menlo Park, California.

• Preface. By Reva Freedman and Charles Callaway. "Human-computer dialogue systems, including spoken, written, GUI-based and multimodal, have increased in number, type and complexity over the last decade. Previous work in the artificial intelligence and computational linguistics communities has focused on the building and evaluation of these systems in a variety of domains. At the same time, other forums for natural language generation research have been discussing the generation of everything from noun phrases to longer monologues. The papers contained in the proceedings show that a large number of researchers in both industry and academia, and from around the world, have taken up the challenge to build dialogue systems with a substantial amount of computer-generated rather than pre-written dialogue."

A collection of publications from the HealthDoc Project ( "What is needed is a natural language generation system for the production of tailored health-information and patient-education documents, that would, on demand, customize a 'master document' to the needs of a particular individual. Building such a system is the goal of the HealthDoc project.")

• Also see: Sentence Planning. From the Information Sciences Instituteof the University of Southern California, a HealthDoc Project collaboratior.

Intelligent Text Summarization: Papers from the 1998 Spring Symposium, ed. Eduard Hovy and Dragomir Radev. Technical Report SS-98-06. American Association for Artificial Intelligence, Menlo Park, California.

Natural Language Understanding Lecture Notes from Professors Tomás Lozano-Pérez & Leslie Kaelbling's Spring 2003 course: Artificial Intelligence. Available from MIT OpenCourseWare.

Natural Language Understanding and Conversational Dialogue - A Different Kind of Self-Service Speech Recognition. By Stefania Viscusi. TMCnet (October 2, 2007). "For more insight into natural language understanding in speech technologies, I took some time to ask Luis Valles, Chief Scientist at GyrusLogic, some questions on the topic. [Q] What is Natural Language Understanding? [A] Natural Language Understanding (NLU) or Conversational Dialogue is the capability for a user to say and/or ask anything, and the system understanding what the user meant, together with the system finding an appropriate response -- as with any other conversation between humans. [Q] How is this deployed with Speech Recognition? ... [Q] Can you tell me a little about the solution you have developed to provide Natural Language Understanding capabilities? [A] GyrusLogic’s Platica product is patented artificial intelligence (AI) technology built with computational linguistic models for customers, employees or any other stakeholder to enter into a fully-automated conversational dialog. ..."

• Visit GyrusLogic.

Cognitive Machines Group at the MIT Media Lab. "Our goal is to create machines which learn to communicate on human terms. ... An emphasis is placed on machine understanding and generation of semantically-grounded spoken language, leading to numerous applications for human-machine communication and human-human communication aids."

When Will HAL Understand What We Are Saying ? Computer Speech Recognition and Understanding. By Raymond Kurzweil. From HAL's Legacy: 2001's Computer as Dream and Reality (MIT Press, 1996). "Let's talk about how to wreck a nice beach. Well, actually, if I were presenting this chapter verbally, you would have little difficulty understanding the preceding sentence as Let's talk about how to recognize speech."

Interactive Systems Labs. Located at Carnegie Mellon University and at University of Karlsruhe, these laboratories are engaged in the development of "user interfaces that improve human-machine and human-to-human communication...[Their] first demonstration of JANUS (in early '93) showed that speaker-independent, continuous speech-to-speech translation is possible." When you visit this site, be sure to explore the Projects, Demos, Publications and other options offered on the navigation bar.

"What do you say to a twenty-five-pound computer? If the Natural Language Processing (NLP) research group at Microsoft has their way--anything you want. The group is building a "gigantic brain" that will understand natural language text input in seven different languages. Visit the Microsoft NLP group's home page and see what other projects they're working on.

Understanding Natural Language. PARC (Palo Alto Research Center). "The problem with creating useful representations of natural language is that they need to go below the surface. Each word in a phrase can have many possible meanings, and phrases can be combined in a number of different ways. PARC's leading-edge deep parsing and semantics algorithms generate possible valid interpretations based on grammatical factors -- such as whether words are verbs, nouns, subjects or objects, active or passive, etc. -- and the relationships among words and phrases. The inherent ambiguity of natural language means that parsing and semantic algorithms often generate large numbers of alternative interpretations." Be sure to see their links to related information, publications, and research groups.

Parsing Demos - just a few of the many that are available online:

• "The Link Grammar Parser is a syntactic parser of English, based on link grammar, an original theory of English syntax. Given a sentence, the system assigns to it a syntactic structure, which consists of a set of labeled links connecting pairs of words. The parser also produces a 'constituent' representation of a sentence (showing noun phrases, verb phrases, etc.)." From Davy Temperley, Daniel Sleator, and John Lafferty, School of Computer Science, at Carnegie Mellon University. Try the demo.
• The Memory-Based Shallow Parser demo. "(MBSP) applies a sequence of modules to English sentences: PoS tags (by MBT), chunks (non-overlapping, non-embedded constituents), and verb-subject and verb-object relations." One of the software demos avilable from the ILK project [Induction of Linguistic Knowledge] at the Computational Linguistics and AI section of the Faculty of Arts of Tilburg University
• Probabilistic Context-Free Parsing demo. One of the several software demos from The National Centre for Language Technology at the School of Computing, Dublin City University, which "conducts research into the processing of human language by computers. This involves the use of computers in speech recognition and production, translation, human-computer interfaces, information retrieval and extraction from the world-wide web, the teaching and learning of languages using computers and software localisation and globalisation."
• Natural Language Parsers and Taggers. "The online demo demonstrates how Connexor Machinese products work in practice. The demo allows you to see what kind of analysis you get with Machinese Phrase Tagger, Machinese Syntax and Machinese Semantics."
• SPARSE II - StudentPARSing Environment II CGI-Web Demonstration Version. Written by Clayton M. Darwin and made available by the The Artificial Intelligence Center at The University of Georgia. "This is a web demonstration version of the SPARSE II parsing program. SPARSE II (Student PARSing Environment) is a parsing program intended to be used as a pedagogical tool to help syntax students grasp the complexity of natural-language grammars and to begin developing their own models. It provides a true introduction to Natural Language Processing without requiring familiarity with Lisp or Prolog."
• XIP demo from Xerox Research Centre Europe.

Special Interest Group on Text Generation (SIGGEN), Association for Computational Linguistics. Lots of information, plus links to research groups, bibliographies, and more.

Gemini Natural-Language Understanding System. "Natural-language research under SRI International's project on Improved Spoken-Language Understanding is focused on the development of Gemini, a natural-language parsing and semantic interpretation system based on unification grammar. 'Unification grammar' means that grammatical categories incorporate features that can be assigned values; so that when grammatical category expressions are matched in the course of parsing or semantic interpretation, the information contained in the features is combined, and if the feature values are incompatible the match fails. Gemini applies a set of syntactic and semantic grammar rules to a word string using a bottom-up parser to generate a logical form, which is a structured representation of the context-independent meaning of the string."

Natural Language Generation for a Speech Prosthesis. From the Stanford University Center for the Study of Language and Information (CSLI). Following the project summary you'll find links to several downloadable papers.

SUMMARIST: Automated Text Summarization. One of the projects from The Natural Language Processing Group at the Information Sciences Institute of the University of Southern California (USC/ISI). "Summarization = Topic Identification + Interpretation + Generation"

Xerox Research Centre Europe (XRCE) - Parsing & Semantics: "ParSem concentrates on automatically making sense of electronic documents, by semantically analyzing them. ParSem concentrates on two main research lines of natural language processing: robust parsing and semantics."

• Robust Parsing:"Robust parsing provides mechanisms for identifying major syntactic structures and major functional relations between words on large collections of unrestricted documents (ex: Web pages, newspapers, scientific literature, encyclopedias). ... Major applications include contextual entity recognition, lexical and structural disambiguation, coreference resolution and more globally knowledge extraction."
• Semantics: "With the goal of transforming documents into “meaningful spaces”, the main focus has to be semantics. Semantics is everywhere, hidden in completely different types of documents (e.g. text, images, videos, programs and audio) and at different levels (e.g. document content, document structure). Because most of the “semantics” that is nowadays accessible in documents lies in texts, we concentrate on the semantic content analysis of the textual parts of documents. ... Our current research themes include:Ontology Acquisition ... Semantic Disambiguation ... Linguistic Normalization ... Co-reference ... Discourse Analysis ..."
• Demos & Videos

Related AI Topics Pages

• Chatbots (@ Natural Language Processing)
• Discourse Analysis
• Drama, Fiction & Poetry
• Human-Computer Interfaces
• Information Retrieval & Extraction
• Machine Translation
• Natural Language Processing (NLP)
• Representation
• Speech
• Turing Test

Other References Offline

House, David. Interactive Text Summarization for Fast Answers. The MITRE Advanced Technology Newsletter (July 1997). "By interactively selecting terms of interest and viewing the corresponding context-dependent summaries, users can quickly find answers relevant to their queries. The technology behind WebSumm exploits recent advances in artificial intelligence and information retrieval."

Hovy, Eduard and Dragomir Radev, Cochairs. Intelligent Text Summarization: Papers from the 1998 AAAI Spring Symposium.