1/16 LELA Language and Computers Harold Somers Professor of Language Engineering

2/16 Getting computer to handle language in a natural way –As part of interface with user (eg alternatives to keyboard/mouse input, text output, support for the disabled) –consulting databases to get information (e.g. library catalogue, train timetable, banking) –As a task in itself: specifically linguistic purposes (e.g. dictation, translation, summarizing, report writing, language teaching) Using computer to “do” linguistics Synonyms (or components?) –Natural Language Processing –Computational Linguistics –Language Engineering Basic tools, techniques and models Applications Computers and Language?

3/16 Syllabus Survey of applications Elements of language, levels of linguistic processing –Sound (speech recognition, synthesis) –Writing (OCR, handwriting, character sets, spelling) –Words (lexicon, morphology) –Sentences (syntax) –Semantics, Pragmatics Branches of language/linguistics –Psycholinguistics and AI –Applied linguistics (teaching, translation) –Information retrieval –Or: your choice of anything else relevant (historical, socio, etc.)

4/16 Mon 10am Newman/G16Fri 10am AlanTuring/G Introduction Sept 24, 28 Introduction, explanation of syllabus What makes language hard for computers? Applications: overview 2. Phonetics Oct 1, 5 Speech recognitionSpeech synthesis 3. Writing Oct 8, 12 Character sets, Unicode, input methodsSpell checkers, grammar checkers 4. Words Oct 15, 19 Dictionaries, ThesauriWordNet 5. Morphology Oct 22, 26* MorphologyTagging READING WEEK 6. Syntax Nov 5, 9 Parsing IParsing II 7. Semantics,Pragmatics Nov 12, 16 Dialogue understandingText generation, Cooperative responses 8. Branches Nov 19, 23 Psycholinguistics: language and AIApplied linguistics: CALL 9. Applied linguistics Nov 26, 30* CALLTranslation aids 10. Translation Dec 3, 7 Machine translation IMachine Translation II 11. Information retrieval Dec 10, 14 Text retrieval; SummarizationQuestion answering, text mining * These lectures to be rearranged due to HS’s absence

5/16 Assessment Examination: 50 multiple choice (or short phrase) answers Essay – 3000 word essay due in at the start of the exam period Choose a particular NLP application, and explain the difficulties that natural language poses to the computer, and how (or whether) they are addressed. How is the study of language relevant to NLP? Focus on one or two areas of NLP only. Choose one branch of language study (eg historical, sociolinguistics, psycholinguistics, child language, applied linguistics, etc.) and discuss what role the computer, and in particular computational linguistics, could play. (Note that all involve taking lectures as starting point, and going into more detail, based on your own “research”)

6/16 What is NLP? Language and Linguistics Psychology Electrical Engineering Computer Science NLP Artificial Intelligence Language Engineering HCI Signal Processing Phonetics PhilosophyLogic

7/16 Language and AI “language ability” is an integral part of Artificial Intelligence, which relates to robotics Computers in SciFi use language readily: how realistic is this? What are the problems? HAL in “2001: A Space Odyssey” (1968)

8/16 How realistic is HAL? Non-linguistic functions include... monitoring and controlling the spaceship playing chess vision general reasoning about the world … especially this particular mission

9/16 HAL’s use of language He has to use language to communicate with the crew, including... chatting sociably discussing general and specific conditions in the spaceship understanding commands initiating conversations, whether “work” or “play”

10/16 What is language? spoken vs. written language speaker vs. hearer = production vs. analysis different levels of language different functions of language

11/16 Computer speech for natural-sounding speech, computer must get individual sounds right, but also combine them correctly intonation stress (pitch, loudness, length) pace (pauses can be significant)

12/16 Speech understanding signal processing (acoustic physics) separating speech from background noise recognizing individual speech sounds (humans can make very fine distinctions) It’s hard to wreck a nice beach What dime’s a neck’s drain to stop port? variability in human voices

13/16 Problems with language in general Words are ambiguous (bank, round, take) Sentences are ambiguous The chicken is ready to eat Visiting relatives can be boring End to free school looms The man saw the girl with a telescope Remove bulb, cover, and replace

14/16 Pragmatic problems We don’t always say what we mean Can you pass the salt? It’s cold in here, isn’t it? I’m sorry (= Say it again) Do you want some more? You’re alright. We don’t always mean what we say It’s raining cats and dogs I could murder a sandwich

15/16 Solutions 4Linguistics (“grammar”) can often tell us which interpretations are possible, including limited aspects of meaning, e.g. The man saw the girl with a hat 4Restricted domain: if we know what the subject is, a lot of ambiguity disappears Context Real-world knowledge

16/16 Next up A whistle-stop tour of applications and what they involve Then down to business!