1. COMP3410 DB32: Technologies for Knowledge Management 08 : Introduction to Knowledge Discovery By Eric Atwell, School of Computing, University of Leeds (including re-use of teaching resources from other sources, esp. Knowledge Management by Stuart Roberts, School of Computing, University of Leeds)

2. What has Machine Learning got to do with Computing / Information Systems? Most international organizations produce more information in a week than many people could read in a lifetime Adriaans and Zantinge

3. Objectives of knowledge discovery or data mining Data mining is about discovering patterns in data. For this we need: –KD/DM techniques, algorithms, tools, eg BootCat, WEKA –A methodological framework to guide us, in collecting data and applying the best algorithms: CRISP-DM

4. Data Mining, Knowledge Discovery, Text Mining Data Mining was originally about learning patterns from DataBases, data structured as Records, Fields Knowledge Discovery is exotic term for DM??? Increasingly, data is unstructured text (WWW), so Text Mining is a new subfield of DM, focussing on Knowledge Discovery from unstructured text data

5. define: data mining Data mining, also known as knowledge-discovery in databases (KDD), is the practice of automatically searching large stores of data for patterns. To do this, data mining uses computational techniques from artificial intelligence, statistics and pattern recognition. en.wikipedia.org/wiki/Data_mining en.wikipedia.org/wiki/Data_mining

6. define: text mining Text mining, also known as intelligent text analysis, text data mining or knowledge-discovery in text (KDT), refers generally to the process of extracting interesting and non-trivial information and knowledge from unstructured text. Text mining is a young interdisciplinary field which draws on information retrieval, data mining, machine learning, statistics and computational linguistics.... en.wikipedia.org/wiki/Text_mining en.wikipedia.org/wiki/Text_mining

7. define: knowledge discovery Knowledge discovery is the process of finding novel, interesting, and useful patterns in data. Data mining is a subset of knowledge discovery. It lets the data suggest new hypotheses to test. www.purpleinsight.com/downloads/docs/visualizer_ tutorial/glossary/go01.html www.purpleinsight.com/downloads/docs/visualizer_ tutorial/glossary/go01.html Data mining, also known as knowledge-discovery in databases (KDD), is the practice of automatically searching large stores of data for patterns. To do this, data mining uses computational techniques from AI, statistics and pattern recognition. en.wikipedia.org/wiki/Knowledge_discovery en.wikipedia.org/wiki/Knowledge_discovery

8. Data Mining: Overview Concepts, Instances or examples, Attributes Data Mining Concept Descriptions Each instance is an example of the concept to be learned or described. The instance may be described by the values of its attributes.

9. Instances Input to a data mining algorithm is in the form of a set of examples, or instances. Each instance is represented as a set of features or attributes. Usually in DB Data-Mining this set takes the form of a flat file; each instance is a record in the file, each attribute is a field in the record. In text-mining, instance may be word/term in context (surrounding words/document) The concepts to be learned are formed from patterns discovered within the set of instances.

10. concepts The types of concepts we try to learn include: Key differences – terms specific to our domain corpus Clusters or Natural partitions; –Eg we might cluster customers according to their shopping habits. –Eg is this web-page British or American English? Rules for classifying examples into pre-defined classes. –Eg Mature students studying information systems with high grade for General Studies A level are likely to get a 1 st class degree General Associations –Eg People who buy nappies are in general likely also to buy beer

11. More concepts The types of concepts we try to learn include: Numerical prediction –Eg look for rules to predict what salary a graduate will get, given A level results, age, gender, programme of study and degree result – this may give us an equation: Salary = a*A-level + b*Age + c*Gender + d*Prog + e*Degree (but are Gender, Programme really numbers???)

12. DB Example: weather to play?

14. /usr/local/weka-3-4-8a/data/weather.arff @relation weather @attribute outlook {sunny,overcast,rainy} @attribute temperature real @attribute humidity real @attribute windy {TRUE, FALSE} @attribute play {yes, no} @data sunny,85,85,FALSE,no sunny,80,90,TRUE,no overcast,83,86,FALSE,yes rainy,70,96,FALSE,yes rainy,68,80,FALSE,yes rainy,65,70,TRUE,no overcast,64,65,TRUE,yes sunny,72,95,FALSE,no sunny,69,70,FALSE,yes rainy,75,80,FALSE,yes

15. Text mining example: Which English dominates the WWW, UK or US? First catch your rabbit (Mrs Beatons cookbook): Other tools are possible, but WWW-BootCat was easier to use … First: sign up for Domain, SketchEngine account, Google key; download seeds-en from http://corpus.leeds.ac.uk/internet.html http://corpus.leeds.ac.uk/internet.html (see Atwell et al 2007 for details)

16. Data Mining for a thesaurus? WordNet is freely-available for general English What about other languages? – EuroWordnet, BalkaNet, … (but not ALL languages!) What about specific domains? Domain-specific ONTOLOGIES have been devised (by experts) What about my own specific domain/language? Automatic extraction from example documents (machine learning / knowledge discovery)

17. Automatic thesaurus construction based on documents (either retrieved set or the whole collection) as Corpus – training text set define a measure of how close one index term is to another – in meaning-space, ?or literal distance? for each term, form a neighbourhood comprising the nearest n terms treat these neighbourhoods like related thesaurus classes terms with similar neighbourhoods are treated as synonyms.

18. Finding coordinate terms One attempt to define how close a term is to another: If two terms are both used to index the same document many times in the collection, then they are deemed to be close. From document-term matrix, compute term-correlation matrix The term correlation matrix can be normalised so that terms that index a lot of documents dont have an unfair chance.

19. Other ways to find specialist terms Other ways to find domain-specific terms and relations: Collect a domain corpus, find terms different from a generic gold standard corpus: British National Corpus Collocation-groups: For each term, collect its collocations in the Corpus: other words it appears next to (or near to). If two terms have similar collocation- sets, then they are deemed to be close. Association matrix based on proximity: compute average distance between pairs of terms (no. of words between them, literally), use this as closeness metric

20. Why build a thesaurus? a thesaurus or ontology can be used to normalise a vocabulary and queries (?or documents?) it can be used (with some human intervention) to increase recall and precision generic thesaurus/ontology may not be effective in specialized collections and/or queries Semi-automatic construction of thesaurus/ontology based on the retrieved set of documents has produced some promising results, e.g. Semantic Web

21. Knowledge Discovery: Key points Knowledge Discovery (Data Mining) tools semi- automate the process of discovering patterns in data. Tools differ in terms of what concepts they discover (differences, key-terms, clusters, decision-trees, rules)… … and in terms of the output they provide (eg clustering algorithms provide a set of subclasses) Selecting the right tools for the job is based on business objectives: what is the USE for the knowledge discovered

22. A Data Mining consultant… You should be able to: –Decide which is the appropriate data mining technique for a given a problem defined in terms of business objectives. –Decide which is the most appropriate form of input (which attributes/features will be useful for learning) and output (what does your client want to see?)