Text Mining: Finding Nuggets in Mountains of Textual Data Jochen Dörre, Peter Gerstl, and Roland Seiffert Presented By: Jake Happs, 4.11.01.

Text Mining: Finding Nuggets in Mountains

of Textual Data

Jochen Dörre, Peter Gerstl, and Roland Seiffert

Presented By: Jake Happs, 4.11.01

Overview

• Reasons for Text Mining

• Special Tasks in Mining Text

• Disambiguating Proper Names

• Application Types

• Customer Intelligence

Reasons for Text Mining

• Corporate Knowledge “Ore”

• Exploiting the Knowledge in Text

• The Value of Mining Text

• Typical Applications

Corporate Knowledge “Ore”

• Email• Insurance claims• News articles• Web pages• Patent portfolios

• Customer complaint letters

• Contracts• Transcripts of phone

calls with customers• Technical documents

Exploiting Textual Knowledge

• Knowledge Discovery

• Knowledge Management

Value of Text Mining

• Rapid digestion of large corporate documents, faster than human knowledge brokers

• Objective and customizable analysis

• Automation of routine tasks

Typical Applications

• Summarizing documents

• Monitoring relations among people, places, and organizations

• Organize documents by content

• Organize indices for search and retrieval

• Retrieve documents by content

Special Tasks in Mining Text

• Interpreting Natural Language

• Comparison with Data Mining

• Extracting Terminology and Relations

• Classifying Documents

Interpreting Natural Language

• Extracting terminology

• Extracting relations

• Summarizing documents

• Extracting models

Comparison of Procedures

Data Mining• Identify data sets.• Select features

manually.• Prepare data.• Analyze distribution.

Text Mining• Identify documents.• Extract features.• Select features by

algorithm.• Prepare data.• Analyze distribution

Terminology and Relations

• What Terminology Is

• Classes of Terms

• Instances of Relations

• Canonical Forms

What Terminology Is

• Function words

• General-purpose content words and phrases

• Technical content words and phrases

• Relations

Classes of Terminology

• Proper names

• Technical phrases

• Abbreviations and acronyms

Instances of Relations

• Facts

• Dates

• Currency values

• Percentages

• Other measurements

Canonical Forms

• Numbers convert to normal form.

• Dates convert to normal form.

• Inflected forms convert to common form.

• Alternative names convert to explicit form.

Classifying Documents

• Hierarchical clustering

• Binary relational clustering

• Supervised learning

Disambiguating Proper Names

• Principles of Nominator Design

• The Process in Nominator

Principles of Nominator Design

• Apply heuristics to strings, instead of interpreting semantics.

• The unit of context for extraction is a document.

• The unit of context for aggregation is a corpus.

• The heuristics represent English naming conventions.

Extracting Proper Names

• Tokenize the words in a document.

• Build list of candidate names in document.

• Break candidates into smaller names.

• Group names into equivalence classes.

• Aggregate classes from multiple documents.

Candidate Names

• Extract all sequences of capitalized tokens.• Exclude adjectives of provenance (e.g. Mr., Dr.,

etc.).• Exclude certain non-name acronyms (e.g. M.D.,

PhD.).• Include numerals, unless following a preposition,

comma, date, or number.• Ignore words in section titles.• Exclude initial adverbs in sentences.

Splitting Candidates

• Apply heuristics to conjunctions, prepositions, and possessives.

• Reconstruct shared words.

Building Equivalence Classes

• Discard non-recurring initial words of sentences.

• Unify variants with heuristics.

• Pick canonical name for each class.

• Categorize each class with heuristics.

• Map canonical name to variants.

• Map variants to canonical name.

Aggregating Classes

• Merge classes that share a variant in separate documents.

• Both type and spelling of variant must agree.

• Replace uncertain categories with certain ones.

Application Types

• Knowledge Discovery (Clustering)

• Information Distillation (Categorization)

Knowledge Discovery

Information Distillation

Customer Intelligence

• Goals

• Process

Customer Intelligence Goals

• What do customers want and need?

• What do customers think of the company?

Customer Intelligence Process

• Corpus of communications with customers

• Cluster the documents to identify issues.

• Characterize the clusters to identify the conditions for problems.

• Assign new messages to appropriate clusters.

Summary

• Reasons for Text Mining

• Special Tasks in Mining Text

• Disambiguating Proper Names

• Customer Intelligence

Exam Question #1

• Name an example of each of the two main classes of applications of text mining.– Knowledge Discovery: Discovering a common

customer complaint among much feedback.– Information Distillation: Filtering future

comments into pre-defined categories

Exam Question #2

• How does the procedure for text mining differ from the procedure for data mining?– Adds feature extraction function– Not feasible to have humans select features– Highly dimensional, sparsely populated feature

vectors

Exam Question #3

• In the Nominator program of IBM’s Intelligent Miner for Text, an objective of the design is to enable rapid extraction of names from large amounts of text. How does this decision affect the ability of the program to interpret the semantics of text?– Does not perform in-depth syntactic or

semantic analyses of texts

Questions & Answers