When a data file is imported into Ipiphany it is stored in a Dataset. 

Each Dataset contains records with the same types of attributes.

When creating a 'Discovery' you will need to select a specific Dataset.

Data for one instance in your Dataset. 

Consists of values for named attributes

When importing data each row in the file is stored in a record (within a dataset)

A named field in your Dataset

When importing data each column in the file is mapped to an attribute

Each attribute has a type of: Date, Number, Category, Score, Verbatim Text

An attribute of type ‘Verbatim Text’

When importing data you have one or more columns in the file that store verbatim text e.g. A ‘Recommend Reason’ and ‘Improvement Suggestion’ comments

Each of these columns is mapped to an Attribute of type Verbatim Text – which we refer to as a Verbatim Text Field

Ipiphany provides the option of analysing all or just some of the Verbatim Text Fields

An individual value for a Verbatim Attribute e.g. a specific comment

When importing data each non-blank cell in a column that stores verbatim text data is imported as a Verbatim

When viewing patterns in Ipiphany example Verbatims will be displayed as examples

A feature or pattern derived from Verbatim text using natural language processing. Examples include Keywords, Word Associations, Themes and Concepts

The base form of a word extracted from a Verbatim e.g. ‘fix’, ‘problem’

When processing Verbatim text Ipiphany extracts keywords by: Identifying the part of speech (POS) for all words, extracting those that are Nouns, Verbs, Adjectives and Adverbs and converting them to their base morphological form e.g. fixed -> fix, problems -> problem

Keywords may be used by Ipiphany to identify patterns within Verbatim attributes

An association between two keywords that is derived from their use in a Verbatim

When processing Verbatim text Ipiphany extracts word associations by: Determining the syntax of the language used, identifying the roles that words play and the relationships between them e.g. In “He fixed my problem” the target of the word ‘fixed’ is ‘problem’, and extracting important relationships and storing them as an association between two keywords e.g. fix<->problem

Word Associations may be used by Ipiphany to identify patterns within Verbatim attributes

A mathematical vector that stores semantic knowledge about a word, concept or other type of verbatim aspect.

When Ipiphany digests knowledge bases for use in semantic processing part of this knowledge is stored as compressed Word or Concept vectors

Word and Concept vectors are used by Ipiphany to understand that two words or concepts have a similar meaning or are related in some way e.g.

The word ‘problem’ is similar to the word ‘issue’ 

The word ‘solution’ is related to the word ‘problem’

A Verbatim Aspect that is derived directly from a curated Knowledge Source i.e. an Ontology or Knowledge Base e.g. ‘Problem Resolution’

Concepts are defined through natural language patterns that consist of keywords, word associations and other verbatim aspects

Concepts may be used by Ipiphany to identify patterns within Verbatim Text attributes

A collection of concept definitions and relationships

Ipiphany includes a number of general and industry specific ontologies

Custom ontologies can be developed by Touchpoint Group for specific companies or domains

A public store of general or industry specific knowledge

Ipiphany uses knowledge derived from public knowledge bases such as WordNet, Wikipedia/DBPedia and Concept Net

A Verbatim Aspect that defines an object or action, that is automatically derived directly from understanding the meaning of the verbatim text

When mining patterns from Verbatim text, Ipiphany uses semantic processing to understand where different words or associations have a similar meaning

These are then clustered into Broad or Narrow Themes

Themes may be used by Ipiphany to identify patterns within Verbatim attributes

Measures how often patterns occur in the data

Displayed as a Count or % 

Frequency % is calculated as: Pattern Count / Data Selection Count

Frequency is always calculated as a Record Count – Except within the Verbatim Panel where it is a Verbatim Count

A calculated statistic derived from a structured data attribute used as the main target metric for describing the importance of patterns with respect to your business objective e.g. NPS of Recommend Score, Average Call Time

A secondary statistic derived from the target measure or pattern frequency that is used to aid in comparing the importance of patterns.

There are three comparison metrics available: Difference, Impact and Relevance

Calculates the difference between the metric calculated for a specific pattern and the metric for data selection

Sometimes referred to as 'Delta'

Takes into account Difference and relative frequencies to calculate the effect that a pattern has on the metric for the data selection i.e. how much a pattern pushes the overall data selection metric up or down

Calculates how relevant a sub-pattern is with respect to either a pattern in your data, or to all your data. You can think of Relevance as finding the ‘uncommonly common’ sub-patterns that occur more frequently with the pattern than they do in your overall data selection – and calculating a score that combines both the uniqueness factor and the frequency. You will often find that sub-patterns with a high Relevance will contain important descriptive and root cause information.

A condition in the form Attribute=Value that selects a subset of your Dataset for analysis.

When querying data multiple filters are combined into a query

Filters for the same Attributes or Verbatim Aspects are combined using OR Filters for different Attributes or Verbatim Aspects are combined using AND

A type of Filter that matches to Verbatim text. 

There are 4 types of Search Terms:

Words - match to similar words  e.g. “account” matches “account”, “accounts”

Phrase – matches quoted words to similar phrase  e.g. “close account”matches “closed account”, “closing accounts”

Wildcards – match to patterns e.g. “cl*d” matches “closed”, “cloud”

Expression – matches using boolean logic e.g. “open & account” will match to “I opened an account”.

Keyword in Context – a way of visualising verbatim text so key aspects aligned and highlighted so they are quick to read and understand