CB2500 Chap.8 Business Intelligence and Text Topic Analysis
Business Intelligence and Text Topic Analysis
The chapter the project cannot be completed without
The brief is unusually specific here. A team must collect data with FacePager from a page the target company hosts and run a text topic analysis in SAS Enterprise Miner, using the techniques taught in the first three compulsory modules of the social analytics and business intelligence material.
Qualitative reading of news articles and industry reports may complement that analysis; it may not replace it.
The requirement exists because a consultation report has to be defensible to somebody who did not do the reading, and a topic derived from hundreds of documents with a stated method can be checked while an impression from fifty cannot.
Four nodes, four decisions
The chain shown in the university's own access hand-out runs from file import through text parsing and text filtering to text topic.
Each node decides something. Import fixes what counts as a document, a post alone or a post with its comments. Parsing decides what counts as a term, which is where punctuation, plurals and mixed-language text are handled. Filtering removes terms too rare or too common to separate anything, and it changes the result more than any other step.
The topic node then groups terms that co-occur, and its properties, how many single-term and multi-term topics to learn and whether topics may correlate, are settings a team chooses rather than defaults to accept.
Getting to the software without losing a week
Two routes exist: the tutorial laboratories on campus, or a remote route from a personal laptop through the network client and then the virtual desktop service.
The remote route works and carries a time limit, with the desktop returning to its initial state when a session ends, so work has to be saved outside the session. The hand-out states the limit in two places and the two statements differ, so plan on the shorter and confirm the current figure before a long sitting.
A topic is not yet a finding
The tool outputs terms and counts and supplies no meaning.
Three checks turn a topic into something a report can rest on. Open the heaviest documents and see whether the label survives reading them. Ask what the count is relative to, since the same proportion means different things on pages of different sizes. And check the period, because a topic driven by a single week is an event rather than a standing characteristic.
What this chapter covers
- 01
Collection with FacePager as a required step
- 02
The compulsory analytics modules behind the method
- 03
File import and what counts as one document
- 04
Parsing terms out of text
- 05
Filtering, and why it changes the result most
- 06
Topic settings that a team must choose
- 07
Two access routes and the session time limit
- 08
Reading heaviest documents before labelling a topic
- 09
What a document count does and does not license
Climb from a topic to a supported recommendation
- 3State what the topic alone entitles you to say.
- 3Read the heaviest documents and narrow the claim.
- 3Add an independent source and name what it corroborates.
- 3Name the mechanism and the change that follows.
Key terms
- Business Intelligence
- The practice of turning recorded data into readable evidence for a management decision.
- Text Topic Analysis
- A text mining step that groups terms which co-occur across documents into recurring topics.
- Text Parsing
- The step that turns raw text into countable terms, handling punctuation, word forms and mixed scripts.
- Text Filtering
- The step that removes terms too rare or too common to distinguish documents, and the one that most changes the result.
- Document Count
- The number of collected items loading onto a topic, which supports a claim about prominence within the collection only.
- Virtual Desktop
- A remote session giving off-campus access to laboratory software, time limited and reset when it ends.
- Corpus Shape
- How many documents were collected, over what period and from where, which every topic reading depends on.
Business Intelligence and Text Topic Analysis FAQ
Can I read the comments myself instead of running the analysis?
Qualitative reading is allowed as a complement and not as a substitute, because the brief requires the collection tool and the topic analysis by name. The two also answer different objections: the quantitative step shows that something recurs across a stated body of material, and the qualitative step shows what it means. A report running only one of them reads as either mechanical or anecdotal.
How many posts should I collect?
Enough that a proportion means something and over a period long enough that one campaign cannot dominate it. No number is published, so report the number you obtained and the date range instead of defending a target. If ninety per cent of your material comes from one promotional week, say so and either extend the window or reframe the study as an analysis of that campaign.
What should I do if changing the filter setting changes my topics?
Report the setting you used and say what changed between the settings you tried. Because filtering removes terms, its threshold decides which topics appear at all, so trying several and presenting only the convenient one stops being analysis. Disclosing the sensitivity is a strength in a consultation report rather than an admission.
Exam move
Run the whole chain once on any public page before your team needs it, even with a small collection, so the first attempt under deadline is not also the first attempt ever. Write down the four numbers that describe your corpus, count, period, source and document definition, at the moment of collection.
Then practise the ladder from evidence to claim on somebody else's published analysis: find a sentence in any news article built on social media data and decide which rung it is standing on.
Working through Business Intelligence and Text Topic Analysis in CB2500? Sia is AskSia’s AI Information Technology tutor — ask any CB2500 Business Intelligence and Text Topic Analysis question and get a clear, step-by-step explanation grounded in how CB2500 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
