QInsights Webinar (english language) 5. November 2024 - Analysing semi-structured data

Introduction to Q Insights

Overview of the Webinar

• The webinar focuses on analyzing structured data, particularly from open-ended questions in surveys and social media comments.

• Emphasizes the difference between written responses (like customer reviews) and spoken interview transcripts, which often contain more complex sentence structures.

Data Importing Capabilities

• Q Insights allows users to import various document types, including Excel files for analysis.

• Preparation is necessary for focus group or interview transcripts; respondents' names must be followed by a colon for accurate data processing.

Transcription and Analysis Features

Language Flexibility

• The platform supports multiple languages; it can analyze documents regardless of their original language (e.g., German articles).

Audio Transcription Services

• Users can upload audio files for transcription, which facilitates analysis without needing to edit every detail manually.

• Currently supports short video transcriptions (up to five minutes), providing an alternative for those lacking other transcription options.

Working with Open-ended Questions

Mixed Method Data Integration

• Users can analyze open-ended survey responses alongside quantitative data (e.g., scaling questions about children).

• Metadata from Excel files can be utilized as filters in analyses, allowing targeted insights based on specific demographics or variables.

Importing Interview Transcripts

Speaker Detection Feature

• When importing documents, users must specify if they want speaker detection enabled; this applies to Word documents and PDFs that may not clearly identify speakers.

Processing Imported Documents

• Upon importing, several processes occur: previews are generated, summaries created, and speakers identified. This ensures efficient handling of large datasets.

Future Developments and User Experience

Platform Growth Plans

Interview Analysis and Data Import Techniques

Overview of Interview Summaries

• The speaker discusses the process of conducting interviews, emphasizing that not all variables can be included at once due to limitations in handling multiple file types.

• Automatic summaries are generated for documents, providing a reminder of key points rather than an in-depth analysis. This feature is particularly useful for quickly recalling discussions.

Importing Excel Files

• When importing data from Excel files, each file type must be handled separately; mixed file types cannot be processed simultaneously.

• An example is given involving ratings and comments about a show, highlighting how qualitative data (qual data) can be analyzed effectively.

Analyzing Open-ended Questions

• The speaker mentions analyzing changes over time by comparing comments from different years (e.g., 2013 vs. 2016).

• For open-ended questions, it’s crucial to analyze one Excel file at a time to maintain clarity and focus on specific variables or themes.

Analysis Options Available

• Four main analysis options are available: header and comment analysis, sentiment analysis for structured data, and grid analysis tailored for interviews.

• Sentiment analysis is noted as more effective with structured data rather than interview transcripts due to potential inaccuracies in interpretation.

Future Directions in Data Analysis

• The speaker expresses optimism about future analytical methods that may eliminate the need for coding or tagging data manually.

• Emphasis is placed on asking targeted questions during the analysis phase to derive insights without complex coding processes.

Practical Application Examples

• A study example involves analyzing YouTube comments regarding a new vehicle introduction. This illustrates practical applications of sentiment and thematic analyses.

Analysis of Brand Loyalty and Competitor Insights

Overview of Analysis Structure

• The analysis focuses on competitor insights, future expectations, feedback recommendations, and brand loyalty perceptions.

• A single project file is utilized to contain all analyses for efficiency, avoiding multiple projects for open-ended questions.

• It’s important to archive all chats and analyses systematically to build a comprehensive understanding over time.

Building the Analysis

• Each chat can be archived, allowing for a step-by-step buildup of analysis on various topics like brand loyalty or product features.

• An experiment with AI creativity levels is discussed; higher creativity may lead to less reliable outputs (hallucinations).

• The AI assistant primarily analyzes uploaded data first, minimizing hallucination risks by focusing solely on available information.

Key Concepts in Data Analysis

• Important concepts include excitement about new models, concerns regarding hybrid designs, pricing value comparisons, and brand identity issues.

• Comments can be filtered based on popularity (likes), providing insight into what aspects are deemed significant by others.

Seam Analysis Process

• Running seam analysis should be done judiciously as it incurs costs; unnecessary repetitions should be avoided.

• Communication errors may occur during analysis due to internet connectivity issues rather than program faults.

Detailed Information Retrieval

• Users can start with themes from the data before diving deeper into dialogues for more detailed insights.

Understanding Data Analysis in Excel and Interviews

Theme Occurrence in Documents

• The discussion highlights the importance of analyzing how frequently a theme appears within a single document, particularly in interview data.

• For Excel files, the focus shifts to counting occurrences across multiple rows, indicating a different approach for data analysis.

Challenges with Language and Clarity

• The speaker notes that even if grammar or spelling is not perfect, the language model (LLM) can still comprehend user input.

• Short responses from interviews limit depth; thus, follow-up questions are essential for extracting more detailed insights.

Pricing and Transparency Concerns

• Comments reveal concerns about pricing transparency related to new models, suggesting potential issues that may extend beyond just pricing.

• A significant concern raised is the lack of detailed information regarding certain features, prompting further inquiries into transparency issues.

Data Extraction and User Interaction

• Users can request specific data points or quotes related to transparency issues; however, this feature is not yet available for Excel files.

• In interview-based analyses, users can reference original sources for verification of extracted answers—a feature lacking in current Excel functionalities.

Brand Loyalty Insights

• The dialogue transitions to brand loyalty themes based on saved comments; it’s noted that social media comments may come from repeat contributors.

• An inquiry into strong brand loyalty reveals varied sentiments among different individuals contributing unique perspectives.

Sentiment Analysis Overview

• A summary indicates mixed feelings towards a new product launch; while some express excitement, others show disappointment—highlighting diverse consumer reactions.

• The conversation emphasizes focusing on positive feedback first before delving into areas of disappointment regarding product features.

Understanding Sentiment Analysis and User Interaction

Overview of the Analysis Process

• The analysis provides a summary statement at the end, which can be edited or used as is, offering flexibility in how results are presented.

• Users are encouraged to engage with the summary and recommendations as a starting point for further dialogue and exploration of insights.

Exploring Sentiment Dimensions

• The sentiment analysis allows users to input specific dimensions for evaluation, enhancing the depth of insights gathered.

• Discussion on analyzing performance aspects reveals that different sentiments (like/dislike) can be explored within various contexts.

Performance Insights

• Users can analyze overall sentiment as well as sentiments related to specific aspects, providing a nuanced understanding of feedback.

• A pie chart representation shows mixed sentiments towards the Deep model's performance, indicating predominantly critical views.

Exporting Data and User Feedback

• Currently, pie charts cannot be downloaded directly; however, user feedback is being collected for future improvements in functionality.

• Export options include detailed Excel tables but do not yet support direct image exports; screenshots are suggested instead.

Utilizing Variables in Analysis

• The system recognizes variables from uploaded data (e.g., gender), allowing tailored queries about respondents' perspectives on having children.