│By Sarah L. Ketchley, Sr. Digital Humanities Specialist│
Gale Digital Scholar Lab has introduced a new visualisation feature in the Sentiment Analysis tool: Sentiment by Timeframe. This enables researchers to bring additional depth to sentiment analysis for historical texts. This tool is part of an ongoing effort to expand the capabilities of the Lab’s six digital humanities tools and is designed to support researchers in analysing, interpreting, and visualising data across various historical documents.
The addition of this feature responds to user feedback for the ability to dig deeper into the output from the Sentiment Analysis tool, enabling researchers to perform more nuanced analysis of primary source materials. By allowing users to switch between timeframes easily, Gale Digital Scholar Lab ensures that exploring these insights is both intuitive and efficient.
This blog post will explore how this feature works and why it matters, and as a bonus, will introduce another Lab enhancement: the new markup OCR View. You may also wish to read Understanding Recent Enhancements to Sentiment Analysis in Gale Digital Scholar Lab.
Why Sentiment Analysis is Valuable for Analysing Historical Texts
Sentiment analysis provides a powerful approach to analysing large volumes of historical text by automatically identifying positive, negative, or neutral sentiments. In historical research, where context is critical, sentiment analysis reveals prevailing attitudes, emotions, and perspectives embedded in documents like newspapers, letters, or speeches, offering unique insights into the sentiments of a particular period.
By analysing sentiment over time, researchers can trace changes in public opinion or identify the impact of significant events, such as wars, political movements, or economic crises, on public sentiment. For example, sentiment analysis of wartime newspapers might reveal shifts in optimism or fear as conflicts progress, providing a more nuanced understanding of public morale and opinion than content alone might offer.
By focusing on sentiments expressed in primary sources, sentiment analysis helps reconstruct historical contexts and captures perspectives that might otherwise be difficult to access.
What is Sentiment by Timeframe?
Sentiment by Timeframe is the third visualisation in the Sentiment Analysis tool, providing a unique way to explore sentiment within historical documents by drilling down into specific timeframes, from centuries to months. While previous tools in the Lab, such as Sentiment Over Time, offered high-level sentiment trends across an entire dataset, Sentiment by Timeframe enables users to home in on particular eras or time spans.
This approach allows researchers to better understand the evolution of sentiment within different segments of their content sets, revealing detail that may be missed when looking at aggregated data alone.
How Sentiment by Timeframe Differs from Sentiment Over Time
The Sentiment Over Time visualisation provides an overarching view of sentiment across the entire dataset, offering a valuable first look at general sentiment trends. Sentiment by Timeframe, however, focuses on the details within specific time periods, making it an excellent tool for comparative research or for identifying changes within more concentrated spans of time.
By combining both visualisations, users have a complete toolkit for exploring sentiment data. For example, a researcher might use Sentiment Over Time to gain a high-level understanding of sentiment trends across a century, then turn to Sentiment by Timeframe to look closely at individual decades or years where sentiment shifts significantly. This flexibility supports a wide range of historical research needs, from studying long-term societal changes to investigating specific historical events or periods.
How to Use Sentiment by Timeframe
The new visualisation is accessible on the Analyse page within the Sentiment Analysis tool. When starting a new analysis, users can select the Sentiment by Timeframe option to view their data with this added detail. However, it’s important to note that this visualisation is available only for new runs; existing runs will need to be reinitiated to incorporate this feature.
The tool offers several timespan options: century, decade, year, and month. The availability of these timeframes depends on the metadata within the documents being analysed. For example, if publication months are not specified in the metadata for a given year, the month option will not display, allowing the tool to remain adaptable based on the data’s structure.
The tool offers a variety of export options: you can download the visualisation itself in a variety of image formats, and the statistical data for your analysis at both the content set and document level.
Spotlight on Another New Feature: Markup OCR View
In addition to Sentiment by Timeframe, Gale Digital Scholar Lab recently introduced the markup OCR View, which enhances transparency and usability across the Lab’s text analysis tools, including Parts of Speech, Sentiment Analysis and Named Entity Recognition. This feature allows users to see how individual words have been categorised within a document, adding valuable context to the results generated by the Lab’s tools.
The markup OCR View can be found in the Inspect panel, which can be accessed when viewing any visualisations within these tools. This continuity ensures that researchers can quickly switch between views to cross-reference the original text and the tool-generated insights. These updates align with Gale’s goal of making its tools both comprehensive and user-friendly.
The markup OCR View in Parts of Speech highlights how each word is classified, such as a noun or adjective, giving researchers insights into the linguistic structure of the text. This functionality is particularly beneficial for linguistic research, as users can switch between different word types to examine the document’s construction more closely.
Similarly, the Sentiment Analysis markup indicates which words have been identified as contributing to positive or negative sentiment. This feature not only increases transparency but also bridges the gap between “distant” reading (broad computational analysis of text) and “close” reading (detailed, manual interpretation of text). By giving users a detailed look at how specific words are tagged and interpreted, markup OCR View supports a central aspect of digital humanities research.
Working with markup view in Named Entity Recognition will enable the researcher to develop an understanding of macro and micro-level details of the texts they are working with, categorised according to the entities they wish to work with. This is a powerful way to gain insights into primary sources, opening the door for research insights.
By prioritising both usability and granularity, these tools make it easier than ever to interpret and visualise complex historical data, providing a new dimension of context and detail to digital humanities research.
With these advancements, Gale Digital Scholar Lab continues to be a versatile and powerful tool for historians, linguists, and researchers who seek to bring new perspectives to the study of historical texts.
If you enjoyed reading this blog post, check out others in the ‘Notes from our DH Correspondent’ series, which include:
